Bioinformatics Toolbox™

User's Guide

R2022b
How to Contact MathWorks

Latest news: www.mathworks.com

Sales and services: www.mathworks.com/sales_and_services

User community: www.mathworks.com/matlabcentral

Technical support: www.mathworks.com/support/contact_us

Phone: 508-647-7000

The MathWorks, Inc.

1 Apple Hill Drive
Natick, MA 01760-2098
Bioinformatics Toolbox™ User's Guide
© COPYRIGHT 2003–2022 by The MathWorks, Inc.
The software described in this document is furnished under a license agreement. The software may be used or copied
only under the terms of the license agreement. No part of this manual may be photocopied or reproduced in any form
without prior written consent from The MathWorks, Inc.
FEDERAL ACQUISITION: This provision applies to all acquisitions of the Program and Documentation by, for, or through
the federal government of the United States. By accepting delivery of the Program or Documentation, the government
hereby agrees that this software or documentation qualifies as commercial computer software or commercial computer
software documentation as such terms are used or defined in FAR 12.212, DFARS Part 227.72, and DFARS 252.227-7014.
Accordingly, the terms and conditions of this Agreement and only those rights specified in this Agreement, shall pertain
to and govern the use, modification, reproduction, release, performance, display, and disclosure of the Program and
Documentation by the federal government (or other entity acquiring for or through the federal government) and shall
supersede any conflicting contractual terms or conditions. If this License fails to meet the government's needs or is
inconsistent in any respect with federal procurement law, the government agrees to return the Program and
Documentation, unused, to The MathWorks, Inc.
Trademarks
MATLAB and Simulink are registered trademarks of The MathWorks, Inc. See
www.mathworks.com/trademarks for a list of additional trademarks. Other product or brand names may be
trademarks or registered trademarks of their respective holders.
Patents
MathWorks products are protected by one or more U.S. patents. Please see www.mathworks.com/patents for
more information.
Revision History
September 2003 Online only New for Version 1.0 (Release 13SP1+)
June 2004 Online only Revised for Version 1.1 (Release 14)
November 2004 Online only Revised for Version 2.0 (Release 14SP1+)
March 2005 Online only Revised for Version 2.0.1 (Release 14SP2)
May 2005 Online only Revised for Version 2.1 (Release 14SP2+)
September 2005 Online only Revised for Version 2.1.1 (Release 14SP3)
November 2005 Online only Revised for Version 2.2 (Release 14SP3+)
March 2006 Online only Revised for Version 2.2.1 (Release 2006a)
May 2006 Online only Revised for Version 2.3 (Release 2006a+)
September 2006 Online only Revised for Version 2.4 (Release 2006b)
March 2007 Online only Revised for Version 2.5 (Release 2007a)
April 2007 Online only Revised for Version 2.6 (Release 2007a+)
September 2007 Online only Revised for Version 3.0 (Release 2007b)
March 2008 Online only Revised for Version 3.1 (Release 2008a)
October 2008 Online only Revised for Version 3.2 (Release 2008b)
March 2009 Online only Revised for Version 3.3 (Release 2009a)
September 2009 Online only Revised for Version 3.4 (Release 2009b)
March 2010 Online only Revised for Version 3.5 (Release 2010a)
September 2010 Online only Revised for Version 3.6 (Release 2010b)
April 2011 Online only Revised for Version 3.7 (Release 2011a)
September 2011 Online only Revised for Version 4.0 (Release 2011b)
March 2012 Online only Revised for Version 4.1 (Release 2012a)
September 2012 Online only Revised for Version 4.2 (Release 2012b)
March 2013 Online only Revised for Version 4.3 (Release 2013a)
September 2013 Online only Revised for Version 4.3.1 (Release 2013b)
March 2014 Online only Revised for Version 4.4 (Release 2014a)
October 2014 Online only Revised for Version 4.5 (Release 2014b)
March 2015 Online only Revised for Version 4.5.1 (Release 2015a)
September 2015 Online only Revised for Version 4.5.2 (Release 2015b)
March 2016 Online only Revised for Version 4.6 (Release 2016a)
September 2016 Online only Revised for Version 4.7 (Release 2016b)
March 2017 Online only Revised for Version 4.8 (Release 2017a)
September 2017 Online only Revised for Version 4.9 (Release 2017b)
March 2018 Online only Revised for Version 4.10 (Release 2018a)
September 2018 Online only Revised for Version 4.11 (Release 2018b)
March 2019 Online only Revised for Version 4.12 (Release 2019a)
September 2019 Online only Revised for Version 4.13 (Release 2019b)
March 2020 Online only Revised for Version 4.14 (Release 2020a)
September 2020 Online only Revised for Version 4.15 (Release 2020b)
March 2021 Online only Revised for Version 4.15.1 (Release 2021a)
September 2021 Online only Revised for Version 4.15.2 (Release 2021b)
March 2022 Online only Revised for Version 4.16 (Release 2022a)
September 2022 Online only Revised for Version 4.16.1 (Release 2022b)
 Contents

Getting Started
1
Bioinformatics Toolbox Product Description . . . . . . . . . . . . . . . . . . . . . . . 1-2
Key Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-2

Product Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-3

Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-3
Expected Users . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-4

Data Formats and Databases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-5

Sequence Alignments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-7

Sequence Utilities and Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-8

Protein Property Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-9

Phylogenetic Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-10

Microarray Data Analysis Tools . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-11

Microarray Data Storage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-12

Mass Spectrometry Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-13

Graph Theory Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-15

Statistical Learning and Visualization . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-16

Prototyping and Development Environment . . . . . . . . . . . . . . . . . . . . . . . 1-17

Data Visualization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-18

Exchange Bioinformatics Data Between Excel and MATLAB . . . . . . . . . . 1-19

Using Excel and MATLAB Together . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-19
About the Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-19
Before Running the Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-19
Running the Example for the Entire Data Set . . . . . . . . . . . . . . . . . . . . . 1-20
Editing Formulas to Run the Example on a Subset of the Data . . . . . . . . 1-21
Using the Spreadsheet Link product to Interact With the Data in MATLAB
..................................................... 1-22

Working with Whole Genome Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-25

Comparing Whole Genomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-32

v
 High-Throughput Sequence Analysis
2
Work with Next-Generation Sequencing Data . . . . . . . . . . . . . . . . . . . . . . . 2-2
Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-2
What Files Can You Access? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-2
Before You Begin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-3
Create a BioIndexedFile Object to Access Your Source File . . . . . . . . . . . . 2-3
Determine the Number of Entries Indexed By a BioIndexedFile Object . . . 2-3
Retrieve Entries from Your Source File . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-4
Read Entries from Your Source File . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-4

Manage Sequence Read Data in Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-6

Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-6
Represent Sequence and Quality Data in a BioRead Object . . . . . . . . . . . . 2-7
Represent Sequence, Quality, and Alignment/Mapping Data in a BioMap
Object . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-8
Retrieve Information from a BioRead or BioMap Object . . . . . . . . . . . . . 2-10
Set Information in a BioRead or BioMap Object . . . . . . . . . . . . . . . . . . . 2-12
Determine Coverage of a Reference Sequence . . . . . . . . . . . . . . . . . . . . 2-12
Construct Sequence Alignments to a Reference Sequence . . . . . . . . . . . 2-13
Filter Read Sequences Using SAM Flags . . . . . . . . . . . . . . . . . . . . . . . . . 2-14

Store and Manage Feature Annotations in Objects . . . . . . . . . . . . . . . . . 2-16

Represent Feature Annotations in a GFFAnnotation or GTFAnnotation
Object . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-16
Construct an Annotation Object . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-16
Retrieve General Information from an Annotation Object . . . . . . . . . . . . 2-16
Access Data in an Annotation Object . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-17
Use Feature Annotations with Sequence Read Data . . . . . . . . . . . . . . . . 2-18

Bioinformatics Toolbox Software Support Packages . . . . . . . . . . . . . . . . 2-21

Install Support Package . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-21
Available Support Packages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-21

Count Features from NGS Reads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-23

Identifying Differentially Expressed Genes from RNA-Seq Data . . . . . . . 2-32

Visualize NGS Data Using Genomics Viewer App . . . . . . . . . . . . . . . . . . . 2-52

Open the App . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-52
Add Tracks by Importing Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-52
Visualize Single Nucleotide Variation in Cytochrome P450 . . . . . . . . . . . 2-53

Exploring Genome-Wide Differences in DNA Methylation Profiles . . . . . 2-58

Exploring Protein-DNA Binding Sites from Paired-End ChIP-Seq Data

......................................................... 2-79

Working with Illumina/Solexa Next-Generation Sequencing Data . . . . . 2-97

vi Contents
 Sequence Analysis
3
Exploring a Nucleotide Sequence Using Command Line . . . . . . . . . . . . . . 3-2
Overview of Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-2
Searching the Web for Sequence Information . . . . . . . . . . . . . . . . . . . . . . 3-2
Reading Sequence Information from the Web . . . . . . . . . . . . . . . . . . . . . . 3-4
Determining Nucleotide Composition . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-5
Determining Codon Composition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-8
Open Reading Frames . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-11
Amino Acid Conversion and Composition . . . . . . . . . . . . . . . . . . . . . . . . 3-13

Exploring a Nucleotide Sequence Using the Sequence Viewer App . . . . 3-15

Overview of the Sequence Viewer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-15
Importing a Sequence into the Sequence Viewer . . . . . . . . . . . . . . . . . . 3-15
Viewing Nucleotide Sequence Information . . . . . . . . . . . . . . . . . . . . . . . 3-17
Searching for Words . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-19
Exploring Open Reading Frames . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-22
Closing the Sequence Viewer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-25

Explore a Protein Sequence Using the Sequence Viewer App . . . . . . . . . 3-26

Overview of the Sequence Viewer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-26
Viewing Amino Acid Sequence Statistics . . . . . . . . . . . . . . . . . . . . . . . . . 3-26
Closing the Sequence Viewer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-28
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-29

Compare Sequences Using Sequence Alignment Algorithms . . . . . . . . . 3-30

View and Align Multiple Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-41

Overview of the Sequence Alignment App . . . . . . . . . . . . . . . . . . . . . . . . 3-41
Visualize Multiple Sequence Alignment . . . . . . . . . . . . . . . . . . . . . . . . . . 3-41
Adjust Sequence Alignments Manually . . . . . . . . . . . . . . . . . . . . . . . . . . 3-42
Rearrange Rows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-50
Generate Phylogenetic Tree from Aligned Sequences . . . . . . . . . . . . . . . 3-52

Analyzing Synonymous and Nonsynonymous Substitution Rates . . . . . . 3-55

Investigating the Bird Flu Virus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-65

Exploring Primer Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-81

Identifying Over-Represented Regulatory Motifs . . . . . . . . . . . . . . . . . . . 3-91

Predicting and Visualizing the Secondary Structure of RNA Sequences

........................................................ 3-102

Using HMMs for Profile Analysis of a Protein Family . . . . . . . . . . . . . . 3-114

Predicting Protein Secondary Structure Using a Neural Network . . . . 3-131

Visualizing the Three-Dimensional Structure of a Molecule . . . . . . . . . 3-148

Calculating and Visualizing Sequence Statistics . . . . . . . . . . . . . . . . . . 3-163

vii
 Aligning Pairs of Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-177

Assessing the Significance of an Alignment . . . . . . . . . . . . . . . . . . . . . . 3-185

Using Scoring Matrices to Measure Evolutionary Distance . . . . . . . . . . 3-194

Calling Bioperl Functions from MATLAB . . . . . . . . . . . . . . . . . . . . . . . . 3-198

Accessing NCBI Entrez Databases with E-Utilities . . . . . . . . . . . . . . . . . 3-210

Microarray Analysis
4
Managing Gene Expression Data in Objects . . . . . . . . . . . . . . . . . . . . . . . . 4-2

Representing Expression Data Values in DataMatrix Objects . . . . . . . . . . 4-5

Overview of DataMatrix Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-5
Constructing DataMatrix Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-5
Getting and Setting Properties of a DataMatrix Object . . . . . . . . . . . . . . . 4-6
Accessing Data in DataMatrix Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-6

Representing Expression Data Values in ExptData Objects . . . . . . . . . . . . 4-9

Overview of ExptData Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-9
Constructing ExptData Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-9
Using Properties of an ExptData Object . . . . . . . . . . . . . . . . . . . . . . . . . 4-10
Using Methods of an ExptData Object . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-10
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-11

Representing Sample and Feature Metadata in MetaData Objects . . . . . 4-12

Overview of MetaData Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-12
Constructing MetaData Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-13
Using Properties of a MetaData Object . . . . . . . . . . . . . . . . . . . . . . . . . . 4-15
Using Methods of a MetaData Object . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-15

Representing Experiment Information in a MIAME Object . . . . . . . . . . . 4-16

Overview of MIAME Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-16
Constructing MIAME Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-16
Using Properties of a MIAME Object . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-17
Using Methods of a MIAME Object . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-18

Representing All Data in an ExpressionSet Object . . . . . . . . . . . . . . . . . . 4-19

Overview of ExpressionSet Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-19
Constructing ExpressionSet Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-20
Using Properties of an ExpressionSet Object . . . . . . . . . . . . . . . . . . . . . 4-21
Using Methods of an ExpressionSet Object . . . . . . . . . . . . . . . . . . . . . . . 4-21

Analyzing Illumina Bead Summary Gene Expression Data . . . . . . . . . . . 4-23

Detecting DNA Copy Number Alteration in Array-Based CGH Data . . . . 4-44

Analyzing Array-Based CGH Data Using Bayesian Hidden Markov

Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-60

viii Contents
 Visualizing Microarray Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-74

Gene Expression Profile Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-95

Working with Affymetrix Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-112

Preprocessing Affymetrix Microarray Data at the Probe Level . . . . . . . 4-131

Analyzing Affymetrix SNP Arrays for DNA Copy Number Variants . . . . 4-142

Working with GEO Series Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-162

Identifying Biomolecular Subgroups Using Attractor Metagenes . . . . 4-173

Working with the Clustergram Function . . . . . . . . . . . . . . . . . . . . . . . . . 4-185

Working with Objects for Microarray Experiment Data . . . . . . . . . . . . . 4-203

Phylogenetic Analysis
5
Using the Phylogenetic Tree App . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-2
Overview of the Phylogenetic Tree App . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-2
Opening the Phylogenetic Tree App . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-2
File Menu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-3
Tools Menu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-11
Window Menu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-17
Help Menu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-18

Building a Phylogenetic Tree for the Hominidae Species . . . . . . . . . . . . 5-19

Analyzing the Origin of the Human Immunodeficiency Virus . . . . . . . . . 5-25

Bootstrapping Phylogenetic Trees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-32

Mass Spectrometry and Bioanalytics

6
Preprocessing Raw Mass Spectrometry Data . . . . . . . . . . . . . . . . . . . . . . . 6-2

Visualizing and Preprocessing Hyphenated Mass Spectrometry Data Sets

for Metabolite and Protein/Peptide Profiling . . . . . . . . . . . . . . . . . . . . 6-19

Identifying Significant Features and Classifying Protein Profiles . . . . . . 6-38

Differential Analysis of Complex Protein and Metabolite Mixtures using

Liquid Chromatography/Mass Spectrometry (LC/MS) . . . . . . . . . . . . . 6-52

ix
 Genetic Algorithm Search for Features in Mass Spectrometry Data . . . 6-71

Batch Processing of Spectra Using Sequential and Parallel Computing

......................................................... 6-79

x Contents
 1

Getting Started

• “Bioinformatics Toolbox Product Description” on page 1-2

• “Product Overview” on page 1-3
• “Data Formats and Databases” on page 1-5
• “Sequence Alignments” on page 1-7
• “Sequence Utilities and Statistics” on page 1-8
• “Protein Property Analysis” on page 1-9
• “Phylogenetic Analysis” on page 1-10
• “Microarray Data Analysis Tools” on page 1-11
• “Microarray Data Storage” on page 1-12
• “Mass Spectrometry Data Analysis” on page 1-13
• “Graph Theory Functions” on page 1-15
• “Statistical Learning and Visualization” on page 1-16
• “Prototyping and Development Environment” on page 1-17
• “Data Visualization” on page 1-18
• “Exchange Bioinformatics Data Between Excel and MATLAB” on page 1-19
• “Working with Whole Genome Data” on page 1-25
• “Comparing Whole Genomes” on page 1-32
1 Getting Started

Bioinformatics Toolbox Product Description

Read, analyze, and visualize genomic and proteomic data

Bioinformatics Toolbox provides algorithms and apps for Next Generation Sequencing (NGS),
microarray analysis, mass spectrometry, and gene ontology. Using toolbox functions, you can read
genomic and proteomic data from standard file formats such as SAM, FASTA, CEL, and CDF, as well
as from online databases such as the NCBI Gene Expression Omnibus and GenBank®. You can explore
and visualize this data with sequence browsers, spatial heatmaps, and clustergrams. The toolbox also
provides statistical techniques for detecting peaks, imputing values for missing data, and selecting
features.

You can combine toolbox functions to support common bioinformatics workflows. You can use ChIP-
Seq data to identify transcription factors; analyze RNA-Seq data to identify differentially expressed
genes; identify copy number variants and SNPs in microarray data; and classify protein profiles using
mass spectrometry data.

Key Features
• Next Generation Sequencing analysis and browser
• Sequence analysis and visualization, including pairwise and multiple sequence alignment and
peak detection
• Microarray data analysis, including reading, filtering, normalizing, and visualization
• Mass spectrometry analysis, including preprocessing, classification, and marker identification
• Phylogenetic tree analysis
• Graph theory functions, including interaction maps, hierarchy plots, and pathways
• Data import from genomic, proteomic, and gene expression files, including SAM, FASTA, CEL, and
CDF, and from databases such as NCBI and GenBank

1-2
 Product Overview

Product Overview

Features
The Bioinformatics Toolbox product extends the MATLAB® environment to provide an integrated
software environment for genome and proteome analysis. Scientists and engineers can answer
questions, solve problems, prototype new algorithms, and build applications for drug discovery and
design, genetic engineering, and biological research. An introduction to these features will help you
to develop a conceptual model for working with the toolbox and your biological data.

The Bioinformatics Toolbox product includes many functions to help you with genome and proteome
analysis. Most functions are implemented in the MATLAB programming language, with the source
available for you to view. This open environment lets you explore and customize the existing toolbox
algorithms or develop your own.

You can use the basic bioinformatic functions provided with this toolbox to create more complex
algorithms and applications. These robust and well-tested functions are the functions that you would
otherwise have to create yourself.

Toolbox features and functions fall within these categories:

• Data formats and databases — Connect to Web-accessible databases containing genomic and
proteomic data. Read and convert between multiple data formats.
• High-throughput sequencing — Gene expression and transcription factor analysis of next-
generation sequencing data, including RNA-Seq and ChIP-Seq.
• Sequence analysis — Determine the statistical characteristics of a sequence, align two
sequences, and multiply align several sequences. Model patterns in biological sequences using
hidden Markov model (HMM) profiles.
• Phylogenetic analysis — Create and manipulate phylogenetic tree data.
• Microarray data analysis — Read, normalize, and visualize microarray data.
• Mass spectrometry data analysis — Analyze and enhance raw mass spectrometry data.
• Statistical learning — Classify and identify features in data sets with statistical learning tools.
• Programming interface — Use other bioinformatic software (BioPerl and BioJava) within the
MATLAB environment.

The field of bioinformatics is rapidly growing and will become increasingly important as biology
becomes a more analytical science. The toolbox provides an open environment that you can customize
for development and deployment of the analytical tools you will need.

• Prototype and develop algorithms — Prototype new ideas in an open and extensible
environment. Develop algorithms using efficient string processing and statistical functions, view
the source code for existing functions, and use the code as a template for customizing, improving,
or creating your own functions. See “Prototyping and Development Environment” on page 1-17.
• Visualize data — Visualize sequences and alignments, gene expression data, phylogenetic trees,
mass spectrometry data, protein structure, and relationships between data with interconnected
graphs. See “Data Visualization” on page 1-18.
• Share and deploy applications — Use an interactive GUI builder to develop a custom graphical
front end for your data analysis programs. Create standalone applications that run separately
from the MATLAB environment.

1-3
1 Getting Started

Expected Users
The Bioinformatics Toolbox product is intended for computational biologists and research scientists
who need to develop new algorithms or implement published ones, visualize results, and create
standalone applications.

• Industry/Professional — Increasingly, drug discovery methods are being supported by

engineering practice. This toolbox supports tool builders who want to create applications for the
biotechnology and pharmaceutical industries.
• Education/Professor/Student — This toolbox is well suited for learning and teaching genome
and proteome analysis techniques. Educators and students can concentrate on bioinformatic
algorithms instead of programming basic functions such as reading and writing to files.

While the toolbox includes many bioinformatic functions, it is not intended to be a complete set of
tools for scientists to analyze their biological data. However, the MATLAB environment is ideal for
rapidly designing and prototyping the tools you need.

1-4
 Data Formats and Databases

Data Formats and Databases

The Bioinformatics Toolbox lets you access many of the databases on the web and other online data
repositories. It lets you copy data into the MATLAB workspace, and read and write to files with
standard bioinformatic formats. It also reads many common genome file formats so that you do not
have to write and maintain your own file readers.

Web-based databases — You can directly access public databases on the Web and copy sequence
and gene expression information into the MATLAB environment.

The sequence databases currently supported are GenBank (getgenbank), GenPept (getgenpept),
European Molecular Biology Laboratory (EMBL) (getembl), and Protein Data Bank (PDB) (getpdb).
You can also access data from the NCBI Gene Expression Omnibus (GEO) Web site by using a single
function (getgeodata).

Get multiply aligned sequences (gethmmalignment), hidden Markov model profiles (gethmmprof),
and phylogenetic tree data (gethmmtree) from the PFAM database.

Gene Ontology database — Load the database from the Web into a gene ontology object (geneont).
Select sections of the ontology with methods for the geneont object (getancestors,
getdescendents, getmatrix, getrelatives), and manipulate data with utility functions
(goannotread, num2goid).

Read data from instruments — Read data generated from gene sequencing instruments (scfread,
joinseq, traceplot), mass spectrometers (jcampread), and Agilent® microarray scanners
(agferead).

Reading data formats — The toolbox provides a number of functions for reading data from common
bioinformatic file formats.

• Sequence data: GenBank (genbankread), GenPept (genpeptread), EMBL (emblread), PDB

(pdbread), and FASTA (fastaread)
• Multiply aligned sequences: ClustalW and GCG formats (multialignread)
• Gene expression data from microarrays: Gene Expression Omnibus (GEO) data (geosoftread),
GenePix® data in GPR and GAL files (gprread, galread), SPOT data (sptread), Affymetrix®
GeneChip® data (affyread), and ImaGene® results files (imageneread)
• Hidden Markov model profiles: PFAM-HMM file (pfamhmmread)

Writing data formats — The functions for getting data from the Web include the option to save the
data to a file. However, there is a function to write data to a file using the FASTA format
(fastawrite).

BLAST searches — Request Web-based BLAST searches (blastncbi), get the results from a search
(getblast) and read results from a previously saved BLAST formatted report file (blastread).

The MATLAB environment has built-in support for other industry-standard file formats including
Microsoft® Excel® and comma-separated-value (CSV) files. Additional functions perform ASCII and
low-level binary I/O, allowing you to develop custom functions for working with any data format.

1-5
1 Getting Started

See Also

More About
• “High-Throughput Sequencing”
• “Microarray Analysis”
• “Sequence Analysis”
• “Structural Analysis”
• “Mass Spectrometry and Bioanalytics”

1-6
 Sequence Alignments

Sequence Alignments
You can select from a list of analysis methods to compare nucleotide or amino acid sequences using
pairwise or multiple sequence alignment functions.

Pairwise sequence alignment — Efficient implementations of standard algorithms such as the

Needleman-Wunsch (nwalign) and Smith-Waterman (swalign) algorithms for pairwise sequence
alignment. The toolbox also includes standard scoring matrices such as the PAM and BLOSUM
families of matrices (blosum, dayhoff, gonnet, nuc44, pam). Visualize sequence similarities with
seqdotplot.

Multiple sequence alignment — Functions for multiple sequence alignment (multialign,

profalign) and functions that support multiple sequences (multialignread, fastaread). There
is also a graphical interface (seqalignviewer) for viewing the results of a multiple sequence
alignment and manually making adjustment.

Multiple sequence profiles — Implementations for multiple alignment and profile hidden Markov
model algorithms (gethmmprof, gethmmalignment, gethmmtree, pfamhmmread, hmmprofalign,
hmmprofestimate, hmmprofgenerate, hmmprofmerge, hmmprofstruct, showhmmprof).

Biological codes — Look up the letters or numeric equivalents for commonly used biological codes
(aminolookup, baselookup, geneticcode, revgeneticcode).

See Also

More About
• “Sequence Utilities and Statistics” on page 1-8
• “Sequence Analysis”
• “Data Formats and Databases” on page 1-5

1-7
1 Getting Started

Sequence Utilities and Statistics

You can manipulate and analyze your sequences to gain a deeper understanding of the physical,
chemical, and biological characteristics of your data. Use a graphical user interface (GUI) with many
of the sequence functions in the toolbox (seqviewer).

Sequence conversion and manipulation — The toolbox provides routines for common operations,
such as converting DNA or RNA sequences to amino acid sequences, that are basic to working with
nucleic acid and protein sequences (aa2int, aa2nt, dna2rna, rna2dna, int2aa, int2nt, nt2aa,
nt2int, seqcomplement, seqrcomplement, seqreverse).

You can manipulate your sequence by performing an in silico digestion with restriction endonucleases
(restrict) and proteases (cleave).

Sequence statistics — Determine various statistics about a sequence (aacount, basecount,

codoncount, dimercount, nmercount, ntdensity, codonbias, cpgisland, oligoprop), search
for specific patterns within a sequence (seqwordcount), or search for open reading frames
(seqshoworfs). In addition, you can create random sequences for test cases (randseq).

Sequence utilities — Determine a consensus sequence from a set of multiply aligned amino acid,
nucleotide sequences (seqconsensus, or a sequence profile (seqprofile). Format a sequence for
display (seqdisp) or graphically show a sequence alignment with frequency data (seqlogo).

Additional MATLAB functions efficiently handle string operations with regular expressions (regexp,
seq2regexp) to look for specific patterns in a sequence and search through a library for string
matches (seqmatch).

Look for possible cleavage sites in a DNA/RNA sequence by searching for palindromes
(palindromes).

See Also

More About
• “Sequence Alignments” on page 1-7
• “Sequence Analysis”
• “Protein and Amino Acid Sequence Analysis”
• “Data Formats and Databases” on page 1-5

1-8
 Protein Property Analysis

Protein Property Analysis

You can use a collection of protein analysis methods to extract information from your data. You can
determine protein characteristics and simulate enzyme cleavage reactions. The toolbox provides
functions to calculate various properties of a protein sequence, such as the atomic composition
(atomiccomp), molecular weight (molweight), and isoelectric point (isoelectric). You can cleave
a protein with an enzyme (cleave, rebasecuts) and create distance and Ramachandran plots for
PDB data (pdbdistplot, ramachandran). The toolbox contains a graphical user interface for
protein analysis (proteinplot) and plotting 3-D protein and other molecular structures with
information from molecule model files, such as PDB files (molviewer).

Amino acid sequence utilities — Calculate amino acid statistics for a sequence (aacount) and get
information about character codes (aminolookup).

See Also

More About
• “Protein and Amino Acid Sequence Analysis”
• “Structural Analysis”

1-9
1 Getting Started

Phylogenetic Analysis
Phylogenetic analysis is the process you use to determine the evolutionary relationships between
organisms. The results of an analysis can be drawn in a hierarchical diagram called a cladogram or
phylogram (phylogenetic tree). The branches in a tree are based on the hypothesized evolutionary
relationships (phylogeny) between organisms. Each member in a branch, also known as a
monophyletic group, is assumed to be descended from a common ancestor. Originally, phylogenetic
trees were created using morphology, but now, determining evolutionary relationships includes
matching patterns in nucleic acid and protein sequences. The Bioinformatics Toolbox provides the
following data structure and functions for phylogenetic analysis.

Phylogenetic tree data — Read and write Newick-formatted tree files (phytreeread,
phytreewrite) into the MATLAB Workspace as phylogenetic tree objects (phytree).

Create a phylogenetic tree — Calculate the pairwise distance between biological sequences
(seqpdist), estimate the substitution rates (dnds, dndsml), build a phylogenetic tree from pairwise
distances (seqlinkage, seqneighjoin, reroot), and view the tree in an interactive GUI that
allows you to view, edit, and explore the data (phytreeviewer or view). This GUI also allows you to
prune branches, reorder, rename, and explore distances.

Phylogenetic tree object methods — You can access the functionality of the phytreeviewer user
interface using methods for a phylogenetic tree object (phytree). Get property values (get) and
node names (getbyname). Calculate the patristic distances between pairs of leaf nodes (pdist,
weights) and draw a phylogenetic tree object in a MATLAB Figure window as a phylogram,
cladogram, or radial treeplot (plot). Manipulate tree data by selecting branches and leaves using a
specified criterion (select, subtree) and removing nodes (prune). Compare trees (getcanonical)
and use Newick-formatted strings (getnewickstr).

See Also

More About
• “Sequence Utilities and Statistics” on page 1-8
• “Sequence Analysis”

1-10
 Microarray Data Analysis Tools

Microarray Data Analysis Tools

The MATLAB environment is widely used for microarray data analysis, including reading, filtering,
normalizing, and visualizing microarray data. However, the standard normalization and visualization
tools that scientists use can be difficult to implement. The toolbox includes these standard functions:

Microarray data — Read Affymetrix GeneChip files (affyread) and plot data (probesetplot),
ImaGene results files (imageneread), SPOT files (sptread) and Agilent microarray scanner files
(agferead). Read GenePix GPR files (gprread) and GAL files (galread). Get Gene Expression
Omnibus (GEO) data from the Web (getgeodata) and read GEO data from files (geosoftread).

A utility function (magetfield) extracts data from one of the microarray reader functions (gprread,
agferead, sptread, imageneread).

Microarray normalization and filtering — The toolbox provides a number of methods for
normalizing microarray data, such as lowess normalization (malowess) and mean normalization
(manorm), or across multiple arrays (quantilenorm). You can use filtering functions to clean raw
data before analysis (geneentropyfilter, genelowvalfilter, generangefilter,
genevarfilter), and calculate the range and variance of values (exprprofrange, exprprofvar).

Microarray visualization — The toolbox contains routines for visualizing microarray data. These
routines include spatial plots of microarray data (maimage, redgreencmap), box plots (maboxplot),
loglog plots (maloglog), and intensity-ratio plots (mairplot). You can also view clustered expression
profiles (clustergram, redgreencmap). You can create 2-D scatter plots of principal components
from the microarray data (mapcaplot).

Microarray utility functions — Use the following functions to work with Affymetrix GeneChip data
sets. Get library information for a probe (probelibraryinfo), gene information from a probe set
(probesetlookup), and probe set values from CEL and CDF information (probesetvalues). Show
probe set information from NetAffx™ Analysis Center (probesetlink) and plot probe set values
(probesetplot).

The toolbox accesses statistical routines to perform cluster analysis and to visualize the results, and
you can view your data through statistical visualizations such as dendrograms, classification, and
regression trees.

See Also

More About
• “Microarray Data Storage” on page 1-12
• “Microarray Analysis”

1-11
1 Getting Started

Microarray Data Storage

The Bioinformatics Toolbox includes functions, objects, and methods for creating, storing, and
accessing microarray data.

The object constructor function, DataMatrix, lets you create a DataMatrix object to encapsulate
data and metadata from a microarray experiment. A DataMatrix object stores experimental data in a
matrix, with rows typically corresponding to gene names or probe identifiers, and columns typically
corresponding to sample identifiers. A DataMatrix object also stores metadata, including the gene
names or probe identifiers (as the row names) and sample identifiers (as the column names).

You can reference microarray expression values in a DataMatrix object the same way you reference
data in a MATLAB array, that is, by using linear or logical indexing. Alternately, you can reference this
experimental data by gene (probe) identifiers and sample identifiers. Indexing by these identifiers lets
you quickly and conveniently access subsets of the data without having to maintain additional index
arrays.

Many MATLAB operators and arithmetic functions are available to DataMatrix objects by means of
methods. These methods let you modify, combine, compare, analyze, plot, and access information
from DataMatrix objects. Additionally, you can easily extend the functionality by using general
element-wise functions, dmarrayfun and dmbsxfun, and by manually accessing the properties of a
DataMatrix object.

Note For more information on creating and using DataMatrix objects, see “Representing Expression
Data Values in DataMatrix Objects” on page 4-5.

See Also

More About
• “Microarray Data Analysis Tools” on page 1-11
• “Microarray Analysis”

1-12
 Mass Spectrometry Data Analysis

Mass Spectrometry Data Analysis

The mass spectrometry functions preprocess and classify raw data from SELDI-TOF and MALDI-TOF
spectrometers and use statistical learning functions to identify patterns.

Reading raw data — Load raw mass/charge and ion intensity data from comma-separated-value
(CSV) files, or read a JCAMP-DX-formatted file with mass spectrometry data (jcampread) into the
MATLAB environment.

You can also have data in TXT files and use the importdata function.

Preprocessing raw data — Resample high-resolution data to a lower resolution (msresample)

where the extra data points are not needed. Correct the baseline (msbackadj). Align a spectrum to a
set of reference masses (msalign) and visually verify the alignment (msheatmap). Normalize the
area between spectra for comparing (msnorm), and filter out noise (mslowess and mssgolay).

Spectrum analysis — Load spectra into a GUI (msviewer) for selecting mass peaks and further
analysis.

The following graphic illustrates the roles of the various mass spectrometry functions in the toolbox.

1-13
1 Getting Started

See Also

More About
• “Mass Spectrometry and Bioanalytics”
• “Data Formats and Databases” on page 1-5

1-14
 Graph Theory Functions

Graph Theory Functions

Graph theory functions in the Bioinformatics Toolbox apply basic graph theory algorithms to sparse
matrices. A sparse matrix represents a graph, any nonzero entries in the matrix represent the edges
of the graph, and the values of these entries represent the associated weight (cost, distance, length,
or capacity) of the edge. Graph algorithms that use the weight information will cancel the edge if a
NaN or an Inf is found. Graph algorithms that do not use the weight information will consider the
edge if a NaN or an Inf is found, because these algorithms look only at the connectivity described by
the sparse matrix and not at the values stored in the sparse matrix.

Sparse matrices can represent four types of graphs:

• Directed Graph — Sparse matrix, either double real or logical. Row (column) index indicates the
source (target) of the edge. Self-loops (values in the diagonal) are allowed, although most of the
algorithms ignore these values.
• Undirected Graph — Lower triangle of a sparse matrix, either double real or logical. An
algorithm expecting an undirected graph ignores values stored in the upper triangle of the sparse
matrix and values in the diagonal.
• Direct Acyclic Graph (DAG) — Sparse matrix, double real or logical, with zero values in the
diagonal. While a zero-valued diagonal is a requirement of a DAG, it does not guarantee a DAG. An
algorithm expecting a DAG will not test for cycles because this will add unwanted complexity.
• Spanning Tree — Undirected graph with no cycles and with one connected component.

There are no attributes attached to the graphs; sparse matrices representing all four types of graphs
can be passed to any graph algorithm. All functions will return an error on nonsquare sparse
matrices.

Graph algorithms do not pretest for graph properties because such tests can introduce a time penalty.
For example, there is an efficient shortest path algorithm for DAG, however testing if a graph is
acyclic is expensive compared to the algorithm. Therefore, it is important to select a graph theory
function and properties appropriate for the type of the graph represented by your input matrix. If the
algorithm receives a graph type that differs from what it expects, it will either:

• Return an error when it reaches an inconsistency. For example, if you pass a cyclic graph to the
graphshortestpath function and specify Acyclic as the method property.
• Produce an invalid result. For example, if you pass a directed graph to a function with an
algorithm that expects an undirected graph, it will ignore values in the upper triangle of the
sparse matrix.

See Also
graphallshortestpaths | graphconncomp | graphshortestpath | graphisdag

1-15
1 Getting Started

Statistical Learning and Visualization

You can classify and identify features in data sets, set up cross-validation experiments, and compare
different classification methods.

The toolbox provides functions that build on the classification and statistical learning tools in the
Statistics and Machine Learning Toolbox™ software (classify, kmeans, fitctree, and
fitrtree).

These functions include imputation tools (knnimpute), and K-nearest neighbor classifiers (fitcknn).

Other functions include set up of cross-validation experiments (crossvalind) and comparison of the
performance of different classification methods (classperf). In addition, there are tools for
selecting diversity and discriminating features (rankfeatures, randfeatures).

1-16
 Prototyping and Development Environment

Prototyping and Development Environment

The MATLAB environment lets you prototype and develop algorithms and easily compare alternatives.

• Integrated environment — Explore biological data in an environment that integrates

programming and visualization. Create reports and plots with the built-in functions for
mathematics, graphics, and statistics.
• Open environment — Access the source code for the toolbox functions. The toolbox includes
many of the basic bioinformatics functions you will need to use, and it includes prototypes for
some of the more advanced functions. Modify these functions to create your own custom solutions.
• Interactive programming language — Test your ideas by typing functions that are interpreted
interactively with a language whose basic data element is an array. The arrays do not require
dimensioning and allow you to solve many technical computing problems,

Using matrices for sequences or groups of sequences allows you to work efficiently and not worry
about writing loops or other programming controls.
• Programming tools — Use a visual debugger for algorithm development and refinement and an
algorithm performance profiler to accelerate development.

1-17
1 Getting Started

Data Visualization
You can visually compare pairwise sequence alignments, multiply aligned sequences, gene expression
data from microarrays, and plot nucleic acid and protein characteristics. The 2-D and volume
visualization features let you create custom graphical representations of multidimensional data sets.
You can also create montages and overlays, and export finished graphics to an Adobe® PostScript®
image file or copy directly into Microsoft PowerPoint®.

1-18
 Exchange Bioinformatics Data Between Excel and MATLAB

Exchange Bioinformatics Data Between Excel and MATLAB

In this section...
“Using Excel and MATLAB Together” on page 1-19
“About the Example” on page 1-19
“Before Running the Example” on page 1-19
“Running the Example for the Entire Data Set” on page 1-20
“Editing Formulas to Run the Example on a Subset of the Data” on page 1-21
“Using the Spreadsheet Link product to Interact With the Data in MATLAB” on page 1-22

Using Excel and MATLAB Together

If you have bioinformatics data in an Excel (2007 or newer) spreadsheet, use Spreadsheet Link to:

• Connect Excel with the MATLAB Workspace to exchange data

• Use MATLAB and Bioinformatics Toolbox computational and visualization functions

About the Example

Note The following example assumes you have Spreadsheet Link software installed on your system.

The Excel file used in the following example contains data from DeRisi, J.L., Iyer, V.R., and Brown, P.O.
(Oct. 24, 1997). Exploring the metabolic and genetic control of gene expression on a genomic scale.
Science 278(5338), 680–686. PMID: 9381177. The data was filtered using the steps described in
“Gene Expression Profile Analysis” on page 4-95.

Before Running the Example

1 If not already done, modify your system path to include the MATLAB root folder as described in
the Spreadsheet Link documentation.
2 If not already done, enable the Spreadsheet Link Add-In as described in “Add-In Setup”
(Spreadsheet Link).
3 Close MATLAB and Excel if they are open.
4 Start Excel. MATLAB and Spreadsheet Link software automatically start.
5 From Excel, open the following file: matlabroot\examples\bioinfo\data
\Filtered_Yeastdata.xlsm

Note matlabroot is the MATLAB root folder, which is where MATLAB software is installed on
your system.
6 In the Excel software, enable macros. Click the Developer tab, and then select Macro Security
from the Code group. If the Developer tab is not displayed on the Excel ribbon, consult Excel
Help to display it. If you encounter the "Can't find project or library" error, you might need to
update the references in the Visual Basic software. Open Visual Basic by clicking the Developer

1-19
1 Getting Started

tab and selecting Visual Basic. Then select Tools > References > SpreadsheetLink. If the
MISSING: exclink2007.xlam check box is selected, clear it.

Running the Example for the Entire Data Set

1 In the provided Excel file, note that columns A through H contain data from DeRisi et al. Also
note that cells J5, J6, J7, and J12 contain formulas using Spreadsheet Link functions
MLPutMatrix and MLEvalString.

Tip To view a cell's formula, select the cell, and then view the formula in the formula bar
at the top of the Excel window.
2 Execute the formulas in cells J5, J6, J7, and J12, by selecting the cell, pressing F2, and then
pressing Enter.

Each of the first three cells contains a formula using the Spreadsheet Link function
MLPutMatrix, which creates a MATLAB variable from the data in the spreadsheet. Cell J12
contains a formula using the Spreadsheet Link function MLEvalString, which runs the
Bioinformatics Toolbox clustergram function using the three variables as input. For more
information on adding formulas using Spreadsheet Link functions, see “Create Diagonal Matrix
Using Worksheet Cells” (Spreadsheet Link).

1-20
 Exchange Bioinformatics Data Between Excel and MATLAB

3 Note that cell J17 contains a formula using a macro function Clustergram, which was created in
the Visual Basic® Editor. Running this macro does the same as the formulas in cells J5, J6, J7, and
J12. Optionally, view the Clustergram macro function by clicking the Developer tab, and then

clicking the Visual Basic button . (If the Developer tab is not on the Excel ribbon, consult
Excel Help to display it.)

For more information on creating macros using Visual Basic Editor, see “Create Diagonal Matrix
Using VBA Macro” (Spreadsheet Link).
4 Execute the formula in cell J17 to analyze and visualize the data:

a Select cell J17.

b Press F2.
c Press Enter.

The macro function Clustergram runs creating three MATLAB variables (data, Genes, and
TimeSteps) and displaying a Clustergram window containing dendrograms and a heat map of
the data.

Editing Formulas to Run the Example on a Subset of the Data

1 Edit the formulas in cells J5 and J6 to analyze a subset of the data. Do this by editing the
formulas’ cell ranges to include data for only the first 30 genes:

1-21
1 Getting Started

a Select cell J5, and then press F2 to display the formula for editing. Change H617 to H33,
and then press Enter.

b Select cell J6, then press F2 to display the formula for editing. Change A617 to A33, and
then press Enter.

2 Run the formulas in cells J5, J6, J7, and J12 to analyze and visualize a subset of the data:

a Select cell J5, press F2, and then press Enter.

b Select cell J6, press F2, and then press Enter.
c Select cell J7, press F2, and then press Enter.
d Select cell J12, press F2, and then press Enter.

Using the Spreadsheet Link product to Interact With the Data in

MATLAB
Use the MATLAB group on the right side of the Home tab to interact with the data:

1-22
 Exchange Bioinformatics Data Between Excel and MATLAB

For example, create a variable in MATLAB containing a 3-by-7 matrix of the data, plot the data in a
Figure window, and then add the plot to your spreadsheet:

1 Click-drag to select cells B5 through H7.

2 From the MATLAB group, select Send data to MATLAB.

3 Type YAGenes for the variable name, and then click OK.

The variable YAGenes is added to the MATLAB Workspace as a 3-by-7 matrix.

4 From the MATLAB group, select Run MATLAB command.
5 Type plot(YAGenes') for the command, and then click OK.

A Figure window displays a plot of the data.

Note Make sure you use the ' (transpose) symbol when plotting the data in this step. You need
to transpose the data in YAGenes so that it plots as three genes over seven time intervals.
6 Select cell J20, and then click from the MATLAB group, select Get MATLAB figure.

The figure is added to the spreadsheet.

1-23
1 Getting Started

1-24
 Working with Whole Genome Data

Working with Whole Genome Data

This example shows how to create a memory mapped file for sequence data and work with it without
loading all the genomic sequence into memory. Whole genomes are available for human, mouse, rat,
fugu, and several other model organisms. For many of these organisms one chromosome can be
several hundred million base pairs long. Working with such large data sets can be challenging as you
may run into limitations of the hardware and software that you are using. This example shows one
way to work around these limitations in MATLAB®.

Large Data Set Handling Issues

Solving technical computing problems that require processing and analyzing large amounts of data
puts a high demand on your computer system. Large data sets take up significant memory during
processing and can require many operations to compute a solution. It can also take a long time to
access information from large data files.

Computer systems, however, have limited memory and finite CPU speed. Available resources vary by
processor and operating system, the latter of which also consumes resources. For example:

32-bit processors and operating systems can address up to 2^32 = 4,294,967,296 = 4 GB of memory
(also known as virtual address space). Windows® XP and Windows® 2000 allocate only 2 GB of this
virtual memory to each process (such as MATLAB). On UNIX®, the virtual memory allocated to a
process is system-configurable and is typically around 3 GB. The application carrying out the
calculation, such as MATLAB, can require storage in addition to the user task. The main problem
when handling large amounts of data is that the memory requirements of the program can exceed
that available on the platform. For example, MATLAB generates an "out of memory" error when data
requirements exceed approximately 1.7 GB on Windows XP.

For more details on memory management and large data sets, see “Performance and Memory”.

On a typical 32-bit machine, the maximum size of a single data set that you can work with in MATLAB
is a few hundred MB, or about the size of a large chromosome. Memory mapping of files allows
MATLAB to work around this limitation and enables you to work with very large data sets in an
intuitive way.

Whole Genome Data Sets

The latest whole genome data sets can be downloaded from the Ensembl Website. The data are
provided in several formats. These are updated regularly as new sequence information becomes
available. This example will use human DNA data stored in FASTA format. Chromosome 1 is (in the
GRCh37.56 Release of September 2009) a 65.6 MB compressed file. After uncompressing the file it is
about 250MB. MATLAB uses 2 bytes per character, so if you read the file into MATLAB, it will require
about 500MB of memory.

This example assumes that you have already downloaded and uncompressed the FASTA file into your
local directory. Change the name of the variable FASTAfilename if appropriate.

FASTAfilename = 'Homo_sapiens.GRCh37.56.dna.chromosome.1.fa';
fileInfo = dir(which(FASTAfilename))

fileInfo =

1-25
1 Getting Started

struct with fields:

name: 'Homo_sapiens.GRCh37.56.dna.chromosome.1.fa'
folder: 'I:\qe\test_data\Bioinformatics_Toolbox\v000\demoData\biomemorymapdemo'
date: '01-Feb-2013 11:54:41'
bytes: 253404851
isdir: 0
datenum: 7.3527e+05

Memory Mapped Files

Memory mapping allows MATLAB to access data in a file as though it is in memory. You can use
standard MATLAB indexing operations to access data. See the documentation for memmapfile for
more details.

You could just map the FASTA file and access the data directly from there. However the FASTA format
file includes new line characters. The memmapfile function treats these characters in the same way
as all other characters. Removing these before memory mapping the file will make indexing
operations simpler. Also, memory mapping does not work directly with character data so you will
have to treat the data as 8-bit integers (uint8 class). The function nt2int in the Bioinformatics
Toolbox™ can be used to convert character information into integer values. int2nt is used to
convert back to characters.

First open the FASTA file and extract the header.

fidIn = fopen(FASTAfilename,'r');
header = fgetl(fidIn)

header =

'>1 dna:chromosome chromosome:GRCh37:1:1:249250621:1'

Open the file to be memory mapped.

[fullPath, filename, extension] = fileparts(FASTAfilename);
mmFilename = [filename '.mm']
fidOut = fopen(mmFilename,'w');

mmFilename =

'Homo_sapiens.GRCh37.56.dna.chromosome.1.mm'

Read the FASTA file in blocks of 1MB, remove new line characters, convert to uint8, and write to the
MM file.
newLine = sprintf('\n');
blockSize = 2^20;
while ~feof(fidIn)
% Read in the data
charData = fread(fidIn,blockSize,'*char')';
% Remove new lines
charData = strrep(charData,newLine,'');
% Convert to integers

1-26
 Working with Whole Genome Data

intData = nt2int(charData);
% Write to the new file
fwrite(fidOut,intData,'uint8');
end

Close the files.

fclose(fidIn);
fclose(fidOut);

The new file is about the same size as the old file but does not contain new lines or the header
information.

mmfileInfo = dir(mmFilename)

mmfileInfo =

struct with fields:

name: 'Homo_sapiens.GRCh37.56.dna.chromosome.1.mm'
folder: 'C:\TEMP\Bdoc22b_2134332_7304\ibD166FA\17\tp9f51422f\bioinfo-ex57563178'
date: '26-Nov-2022 10:52:13'
bytes: 249250621
isdir: 0
datenum: 7.3885e+05

Accessing the Data in the Memory Mapped File

The command memmapfile constructs a memmapfile object that maps the new file to memory. In
order to do this, it needs to know the format of the file. The format of this file is simple, though much
more complicated formats can be mapped.

chr1 = memmapfile(mmFilename, 'format', 'uint8')

chr1 =

Filename: 'C:\TEMP\Bdoc22b_2134332_7304\ibD166FA\17\tp9f51422f\bioinfo-ex57563178\Homo_sapien
Writable: false
Offset: 0
Format: 'uint8'
Repeat: Inf
Data: 249250621x1 uint8 array

The MEMMAPFILE Object

The memmapfile object has various properties. Filename stores the full path to the file. Writable
indicates whether or not the data can be modified. Note that if you do modify the data, this will also
modify the original file. Offset allows you to specify the space used by any header information.
Format indicates the data format. Repeat is used to specify how many blocks (as defined by
Format) to map. This can be useful for limiting how much memory is used to create the memory map.
These properties can be accessed in the same way as other MATLAB data. For more details see type
help memmapfile or doc memmapfile.

chr1.Data(1:10)

1-27
1 Getting Started

ans =

10x1 uint8 column vector

15
15
15
15
15
15
15
15
15
15

You can access any region of the data using indexing operations.
chr1.Data(10000000:10000010)'

ans =

1x11 uint8 row vector

1 1 2 2 2 2 3 4 2 4 2

Remember that the nucleotide information was converted to integers. You can use int2nt to get the
sequence information back.
int2nt(chr1.Data(10000000:10000010)')

ans =

'AACCCCGTCTC'

Or use seqdisp to display the sequence.

seqdisp(chr1.Data(10000000:10001000)')

ans =

17x71 char array

' 1 AACCCCGTCT CTACAATAAA TTAAAATATT AGCTGGGCAT GGTGGTGTGT GCTTGTAGTC'

' 61 CCAGCTACTT GGCGGGCTGA GGTGGGAGAA TCATCCAAGC CTTGGAGGCA GAGGTTGCAG'
' 121 TGAGCTGAGA TTGTGACACT GCACTCCAGC CTGGGAGACA GAGTGAGACT CCTACTCAAA'
' 181 AAAAAACAAA AAACAAAAAA CAAACCACAA AACTTTCCAG GTAACTTATT AAAACATGTT'
' 241 TTTTGTTTGT TTTGAGACAG AGTCTTGCTC TGTCGCCCAG GCTGGAGTGC AGTGGAGCAA'
' 301 TCTCAGCTCA CTGCAAGCTC CGCCTCCCGG GTTCACACCA TTCTCCTGCC TCAGCCTCCC'
' 361 GAGTAGCTAG GACTATAGGC ACCCGCCACC ACGCCCAGCT TATTTTTTTT GTATTTTTTA'
' 421 GTAGAGACGG GGTTTCATCG TGTTAGCCAG GATGGTCTCG ATCTCCTGAC CTCGTGATCC'
' 481 GCCCACCTCA GCCTCCCAAA GTGCTGGGAT TACAGGCGTG AGCCACTGCA CCCGGCCTAG'
' 541 TTTTTGTATA TTTTTTTTAG TAGAGACAGG GTTTCACCAT GTTAGCCAGG ATGGTCTCAA'

1-28
 Working with Whole Genome Data

' 601 TCTCCTGACC TCGTGATCCG CCCGCCTCGG CCTCCCAAAG TGCTGGGGTT ACAGGCGTGA'

' 661 GCCACCGCAC ACAGCATTAA AGCATGTTTT ATTTTCCTAC ACATAATGAA ATCATTACCA'
' 721 GATGATTTGA CATGTGTACT TCATTGGAGA GGATTCTTAC AGTATATTCA AAATTAAATA'
' 781 TAATGACAAA AAATTACTAC CTAATCTATT AAAATTGGCA TAAGTCATCT ATGATCATTA'
' 841 ATGATATGCA AACATAAACA AGTATTATAC CCAGAAGTGT AATTTATTGT AGCTACATCT'
' 901 TATGTATAAT AGTTTAGTGG ATTTTTCCTG GAAATTGTCC ATTTTAATTT TTCTCTTAAG'
' 961 TCTGTGGAAT TTTCCAGTAA AAGTCAAGGC AAACCCAAGA T '

Analysis of the Whole Chromosome

Now that you can easily access the whole chromosome, you can analyze the data. This example shows
one way to look at the GC content along the chromosome.

You extract blocks of 500000bp and calculate the GC content.

Calculate how many blocks to use.

numNT = numel(chr1.Data);
blockSize = 500000;
numBlocks = floor(numNT/blockSize);

One way to help MATLAB performance when working with large data sets is to "preallocate" space
for data. This allows MATLAB to allocate enough space for all of the data rather than having to grow
the array in small chunks. This will speed things up and also protect you from problems of the data
getting too large to store. For more details on pre-allocating arrays, see: https://www.mathworks.com/
matlabcentral/answers/99124-how-do-i-pre-allocate-memory-when-using-matlab

An easy way to preallocate an array is to use the zeros function.

ratio = zeros(numBlocks+1,1);

Loop over the data looking for C or G and then divide this number by the total number of A, T, C, and
G. This will take about a minute to run.

A = nt2int('A'); C = nt2int('C'); G = nt2int('G'); T = nt2int('T');

for count = 1:numBlocks

% calculate the indices for the block
start = 1 + blockSize*(count-1);
stop = blockSize*count;
% extract the block
block = chr1.Data(start:stop);
% find the GC and AT content
gc = (sum(block == G | block == C));
at = (sum(block == A | block == T));
% calculate the ratio of GC to the total known nucleotides
ratio(count) = gc/(gc+at);
end

The final block is smaller so treat this as a special case.

block = chr1.Data(stop+1:end);
gc = (sum(block == G | block == C));
at = (sum(block == A | block == T));
ratio(end) = gc/(gc+at);

1-29
1 Getting Started

Plot of the GC Content for the Homo Sapiens Chromosome 1

xAxis = [1:blockSize:numBlocks*blockSize, numNT];

plot(xAxis,ratio)
xlabel('Base pairs');
ylabel('Relative GC content');
title('Relative GC content of Homo Sapiens Chromosome 1')

The region in the center of the plot around 140Mbp is a large region of Ns.

seqdisp(chr1.Data(140000000:140001000))

ans =

17x71 char array

' 1 NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN'

' 61 NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN'
' 121 NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN'
' 181 NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN'
' 241 NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN'
' 301 NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN'
' 361 NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN'
' 421 NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN'
' 481 NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN'
' 541 NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN'
' 601 NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN'

1-30
 Working with Whole Genome Data

' 661 NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN'

' 721 NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN'
' 781 NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN'
' 841 NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN'
' 901 NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN'
' 961 NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN N '

Finding Regions of High GC Content

You can use find to identify regions of high GC content.

indices = find(ratio>0.5);
ranges = [(1 + blockSize*(indices-1)), blockSize*indices];
fprintf('Region %d:%d has GC content %f\n',[ranges ,ratio(indices)]')

Region 500001:1000000 has GC content 0.501412

Region 1000001:1500000 has GC content 0.598332
Region 1500001:2000000 has GC content 0.539498
Region 2000001:2500000 has GC content 0.594508
Region 2500001:3000000 has GC content 0.568620
Region 3000001:3500000 has GC content 0.584572
Region 3500001:4000000 has GC content 0.548137
Region 6000001:6500000 has GC content 0.545072
Region 9000001:9500000 has GC content 0.506692
Region 10500001:11000000 has GC content 0.511386
Region 11500001:12000000 has GC content 0.519874
Region 16000001:16500000 has GC content 0.513082
Region 17500001:18000000 has GC content 0.513392
Region 21500001:22000000 has GC content 0.505598
Region 156000001:156500000 has GC content 0.504446
Region 156500001:157000000 has GC content 0.504090
Region 201000001:201500000 has GC content 0.502976
Region 228000001:228500000 has GC content 0.511946

If you want to remove the temporary file, you must first clear the memmapfile object.

clear chr1
delete(mmFilename)

1-31
1 Getting Started

Comparing Whole Genomes

This example shows how to compare whole genomes for organisms, which allows you to compare the
organisms at a very different resolution relative to single gene comparisons. Instead of just focusing
on the differences between homologous genes you can gain insight into the large-scale features of
genomic evolution.

This example uses two strains of Chlamydia, Chlamydia trachomatis and Chlamydophila pneumoniae.
These are closely related bacteria that cause different, though both very common, diseases in
humans. Whole genomes are available in the GenBank® database for both organisms.

Retrieving the Genomes

You can download these genomes using the getgenbank function. First, download the Chlamydia
trachomatis genome. Notice that the genome is circular and just over one million bp in length. These
sequences are quite large so may take a while to download.

seqtrachomatis = getgenbank('NC_000117');

Next, download Chlamydophila pneumoniae. This genome is also circular and a little longer at 1.2
Mbp.

seqpneumoniae = getgenbank('NC_002179');

For your convenience, previously downloaded sequences are included in a MAT-file. Note that data in
public repositories is frequently curated and updated. Hence, the results of this example might be
slightly different when you use up-to-date datasets.

load('chlamydia.mat','seqtrachomatis','seqpneumoniae')

A very simple approach for comparing the two genomes is to perform pairwise alignment between all
genes in the genomes. Given that these are bacterial genomes, a simple approach would be to
compare all ORFs in the two genomes. However, the GenBank data includes more information about
the genes in the sequences. This is stored in the CDS field of the data structure. Chlamydia
trachomatis has 895 coding regions, while Chlamydophila pneumoniae has 1112.

M = numel(seqtrachomatis.CDS)
N = numel(seqpneumoniae.CDS)

M =

895

N =

1112

Most of the CDS records contain the translation to amino acid sequences. The first CDS record in the
Chlamydia trachomatis data is a hypothetical protein of length 591 residues.

seqtrachomatis.CDS(1)

1-32
 Comparing Whole Genomes

ans =

struct with fields:

location: 'join(1041920..1042519,1..1176)'
gene: []
product: 'hypothetical protein'
codon_start: '1'
indices: [1041920 1042519 1 1176]
protein_id: 'NP_219502.1'
db_xref: 'GeneID:884145'
note: []
translation: 'MSIRGVGGNGNSRIPSHNGDGSNRRSQNTKGNNKVEDRVCSLYSSRSNENRESPYAVVDVSSMIESTPTSGETTRASRG
text: [19x58 char]

The fourth CDS record is for the gatA gene, which has product glutamyl-tRNA amidotransferase
subunit A. The length of the product sequence is 491 residues.

seqtrachomatis.CDS(4)

ans =

struct with fields:

location: '2108..3583'
gene: 'gatA'
product: [2x47 char]
codon_start: '1'
indices: [2108 3583]
protein_id: 'NP_219505.1'
db_xref: 'GeneID:884087'
note: [7x58 char]
translation: 'MYRKSALELRDAVVNRELSVTAITEYFYHRIESHDEQIGAFLSLCKERALLRASRIDDKLAKGDPIGLLAGIPIGVKDN
text: [26x58 char]

A few of the Chlamydophila pneumoniae CDS have empty translations. Fill them in as follows. First,
find all empty translations, then display the first empty translation.

missingPn = find(cellfun(@isempty,{seqpneumoniae.CDS.translation}));
seqpneumoniae.CDS(missingPn(1))

ans =

struct with fields:

location: 'complement(73364..73477)'
gene: []
product: 'hypothetical protein'
codon_start: '1'
indices: [73477 73364]
protein_id: 'NP_444613.1'
db_xref: 'GeneID:963699'
note: 'hypothetical protein; identified by Glimmer2'

1-33
1 Getting Started

translation: []
text: [10x52 char]

The function featureparse extracts features, such as the CDS, from the sequence structure. You
can then use cellfun to apply nt2aa to the sequences with missing translations.

allCDS = featureparse(seqpneumoniae,'Feature','CDS','Sequence',true);
missingSeqs = cellfun(@nt2aa,{allCDS(missingPn).Sequence},'uniform',false);
[seqpneumoniae.CDS(missingPn).translation] = deal(missingSeqs{:});
seqpneumoniae.CDS(missingPn(1))

ans =

struct with fields:

location: 'complement(73364..73477)'
gene: []
product: 'hypothetical protein'
codon_start: '1'
indices: [73477 73364]
protein_id: 'NP_444613.1'
db_xref: 'GeneID:963699'
note: 'hypothetical protein; identified by Glimmer2'
translation: 'MLTDQRKHIQMLHKHNSIEIFLSNMVVEVKLFFKTLK*'
text: [10x52 char]

Performing Gene Comparisons

To compare the gatA gene in Chlamydia trachomatis with all the CDS genes in Chlamydophila
pneumoniae, put a for loop around the nwalign function. You could alternatively use local
alignment (swalign).

tic
gatAScores = zeros(1,N);
for inner = 1:N
gatAScores(inner) = nwalign(seqtrachomatis.CDS(4).translation,...
seqpneumoniae.CDS(inner).translation);
end
toc % |tic| and |toc| are used to report how long the calculation takes.

Elapsed time is 3.185265 seconds.

A histogram of the scores shows a large number of negative scores and one very high positive score.

hist(gatAScores,100)
title(sprintf(['Alignment Scores for Chlamydia trachomatis %s\n',...
'with all CDS in Chlamydophila pneumoniae'],seqtrachomatis.CDS(4).gene))

1-34
 Comparing Whole Genomes

As expected, the high scoring match is with the gatA gene in Chlamydophila pneumoniae.

[gatABest, gatABestIdx] = max(gatAScores);

seqpneumoniae.CDS(gatABestIdx)

ans =

struct with fields:

location: 'complement(838828..840306)'
gene: 'gatA'
product: [2x47 char]
codon_start: '1'
indices: [840306 838828]
protein_id: 'NP_445311.1'
db_xref: 'GeneID:963139'
note: [7x58 char]
translation: 'MYRYSALELAKAVTLGELTATGVTQHFFHRIEEAEGQVGAFISLCKEQALEQAELIDKKRSRGEPLGKLAGVPVGIKDN
text: [26x58 char]

The pairwise alignment of one gene from Chlamydia trachomatis with all genes from Chlamydophila
pneumoniae takes just under a minute on an Intel® Pentium 4, 2.0 GHz machine running Windows®
XP. To do this calculation for all 895 CDS in Chlamydia trachomatis would take about 12 hours on the
same machine. Uncomment the following code if you want to run the whole calculation.

1-35
1 Getting Started

scores = zeros(M,N);
parfor outer = 1:M
theScore = zeros(1,outer);
theSeq = seqtrachomatis.CDS(outer).translation;
for inner = 1:N
theScore(inner) = ...
nwalign(theSeq,...
seqpneumoniae.CDS(inner).translation);
end
scores(outer,:) = theScore;
end

Note the command parfor is used in the outer loop. If your machine is configured to run multiple
labs then the outer loop will be executed in parallel. For a full understanding of this construct, see
doc parfor.

Investigating the Meaning of the Scores

The distributions of the scores for several genes show a pattern. The CDS(3) of Chlamydia
trachomatis is the gatC gene. This has a relatively short product,aspartyl/glutamyl-tRNA
amidotransferase subunit C, with only 100 residues.

gatCScores = zeros(1,N);
for inner = 1:N
gatCScores(inner) = nwalign(seqtrachomatis.CDS(3).translation,...
seqpneumoniae.CDS(inner).translation);
end
figure
hist(gatCScores,100)
title(sprintf(['Alignment Scores for Chlamydia trachomatis %s\n',...
'with all CDS in Chlamydophila pneumoniae'],seqtrachomatis.CDS(3).gene))
xlabel('Score');ylabel('Number of genes');

1-36
 Comparing Whole Genomes

The best score again corresponds to the same gene in the Chlamydophila pneumoniae.
[gatCBest, gatCBestIdx] = max(gatCScores);
seqpneumoniae.CDS(gatCBestIdx).product

ans =

2x47 char array

'aspartyl/glutamyl-tRNA amidotransferase subunit'

'C '

CDS(339) of Chlamydia trachomatis is the uvrA gene. This has a very long product, excinuclease ABC
subunit A, of length 1786.
uvrAScores = zeros(1,N);
for inner = 1:N
uvrAScores(inner) = nwalign(seqtrachomatis.CDS(339).translation,...
seqpneumoniae.CDS(inner).translation);
end
figure
hist(uvrAScores,100)
title(sprintf(['Alignment Scores for Chlamydia trachomatis %s\n',...
'with all CDS in Chlamydophila pneumoniae'],seqtrachomatis.CDS(339).gene))
xlabel('Score');ylabel('Number of genes');

1-37
1 Getting Started

[uvrABest, uvrABestIdx] = max(uvrAScores);

seqpneumoniae.CDS(uvrABestIdx)

ans =

struct with fields:

location: '716887..722367'
gene: []
product: 'excinuclease ABC subunit A'
codon_start: '1'
indices: [716887 722367]
protein_id: 'NP_445220.1'
db_xref: 'GeneID:963214'
note: [6x58 char]
translation: 'MKSLPVYVSGIKVRNLKNVSIHFNSEEIVLLTGVSGSGKSSIAFDTLYAAGRKRYISTLPTFFATTITTLPNPKVEEIH
text: [46x58 char]

The distribution of the scores is affected by the length of the sequences, with very long sequences
potentially having much higher or lower scores than shorter sequences. You can normalize for this in
a number of ways. One way is to divide by the length of the sequences.

lnormgatABest = gatABest./length(seqtrachomatis.CDS(4).product)
lnormgatCBest = gatCBest./length(seqtrachomatis.CDS(3).product)
lnormuvrABest = uvrABest./length(seqtrachomatis.CDS(339).product)

1-38
 Comparing Whole Genomes

lnormgatABest =

16.8794

lnormgatCBest =

2.2695

lnormuvrABest =

78.9615

An alternative normalization method is to use the self alignment score, that is the score from aligning
the sequence with itself.

gatASelf = nwalign(seqtrachomatis.CDS(4).translation,...
seqtrachomatis.CDS(4).translation);
gatCSelf = nwalign(seqtrachomatis.CDS(3).translation,...
seqtrachomatis.CDS(3).translation);
uvrASelf = nwalign(seqtrachomatis.CDS(339).translation,...
seqtrachomatis.CDS(339).translation);
normgatABest = gatABest./gatASelf
normgatCBest = gatCBest./gatCSelf
normuvrABest = uvrABest./uvrASelf

normgatABest =

0.7380

normgatCBest =

0.5212

normuvrABest =

0.5253

Using Sparse Matrices to Reduce Memory Usage

The all-against-all alignment calculation not only takes a lot of time, it also generates a large matrix
of scores. If you are looking for similar genes across species, then the scores that are interesting are
the positive scores that indicate good alignment. However, most of these scores are negative, and the
actual values are not particularly useful for this type of study. Sparse matrices allow you to store the
interesting values in a more efficient way.

The sparse matrix, spScores, in the MAT-file chlamydia.mat contains the positive values from the
all against all pairwise alignment calculation normalized by self-alignment score.

load('chlamydia.mat','spScores')

1-39
1 Getting Started

With the matrix of scores you can look at the distribution of scores of Chlamydophila pneumoniae
genes aligned with Chlamydia trachomatis and the converse of this, Chlamydia trachomatis genes
aligned with Chlamydophila pneumoniae genes

figure
subplot(2,1,1)
hist(max(spScores),100)
title('Highest Alignment Scores for Chlamydophila pneumoniae Genes')
xlabel('Score');ylabel('Number of genes');
subplot(2,1,2)
hist(max(spScores,[],2),100)
title('Highest Alignment Scores for Chlamydia trachomatis Genes')
xlabel('Score');ylabel('Number of genes');

Remember that there are 1112 CDS in Chlamydophila pneumoniae and only 895 in Chlamydia
trachomatis. The high number of zero scores in the top histogram indicates that many of the extra
CDS in Chlamydophila pneumoniae do not have good matches in Chlamydia trachomatis.

Another way to visualize the data is to look at the positions of points in the scores matrix that are
positive. The sparse function spy is an easy way to quickly view dotplots of matrices. This shows
some interesting structure in the positions of the high scoring matches.

figure
spy(spScores > 0)
title(sprintf('Dot Plot of High-Scoring Alignments.\nNormalized Threshold = 0'))

1-40
 Comparing Whole Genomes

Raise the threshold a little higher to see clear diagonal lines in the plot.

spy(spScores >.1)
title(sprintf('Dot Plot of High-Scoring Alignments.\nNormalized Threshold = 0.1'))

1-41
1 Getting Started

Remember that these are circular genomes, and it seems that the starting points in GenBank are
arbitrary. Permute the scores matrix so that the best match of the first CDS in Chlamydophila
pneumoniae is in the first row to see a clear diagonal plot. This shows the synteny between the two
organisms.

[bestScore bestMatch] = max(spScores(:,1));

spy(spScores([bestMatch:end 1:bestMatch-1],:)>.1);
title('Synteny Plot of Chlamydophila pneumoniae and Chlamydia trachomatis')

1-42
 Comparing Whole Genomes

Looking for Homologous Genes

Genes in different genomes that are related to each other are said to be homologous. Similarity can
be by speciation (orthologous genes) or by replication (paralogous genes). Having the scoring matrix
lets you look for both types of relationships.

The most obvious way to find orthologs is to look for the highest scoring pairing for each gene. If the
score is significant then these best reciprocal pairs are very likely to be orthologous.

[bestScores, bestIndices] = max(spScores);

The variable bestIndices contains the index of the best reciprocal pairs for the genes in
Chlamydophila pneumoniae. Sort the best scores and create a table to compare the description of the
best reciprocal pairs and discover very high similarity between the highest scoring best reciprocal
pairs.

[orderedScores, permScores] = sort(full(bestScores),'descend');

matches = [num2cell(orderedScores)',num2cell(bestIndices(permScores))',...
num2cell((permScores))',...
{seqtrachomatis.CDS(bestIndices(permScores)).product;...
seqpneumoniae.CDS((permScores)).product; }'];

for count = 1:7

fprintf(['Score %f\nChlamydia trachomatis Gene : %s\n',...
'Chlamydophila pneumoniae Gene : %s\n\n'],...
matches{count,1}, matches{count,4}, matches{count,5})
end

1-43
1 Getting Started

Score 0.982993
Chlamydia trachomatis Gene : 50S ribosomal protein L36
Chlamydophila pneumoniae Gene : 50S ribosomal protein L36

Score 0.981818
Chlamydia trachomatis Gene : 30S ribosomal protein S15
Chlamydophila pneumoniae Gene : 30S ribosomal protein S15

Score 0.975422
Chlamydia trachomatis Gene : integration host factor alpha-subunit
Chlamydophila pneumoniae Gene : integration host factor beta-subunit

Score 0.971647
Chlamydia trachomatis Gene : 50S ribosomal protein L16
Chlamydophila pneumoniae Gene : 50S ribosomal protein L16

Score 0.970105
Chlamydia trachomatis Gene : 30S ribosomal protein S10
Chlamydophila pneumoniae Gene : 30S ribosomal protein S10

Score 0.969554
Chlamydia trachomatis Gene : rod shape-determining protein MreB
Chlamydophila pneumoniae Gene : rod shape-determining protein MreB

Score 0.953654
Chlamydia trachomatis Gene : hypothetical protein
Chlamydophila pneumoniae Gene : hypothetical protein

You can use the Variable Editor to look at the data in a spreadsheet format.

open('matches')

Compare the descriptions to see that the majority of the best reciprocal pairs have identical
descriptions.

exactMatches = strcmpi(matches(:,4),matches(:,5));
sum(exactMatches)

ans =

808

Perhaps more interesting are the best reciprocal pairs where the descriptions are not identical. Some
are simply differences in how the same gene is described, but others show quite different
descriptions.

mismatches = matches(~exactMatches,:);
for count = 1:7
fprintf(['Score %f\nChlamydia trachomatis Gene : %s\n',...
'Chlamydophila pneumoniae Gene : %s\n\n'],...
mismatches{count,1}, mismatches{count,4}, mismatches{count,5})
end

Score 0.975422
Chlamydia trachomatis Gene : integration host factor alpha-subunit

1-44
 Comparing Whole Genomes

Chlamydophila pneumoniae Gene : integration host factor beta-subunit

Score 0.929565
Chlamydia trachomatis Gene : low calcium response D
Chlamydophila pneumoniae Gene : type III secretion inner membrane protein SctV

Score 0.905000
Chlamydia trachomatis Gene : NrdR family transcriptional regulator
Chlamydophila pneumoniae Gene : transcriptional regulator NrdR

Score 0.903226
Chlamydia trachomatis Gene : Yop proteins translocation protein S
Chlamydophila pneumoniae Gene : type III secretion inner membrane protein SctS

Score 0.896212
Chlamydia trachomatis Gene : ATP-dependent protease ATP-binding subunit ClpX
Chlamydophila pneumoniae Gene : ATP-dependent protease ATP-binding protein ClpX

Score 0.890705
Chlamydia trachomatis Gene : ribonuclease E
Chlamydophila pneumoniae Gene : ribonuclease G

Score 0.884234
Chlamydia trachomatis Gene : ClpC protease ATPase
Chlamydophila pneumoniae Gene : ATP-dependent Clp protease ATP-binding protein

View data for mismatches.

open('mismatches')

Once you have the scoring matrix this opens up many possibilities for further investigation. For
example, you could look for CDS where there are multiple high scoring reciprocal CDS. See
Cristianini and Hahn [1] for further ideas.

References

[1] Cristianini, N. and Hahn, M.W., "Introduction to Computational Genomics: A Case Studies
Approach", Cambridge University Press, 2007.

See Also
getgenbank | nwalign | featureparse

1-45
 2

High-Throughput Sequence Analysis

• “Work with Next-Generation Sequencing Data” on page 2-2

• “Manage Sequence Read Data in Objects” on page 2-6
• “Store and Manage Feature Annotations in Objects” on page 2-16
• “Bioinformatics Toolbox Software Support Packages” on page 2-21
• “Count Features from NGS Reads” on page 2-23
• “Identifying Differentially Expressed Genes from RNA-Seq Data” on page 2-32
• “Visualize NGS Data Using Genomics Viewer App” on page 2-52
• “Exploring Genome-Wide Differences in DNA Methylation Profiles” on page 2-58
• “Exploring Protein-DNA Binding Sites from Paired-End ChIP-Seq Data” on page 2-79
• “Working with Illumina/Solexa Next-Generation Sequencing Data” on page 2-97
2 High-Throughput Sequence Analysis

Work with Next-Generation Sequencing Data

In this section...
“Overview” on page 2-2
“What Files Can You Access?” on page 2-2
“Before You Begin” on page 2-3
“Create a BioIndexedFile Object to Access Your Source File” on page 2-3
“Determine the Number of Entries Indexed By a BioIndexedFile Object” on page 2-3
“Retrieve Entries from Your Source File” on page 2-4
“Read Entries from Your Source File” on page 2-4

Overview
Many biological experiments produce huge data files that are difficult to access due to their size,
which can cause memory issues when reading the file into the MATLAB Workspace. You can construct
a BioIndexedFile object to access the contents of a large text file containing nonuniform size
entries, such as sequences, annotations, and cross-references to data sets. The BioIndexedFile
object lets you quickly and efficiently access this data without loading the source file into memory.

You can use the BioIndexedFile object to access individual entries or a subset of entries when the
source file is too big to fit into memory. You can access entries using indices or keys. You can read and
parse one or more entries using provided interpreters or a custom interpreter function.

Use the BioIndexedFile object in conjunction with your large source file to:

• Access a subset of the entries for validation or further analysis.

• Parse entries using a custom interpreter function.

What Files Can You Access?

You can use the BioIndexedFile object to access large text files.

Your source file can have these application-specific formats:

• FASTA
• FASTQ
• SAM

Your source file can also have these general formats:

• Table — Tab-delimited table with multiple columns. Keys can be in any column. Rows with the
same key are considered separate entries.
• Multi-row Table — Tab-delimited table with multiple columns. Keys can be in any column.
Contiguous rows with the same key are considered a single entry. Noncontiguous rows with the
same key are considered separate entries.
• Flat — Flat file with concatenated entries separated by a character vector, typically //. Within an
entry, the key is separated from the rest of the entry by a white space.

2-2
 Work with Next-Generation Sequencing Data

Before You Begin

Before constructing a BioIndexedFile object, locate your source file on your hard drive or a local
network.

When you construct a BioIndexedFile object from your source file for the first time, you also
create an auxiliary index file, which by default is saved to the same location as your source file.
However, if your source file is in a read-only location, you can specify a different location to save the
index file.

Tip If you construct a BioIndexedFile object from your source file on subsequent occasions, it
takes advantage of the existing index file, which saves time. However, the index file must be in the
same location or a location specified by the subsequent construction syntax.

Tip If insufficient memory is not an issue when accessing your source file, you may want to try an
appropriate read function, such as genbankread, for importing data from GenBank files. .

Additionally, several read functions such as fastaread, fastqread, samread, and sffread include
a Blockread property, which lets you read a subset of entries from a file, thus saving memory.

Create a BioIndexedFile Object to Access Your Source File

To construct a BioIndexedFile object from a multi-row table file:
1 Create a variable containing the full absolute path of your source file. For your source file, use
the yeastgenes.sgd file, which is included with the Bioinformatics Toolbox software.
sourcefile = which('yeastgenes.sgd');
2 Use the BioIndexedFile constructor function to construct a BioIndexedFile object from the
yeastgenes.sgd source file, which is a multi-row table file. Save the index file in the Current
Folder. Indicate that the source file keys are in column 3. Also, indicate that the header lines in
the source file are prefaced with !, so the constructor ignores them.
gene2goObj = BioIndexedFile('mrtab', sourcefile, '.', ...
'KeyColumn', 3, 'HeaderPrefix','!')

The BioIndexedFile constructor function constructs gene2goObj, a BioIndexedFile object,

and also creates an index file with the same name as the source file, but with an IDX extension. It
stores this index file in the Current Folder because we specified this location. However, the
default location for the index file is the same location as the source file.

Caution Do not modify the index file. If you modify it, you can get invalid results. Also, the
constructor function cannot use a modified index file to construct future objects from the
associated source file.

Determine the Number of Entries Indexed By a BioIndexedFile Object

To determine the number of entries indexed by a BioIndexedFile object, use the NumEntries
property of the BioIndexedFile object. For example, for the gene2goObj object:
gene2goObj.NumEntries

2-3
2 High-Throughput Sequence Analysis

ans =

6476

Note For a list and description of all properties of the object, see BioIndexedFile.

Retrieve Entries from Your Source File

Retrieve entries from your source file using either:

• The index of the entry

• The entry key

Retrieve Entries Using Indices

Use the getEntryByIndex method to retrieve a subset of entries from your source file that
correspond to specified indices. For example, retrieve the first 12 entries from the yeastgenes.sgd
source file:

subset_entries = getEntryByIndex(gene2goObj, [1:12]);

Retrieve Entries Using Keys

Use the getEntryByKey method to retrieve a subset of entries from your source file that are
associated with specified keys. For example, retrieve all entries with keys of AAC1 and AAD10 from
the yeastgenes.sgd source file:

subset_entries = getEntryByKey(gene2goObj, {'AAC1' 'AAD10'});

The output subset_entries is a character vector of concatenated entries. Because the keys in the
yeastgenes.sgd source file are not unique, this method returns all entries that have a key of AAC1
or AAD10.

Read Entries from Your Source File

The BioIndexedFile object includes a read method, which you can use to read and parse a subset
of entries from your source file. The read method parses the entries using an interpreter function
specified by the Interpreter property of the BioIndexedFile object.

Set the Interpreter Property

Before using the read method, make sure the Interpreter property of the BioIndexedFile
object is set appropriately.

If you constructed a BioIndexedFile The Interpreter property ...

object from ...
A source file with an application-specific By default is a handle to a function appropriate for that
format (FASTA, FASTQ, or SAM) file type and typically does not require you to change it.

2-4
 Work with Next-Generation Sequencing Data

If you constructed a BioIndexedFile The Interpreter property ...

object from ...
A source file with a table, multi-row table, or By default is [], which means the interpreter is an
flat format anonymous function in which the output is equivalent to
the input. You can change this to a handle to a function
that accepts a character vector of one or more
concatenated entries and returns a structure or an
array of structures containing the interpreted data.

There are two ways to set the Interpreter property of the BioIndexedFile object:

• When constructing the BioIndexedFile object, use the Interpreter property name/property
value pair
• After constructing the BioIndexedFile object, set the Interpreter property

Note For more information on setting the Interpreter property of the object, see
BioIndexedFile.

Read a Subset of Entries

The read method reads and parses a subset of entries that you specify using either entry indices or
keys.

Example

To quickly find all the gene ontology (GO) terms associated with a particular gene because the entry
keys are gene names:

1 Set the Interpreter property of the gene2goObj BioIndexedFile object to a handle to a

function that reads entries and returns only the column containing the GO term. In this case the
interpreter is a handle to an anonymous function that accepts character vectors and extracts
those that start with the characters GO.

gene2goObj.Interpreter = @(x) regexp(x,'GO:\d+','match')

2 Read only the entries that have a key of YAT2, and return their GO terms.

GO_YAT2_entries = read(gene2goObj, 'YAT2')

GO_YAT2_entries =

'GO:0004092' 'GO:0005737' 'GO:0006066' 'GO:0006066' 'GO:0009437'

2-5
2 High-Throughput Sequence Analysis

Manage Sequence Read Data in Objects

In this section...
“Overview” on page 2-6
“Represent Sequence and Quality Data in a BioRead Object” on page 2-7
“Represent Sequence, Quality, and Alignment/Mapping Data in a BioMap Object” on page 2-8
“Retrieve Information from a BioRead or BioMap Object” on page 2-10
“Set Information in a BioRead or BioMap Object” on page 2-12
“Determine Coverage of a Reference Sequence” on page 2-12
“Construct Sequence Alignments to a Reference Sequence” on page 2-13
“Filter Read Sequences Using SAM Flags” on page 2-14

Overview
High-throughput sequencing instruments produce large amounts of sequence read data that can be
challenging to store and manage. Using objects to contain this data lets you easily access,
manipulate, and filter the data.

Bioinformatics Toolbox includes two objects for working with sequence read data.

Object Contains This Information Construct from One of These

BioRead • Sequence headers • FASTQ file
• Read sequences • SAM file
• Sequence qualities (base calling) • FASTQ structure (created using the
fastqread function)
• SAM structure (created using the
samread function)
• Cell arrays containing header,
sequence, and quality information
(created using the fastqread
function)
BioMap • Sequence headers • SAM file
• Read sequences • BAM file
• Sequence qualities (base calling) • SAM structure (created using the
• Sequence alignment and mapping samread function)
information (relative to a single • BAM structure (created using the
reference sequence), including bamread function)
mapping quality • Cell arrays containing header,
sequence, quality, and mapping/
alignment information (created using
the samread or bamread function)

2-6
 Manage Sequence Read Data in Objects

Represent Sequence and Quality Data in a BioRead Object

Prerequisites

A BioRead object represents a collection of sequence reads. Each element in the object is associated
with a sequence, sequence header, and sequence quality information.

Construct a BioRead object in one of two ways:

• Indexed — The data remains in the source file. Constructing the object and accessing its contents
is memory efficient. However, you cannot modify object properties, other than the Name property.
This is the default method if you construct a BioRead object from a FASTQ- or SAM-formatted
file.
• In Memory — The data is read into memory. Constructing the object and accessing its contents is
limited by the amount of available memory. However, you can modify object properties. When you
construct a BioRead object from a FASTQ structure or cell arrays, the data is read into memory.
When you construct a BioRead object from a FASTQ- or SAM-formatted file, use the InMemory
name-value pair argument to read the data into memory.

Construct a BioRead Object from a FASTQ- or SAM-Formatted File

Note This example constructs a BioRead object from a FASTQ-formatted file. Use similar steps to
construct a BioRead object from a SAM-formatted file.

Use the BioRead constructor function to construct a BioRead object from a FASTQ-formatted file
and set the Name property:

BRObj1 = BioRead('SRR005164_1_50.fastq', 'Name', 'MyObject')

BRObj1 =

BioRead with properties:

Quality: [50x1 File indexed property]

Sequence: [50x1 File indexed property]
Header: [50x1 File indexed property]
NSeqs: 50
Name: 'MyObject'

The constructor function construct a BioRead object and, if an index file does not already exist, it
also creates an index file with the same file name, but with an .IDX extension. This index file, by
default, is stored in the same location as the source file.

Caution Your source file and index file must always be in sync.

• After constructing a BioRead object, do not modify the index file, or you can get invalid results
when using the existing object or constructing new objects.
• If you modify the source file, delete the index file, so the object constructor creates a new index
file when constructing new objects.

2-7
2 High-Throughput Sequence Analysis

Note Because you constructed this BioRead object from a source file, you cannot modify the
properties (except for Name) of the BioRead object.

Represent Sequence, Quality, and Alignment/Mapping Data in a

BioMap Object
Prerequisites

A BioMap object represents a collection of sequence reads that map against a single reference
sequence. Each element in the object is associated with a read sequence, sequence header, sequence
quality information, and alignment/mapping information.

When constructing a BioMap object from a BAM file, the maximum size of the file is limited by your
operating system and available memory.

Construct a BioMap object in one of two ways:

• Indexed — The data remains in the source file. Constructing the object and accessing its contents
is memory efficient. However, you cannot modify object properties, other than the Name property.
This is the default method if you construct a BioMap object from a SAM- or BAM-formatted file.
• In Memory — The data is read into memory. Constructing the object and accessing its contents is
limited by the amount of available memory. However, you can modify object properties. When you
construct a BioMap object from a structure, the data stays in memory. When you construct a
BioMap object from a SAM- or BAM-formatted file, use the InMemory name-value pair argument
to read the data into memory.

Construct a BioMap Object from a SAM- or BAM-Formatted File

Note This example constructs a BioMap object from a SAM-formatted file. Use similar steps to
construct a BioMap object from a BAM-formatted file.

1 If you do not know the number and names of the reference sequences in your source file,
determine them using the saminfo or baminfo function and the ScanDictionary name-value
pair argument.

samstruct = saminfo('ex2.sam', 'ScanDictionary', true);

samstruct.ScannedDictionary

ans =

'seq1'
'seq2'

Tip The previous syntax scans the entire SAM file, which is time consuming. If you are confident
that the Header information of the SAM file is correct, omit the ScanDictionary name-value
pair argument, and inspect the SequenceDictionary field instead.
2 Use the BioMap constructor function to construct a BioMap object from the SAM file and set the
Name property. Because the SAM-formatted file in this example, ex2.sam, contains multiple
reference sequences, use the SelectRef name-value pair argument to specify one reference
sequence, seq1:

2-8
 Manage Sequence Read Data in Objects

BMObj2 = BioMap('ex2.sam', 'SelectRef', 'seq1', 'Name', 'MyObject')

BMObj2 =

BioMap with properties:

SequenceDictionary: 'seq1'
Reference: [1501x1 File indexed property]
Signature: [1501x1 File indexed property]
Start: [1501x1 File indexed property]
MappingQuality: [1501x1 File indexed property]
Flag: [1501x1 File indexed property]
MatePosition: [1501x1 File indexed property]
Quality: [1501x1 File indexed property]
Sequence: [1501x1 File indexed property]
Header: [1501x1 File indexed property]
NSeqs: 1501
Name: 'MyObject'

The constructor function constructs a BioMap object and, if index files do not already exist, it also
creates one or two index files:

• If constructing from a SAM-formatted file, it creates one index file that has the same file name as
the source file, but with an .IDX extension. This index file, by default, is stored in the same
location as the source file.
• If constructing from a BAM-formatted file, it creates two index files that have the same file name
as the source file, but one with a .BAI extension and one with a .LINEARINDEX extension. These
index files, by default, are stored in the same location as the source file.

Caution Your source file and index files must always be in sync.

• After constructing a BioMap object, do not modify the index files, or you can get invalid results
when using the existing object or constructing new objects.
• If you modify the source file, delete the index files, so the object constructor creates new index
files when constructing new objects.

Note Because you constructed this BioMap object from a source file, you cannot modify the
properties (except for Name and Reference) of the BioMap object.

Construct a BioMap Object from a SAM or BAM Structure

Note This example constructs a BioMap object from a SAM structure using samread. Use similar
steps to construct a BioMap object from a BAM structure using bamread.

1 Use the samread function to create a SAM structure from a SAM-formatted file:

SAMStruct = samread('ex2.sam');
2 To construct a valid BioMap object from a SAM-formatted file, the file must contain only one
reference sequence. Determine the number and names of the reference sequences in your SAM-

2-9
2 High-Throughput Sequence Analysis

formatted file using the unique function to find unique names in the ReferenceName field of
the structure:

unique({SAMStruct.ReferenceName})

ans =

'seq1' 'seq2'
3 Use the BioMap constructor function to construct a BioMap object from a SAM structure.
Because the SAM structure contains multiple reference sequences, use the SelectRef name-
value pair argument to specify one reference sequence, seq1:

BMObj1 = BioMap(SAMStruct, 'SelectRef', 'seq1')

BMObj1 =

BioMap with properties:

SequenceDictionary: {'seq1'}
Reference: {1501x1 cell}
Signature: {1501x1 cell}
Start: [1501x1 uint32]
MappingQuality: [1501x1 uint8]
Flag: [1501x1 uint16]
MatePosition: [1501x1 uint32]
Quality: {1501x1 cell}
Sequence: {1501x1 cell}
Header: {1501x1 cell}
NSeqs: 1501
Name: ''

Retrieve Information from a BioRead or BioMap Object

You can retrieve all or a subset of information from a BioRead or BioMap object.

Retrieve a Property from a BioRead or BioMap Object

You can retrieve a specific property from elements in a BioRead or BioMap object.

For example, to retrieve all headers from a BioRead object, use the Header property as follows:

allHeaders = BRObj1.Header;

This syntax returns a cell array containing the headers for all elements in the BioRead object.

Similarly, to retrieve all start positions of aligned read sequences from a BioMap object, use the
Start property of the object:

allStarts = BMObj1.Start;

This syntax returns a vector containing the start positions of aligned read sequences with respect to
the position numbers in the reference sequence in a BioMap object.

2-10
 Manage Sequence Read Data in Objects

Retrieve Multiple Properties from a BioRead or BioMap Object

You can retrieve multiple properties from a BioRead or BioMap object in a single command using the
get method. For example, to retrieve both start positions and headers information of a BioMap
object, use the get method as follows:

multiProp = get(BMObj1, {'Start', 'Header'});

This syntax returns a cell array containing all start positions and headers information of a BioMap
object.

Note Property names are case sensitive.

For a list and description of all properties of a BioRead object, see BioRead class. For a list and
description of all properties of a BioMap object, see BioMap class.

Retrieve a Subset of Information from a BioRead or BioMap Object

Use specialized get methods with a numeric vector, logical vector, or cell array of headers to retrieve
a subset of information from an object. For example, to retrieve the first 10 elements from a BioRead
object, use the getSubset method:

newBRObj = getSubset(BRObj1, [1:10]);

This syntax returns a new BioRead object containing the first 10 elements in the original BioRead
object.

For example, to retrieve the first 12 positions of sequences with headers SRR005164.1,
SRR005164.7, and SRR005164.16, use the getSubsequence method:

subSeqs = getSubsequence(BRObj1, ...

{'SRR005164.1', 'SRR005164.7', 'SRR005164.16'}, [1:12]')

subSeqs =

'TGGCTTTAAAGC'
'CCCGAAAGCTAG'
'AATTTTGCGGCT'

For example, to retrieve information about the third element in a BioMap object, use the getInfo
method:

Info_3 = getInfo(BMObj1, 3);

This syntax returns a tab-delimited character vector containing this information for the third element:

• Sequence header
• SAM flags for the sequence
• Start position of the aligned read sequence with respect to the reference sequence
• Mapping quality score for the sequence
• Signature (CIGAR-formatted character vector) for the sequence
• Sequence

2-11
2 High-Throughput Sequence Analysis

• Quality scores for sequence positions

Note Method names are case sensitive.

For a complete list and description of methods of a BioRead object, see BioRead class. For a
complete list and description of methods of a BioMap object, see BioMap class.

Set Information in a BioRead or BioMap Object

Prerequisites

To modify properties (other than Name and Reference) of a BioRead or BioMap object, the data
must be in memory, and not indexed. To ensure the data is in memory, do one of the following:

• Construct the object from a structure as described in “Construct a BioMap Object from a SAM or
BAM Structure” on page 2-9.
• Construct the object from a source file using the InMemory name-value pair argument.

Provide Custom Headers for Sequences

First, create an object with the data in memory:

BRObj1 = BioRead('SRR005164_1_50.fastq','InMemory',true);

To provide custom headers for sequences of interest (in this case sequences 1 to 5), do the following:

BRObj1.Header(1:5) = {'H1', 'H2', 'H3', 'H4', 'H5'};

Alternatively, you can use the setHeader method:

BRObj1 = setHeader(BRObj1, {'H1', 'H2', 'H3', 'H4', 'H5'}, [1:5]);

Several other specialized set methods let you set the properties of a subset of elements in a
BioRead or BioMap object.

Note Method names are case sensitive.

For a complete list and description of methods of a BioRead object, see BioRead class. For a
complete list and description of methods of a BioMap object, see BioMap class.

Determine Coverage of a Reference Sequence

When working with a BioMap object, you can determine the number of read sequences that:

• Align within a specific region of the reference sequence

• Align to each position within a specific region of the reference sequence

For example, you can compute the number, indices, and start positions of the read sequences that
align within the first 25 positions of the reference sequence. To do so, use the getCounts,
getIndex, and getStart methods:

Cov = getCounts(BMObj1, 1, 25)

2-12
 Manage Sequence Read Data in Objects

Cov =

12

Indices = getIndex(BMObj1, 1, 25)

Indices =

1
2
3
4
5
6
7
8
9
10
11
12

startPos = getStart(BMObj1, Indices)

startPos =

1
3
5
6
9
13
13
15
18
22
22
24

The first two syntaxes return the number and indices of the read sequences that align within the
specified region of the reference sequence. The last syntax returns a vector containing the start
position of each aligned read sequence, corresponding to the position numbers of the reference
sequence.

For example, you can also compute the number of the read sequences that align to each of the first
10 positions of the reference sequence. For this computation, use the getBaseCoverage method:

Cov = getBaseCoverage(BMObj1, 1, 10)

Cov =

1 1 2 2 3 4 4 4 5 5

Construct Sequence Alignments to a Reference Sequence

It is useful to construct and view the alignment of the read sequences that align to a specific region of
the reference sequence. It is also helpful to know which read sequences align to this region in a
BioMap object.

2-13
2 High-Throughput Sequence Analysis

For example, to retrieve the alignment of read sequences to the first 12 positions of the reference
sequence in a BioMap object, use the getAlignment method:

[Alignment_1_12, Indices] = getAlignment(BMObj2, 1, 12)

Alignment_1_12 =

CACTAGTGGCTC
CTAGTGGCTC
AGTGGCTC
GTGGCTC
GCTC

Indices =

1
2
3
4
5

Return the headers of the read sequences that align to a specific region of the reference sequence:

alignedHeaders = getHeader(BMObj2, Indices)

alignedHeaders =

'B7_591:4:96:693:509'
'EAS54_65:7:152:368:113'
'EAS51_64:8:5:734:57'
'B7_591:1:289:587:906'
'EAS56_59:8:38:671:758'

Filter Read Sequences Using SAM Flags

SAM- and BAM-formatted files include the status of 11 binary flags for each read sequence. These
flags describe different sequencing and alignment aspects of a read sequence. For more information
on the flags, see the SAM Format Specification. The filterByFlag method lets you filter the read
sequences in a BioMap object by using these flags.

Filter Unmapped Read Sequences

1 Construct a BioMap object from a SAM-formatted file.

BMObj2 = BioMap('ex1.sam');
2 Use the filterByFlag method to create a logical vector indicating the read sequences in a
BioMap object that are mapped.

LogicalVec_mapped = filterByFlag(BMObj2, 'unmappedQuery', false);

3 Use this logical vector and the getSubset method to create a new BioMap object containing
only the mapped read sequences.

filteredBMObj_1 = getSubset(BMObj2, LogicalVec_mapped);

2-14
 Manage Sequence Read Data in Objects

Filter Read Sequences That Are Not Mapped in a Pair

1 Construct a BioMap object from a SAM-formatted file.

BMObj2 = BioMap('ex1.sam');
2 Use the filterByFlag method to create a logical vector indicating the read sequences in a
BioMap object that are mapped in a proper pair, that is, both the read sequence and its mate are
mapped to the reference sequence.

LogicalVec_paired = filterByFlag(BMObj2, 'pairedInMap', true);

3 Use this logical vector and the getSubset method to create a new BioMap object containing
only the read sequences that are mapped in a proper pair.

filteredBMObj_2 = getSubset(BMObj2, LogicalVec_paired);

2-15
2 High-Throughput Sequence Analysis

Store and Manage Feature Annotations in Objects

In this section...
“Represent Feature Annotations in a GFFAnnotation or GTFAnnotation Object” on page 2-16
“Construct an Annotation Object” on page 2-16
“Retrieve General Information from an Annotation Object” on page 2-16
“Access Data in an Annotation Object” on page 2-17
“Use Feature Annotations with Sequence Read Data” on page 2-18

Represent Feature Annotations in a GFFAnnotation or GTFAnnotation

Object
The GFFAnnotation and GTFAnnotation objects represent a collection of feature annotations for
one or more reference sequences. You construct these objects from GFF (General Feature Format)
and GTF (Gene Transfer Format) files. Each element in the object represents a single annotation. The
properties and methods associated with the objects let you investigate and filter the data based on
reference sequence, a feature (such as CDS or exon), or a specific gene or transcript.

Construct an Annotation Object

Use the GFFAnnotation constructor function to construct a GFFAnnotation object from either a
GFF- or GTF-formatted file:

GFFAnnotObj = GFFAnnotation('tair8_1.gff')

GFFAnnotObj =

GFFAnnotation with properties:

FieldNames: {1x9 cell}

NumEntries: 3331

Use the GTFAnnotation constructor function to construct a GTFAnnotation object from a GTF-
formatted file:

GTFAnnotObj = GTFAnnotation('hum37_2_1M.gtf')

GTFAnnotObj =

GTFAnnotation with properties:

FieldNames: {1x11 cell}

NumEntries: 308

Retrieve General Information from an Annotation Object

Determine the field names and the number of entries in an annotation object by accessing the
FieldNames and NumEntries properties. For example, to see the field names for each annotation
object constructed in the previous section, query the FieldNames property:

GFFAnnotObj.FieldNames

2-16
 Store and Manage Feature Annotations in Objects

ans =

Columns 1 through 6

'Reference' 'Start' 'Stop' 'Feature' 'Source' 'Score'

Columns 7 through 9

'Strand' 'Frame' 'Attributes'

GTFAnnotObj.FieldNames

ans =

Columns 1 through 6

'Reference' 'Start' 'Stop' 'Feature' 'Gene' 'Transcript'

Columns 7 through 11

'Source' 'Score' 'Strand' 'Frame' 'Attributes'

Determine the range of the reference sequences that are covered by feature annotations by using the
getRange method with the annotation object constructed in the previous section:

range = getRange(GFFAnnotObj)

range =

3631 498516

Access Data in an Annotation Object

Create a Structure of the Annotation Data

Creating a structure of the annotation data lets you access the field values. Use the getData method
to create a structure containing a subset of the data in a GFFAnnotation object constructed in the
previous section.

% Extract annotations for positions 1 through 10000 of the

% reference sequence
AnnotStruct = getData(GFFAnnotObj,1,10000)

AnnotStruct =

60x1 struct array with fields:

Reference
Start
Stop
Feature
Source
Score
Strand
Frame
Attributes

2-17
2 High-Throughput Sequence Analysis

Access Field Values in the Structure

Use dot indexing to access all or specific field values in a structure.

For example, extract the start positions for all annotations:

Starts = AnnotStruct.Start;

Extract the start positions for annotations 12 through 17. Notice that you must use square brackets
when indexing a range of positions:
Starts_12_17 = [AnnotStruct(12:17).Start]

Starts_12_17 =

4706 5174 5174 5439 5439 5631

Extract the start position and the feature for the 12th annotation:
Start_12 = AnnotStruct(12).Start

Start_12 =

4706

Feature_12 = AnnotStruct(12).Feature

Feature_12 =

CDS

Use Feature Annotations with Sequence Read Data

Investigate the results of HTS sequencing experiments by using GFFAnnotation and
GTFAnnotation objects with BioMap objects. For example, you can:

• Determine counts of sequence reads aligned to regions of a reference sequence associated with
specific annotations, such as in RNA-Seq workflows.
• Find annotations within a specific range of a peak of interest in a reference sequence, such as in
ChIP-Seq workflows.

Determine Annotations of Interest

1 Construct a GTFAnnotation object from a GTF- formatted file:

GTFAnnotObj = GTFAnnotation('hum37_2_1M.gtf');
2 Use the getReferenceNames method to return the names for the reference sequences for the
annotation object:
refNames = getReferenceNames(GTFAnnotObj)

refNames =

'chr2'
3 Use the getFeatureNames method to retrieve the feature names from the annotation object:
featureNames = getFeatureNames(GTFAnnotObj)

2-18
 Store and Manage Feature Annotations in Objects

featureNames =

'CDS'
'exon'
'start_codon'
'stop_codon'
4 Use the getGeneNames method to retrieve a list of the unique gene names from the annotation
object:
geneNames = getGeneNames(GTFAnnotObj)

geneNames =

'uc002qvu.2'
'uc002qvv.2'
'uc002qvw.2'
'uc002qvx.2'
'uc002qvy.2'
'uc002qvz.2'
'uc002qwa.2'
'uc002qwb.2'
'uc002qwc.1'
'uc002qwd.2'
'uc002qwe.3'
'uc002qwf.2'
'uc002qwg.2'
'uc002qwh.2'
'uc002qwi.3'
'uc002qwk.2'
'uc002qwl.2'
'uc002qwm.1'
'uc002qwn.1'
'uc002qwo.1'
'uc002qwp.2'
'uc002qwq.2'
'uc010ewe.2'
'uc010ewf.1'
'uc010ewg.2'
'uc010ewh.1'
'uc010ewi.2'
'uc010yim.1'

The previous steps gave us a list of available reference sequences, features, and genes associated
with the available annotations. Use this information to determine annotations of interest. For
instance, you might be interested only in annotations that are exons associated with the uc002qvv.2
gene on chromosome 2.

Filter Annotations

Use the getData method to filter the annotations and create a structure containing only the
annotations of interest, which are annotations that are exons associated with the uc002qvv.2 gene on
chromosome 2.
AnnotStruct = getData(GTFAnnotObj,'Reference','chr2',...
'Feature','exon','Gene','uc002qvv.2')

AnnotStruct =

2-19
2 High-Throughput Sequence Analysis

12x1 struct array with fields:

Reference
Start
Stop
Feature
Gene
Transcript
Source
Score
Strand
Frame
Attributes

The return structure contains 12 elements, indicating there are 12 annotations that meet your filter
criteria.

Extract Position Ranges for Annotations of Interest

After filtering the data to include only annotations that are exons associated with the uc002qvv.2
gene on chromosome 2, use the Start and Stop fields to create vectors of the start and end positions
for the ranges associated with the 12 annotations.

StartPos = [AnnotStruct.Start];
EndPos = [AnnotStruct.Stop];

Determine Counts of Sequence Reads Aligned to Annotations

Construct a BioMap object from a BAM-formatted file containing sequence read data aligned to
chromosome 2.

BMObj3 = BioMap('ex3.bam');

Then use the range for the annotations of interest as input to the getCounts method of a BioMap
object. This returns the counts of short reads aligned to the annotations of interest.

counts = getCounts(BMObj3,StartPos,EndPos,'independent', true)

counts =

1399
1
54
221
97
125
0
1
0
65
9
12

2-20
 Bioinformatics Toolbox Software Support Packages

Bioinformatics Toolbox Software Support Packages

Bioinformatics Toolbox provides support packages for various next-generation sequencing workflows
and analyses. To make a support package available in your MATLAB command line, you must first
install it.

Install Support Package

Follow these steps to install a support package.

1 In the Environment section of the MATLAB toolstrip, select Add-Ons > Get Add-Ons.
2 In the Add-On Explorer, search for the support package that you want to install by entering its
name.
3 Install the support package.

For details about installing add-ons, see “Get and Manage Add-Ons”. For other information, see “Add-
Ons”.

Available Support Packages

The following table lists all the Bioinformatics Toolbox support packages that are available for
download as Add-Ons.

Support Package Name Version† Corresponding MATLAB functions Supporte

d OS
Bowtie 2 Support Package for 2.3.2 bowtie2, bowtie2build, bowtie2inspect. Windows
®‡
Bioinformatics Toolbox [1] , Mac ,
(download link) and
UNIX®
Cufflinks Support Package for 2.2.1 cufflinks, cuffcompare, cuffdiff, Windows‡
the Bioinformatics Toolbox [2] cuffgffread, cuffgtf2sam, cuffmerge, , Mac ,
(download link) cuffnorm, cuffquant. and UNIX
BWA Support Package for 0.7.17 bwaindex, bwamem. Windows‡
Bioinformatics Toolbox [3][4] , Mac ,
(download link) and UNIX

†
Version of the original (third-party) software

‡
You need to install Windows Subsystem for Linux (WSL) and a Linux distribution on your Windows
machine. For details on installing WSL, see here.

See Also

More About
• “Count Features from NGS Reads” on page 2-23
• “High-Throughput Sequencing”

2-21
2 High-Throughput Sequence Analysis

References
[1] Langmead, Ben, and Steven L Salzberg. “Fast Gapped-Read Alignment with Bowtie 2.” Nature
Methods 9, no. 4 (April 2012): 357–59. https://doi.org/10.1038/nmeth.1923.

[2] Trapnell, Cole, Brian A Williams, Geo Pertea, Ali Mortazavi, Gordon Kwan, Marijke J van Baren,
Steven L Salzberg, Barbara J Wold, and Lior Pachter. “Transcript Assembly and Quantification
by RNA-Seq Reveals Unannotated Transcripts and Isoform Switching during Cell
Differentiation.” Nature Biotechnology 28, no. 5 (May 2010): 511–15.

[3] Li, Heng, and Richard Durbin. “Fast and Accurate Long-Read Alignment with Burrows–Wheeler
Transform.” Bioinformatics 26, no. 5 (March 1, 2010): 589–95. https://doi.org/10.1093/
bioinformatics/btp698.

[4] Li, Heng, and Richard Durbin. “Fast and Accurate Short Read Alignment with Burrows-Wheeler
Transform.” Bioinformatics 25, no. 14 (July 15, 2009): 1754–60. https://doi.org/10.1093/
bioinformatics/btp324.

2-22
 Count Features from NGS Reads

Count Features from NGS Reads

This example shows how to count features from paired-end sequencing reads after aligning them to
the whole human genome curated by the Genome Reference Consortium. This example uses Genome
Reference Consortium Human Build 38 patch release 12 (GRCh38.p12) as the human genome
reference.

Prerequisites and Data Set

This example works on the UNIX® and Mac platforms only. Download the Bioinformatics Toolbox™
Interface for Bowtie Aligner support package from the Add-On Explorer. For details, see
“Bioinformatics Toolbox Software Support Packages” on page 2-21.

This example assumes you have:

• Downloaded and extracted the RefSeq assembly from Genome Reference Consortium Human
Build 38 patch release 12 (GRCh38.p12).
• Downloaded and organized some paired-end reads data. This example uses the exome sequencing
data from the 1000 genomes project. Paired-end reads are indicated by '_1' and '_2' in the
filenames.

Build Index

Construct an index for aligning reads to the reference using bowtie2build. The file
GCF_000001405.38_GRCh38.p12_genomic.fna contains the human reference genome in the
FASTA format. bowtieIdx is the base name of the reference index files. The '--threads 8' option
specifies the number of parallel threads to build index files faster. You do not need to specify full file
paths for *.fna or *.index files if you are running the example from the same folder location. Specify
the full paths if you wish to store the files elsewhere or run the example from a different folder.

bowtieIdx = 'GCF_000001405.38_GRCh38.p12_genomic.index';
buildFlag = bowtie2build('GCF_000001405.38_GRCh38.p12_genomic.fna',...
bowtieIdx,'--threads 8');

Align Reads to Reference

Align paired-end reads to the reference using bowtie2. You can create a Bowtie2AlignOptions
object to specify different options, such as the number of parallel threads to use.

opt = Bowtie2AlignOptions;
opt.NumThreads = 8;
reads1 = 'HG00096_1.fastq';
reads2 = 'HG00096_2.fastq';
bowtie2(bowtieIdx,reads1,reads2,'HG00096.sam',opt);

Selectively Align to Gene of Interest

SAM files can be very large. Use BioMap to select only the data for the correct reference. For this
example, consider APOE, which is a gene on Chromosome 19 linked to Alzheimer's disease. Create a
smaller BAM file for APOE to improve performance.

apoeRef = 'NC_000019.10'; % Reference name for Chromosome 19 in HG38

bm = BioMap('HG00096.sam','SelectReference',apoeRef);
write(bm, 'HG00096.bam','Format','bam');

2-23
2 High-Throughput Sequence Analysis

Warning: Found invalid tag in header type: 'PG'. Ignoring tag 'PN:bowtie2'.
Warning: The read sequences in input SAM file do not appear to be ordered
according to the start position of their alignments with the reference sequence.
Because of this, there will be a decrease in performance when accessing the
reads. For maximum performance, order the read sequences in the SAM file, before
creating a BioMap object.

Summarize Read Counts

Use featurecount to compare the number of transcripts for each APOE variant using a GTF file. A
full table of features is included in the GRCh38.p12 assembly in GFF format, which can be converted
to GTF using cuffgffread. This example uses a simplified GTF based on APOE transcripts.
APOE_gene.gtf is included with the software.

[FeatTable, Summary] = featurecount('APOE_gene.gtf','HG00096.bam',...

'Metafeature','transcript_id');

Processing GTF file APOE_gene.gtf ...

Processing BAM file HG00096.bam ...
Processing reference NC_000019.10 ...
10000 reads processed ...
20000 reads processed ...
30000 reads processed ...
40000 reads processed ...
50000 reads processed ...
60000 reads processed ...
70000 reads processed ...
80000 reads processed ...
90000 reads processed ...
100000 reads processed ...
110000 reads processed ...
120000 reads processed ...
130000 reads processed ...
140000 reads processed ...
150000 reads processed ...
160000 reads processed ...
170000 reads processed ...
180000 reads processed ...
190000 reads processed ...
200000 reads processed ...
210000 reads processed ...
220000 reads processed ...
230000 reads processed ...
240000 reads processed ...
250000 reads processed ...
260000 reads processed ...
270000 reads processed ...
280000 reads processed ...
290000 reads processed ...
300000 reads processed ...
310000 reads processed ...
320000 reads processed ...
330000 reads processed ...
340000 reads processed ...
350000 reads processed ...
360000 reads processed ...
370000 reads processed ...
380000 reads processed ...

2-24
 Count Features from NGS Reads

390000 reads processed ...

400000 reads processed ...
410000 reads processed ...
420000 reads processed ...
430000 reads processed ...
440000 reads processed ...
450000 reads processed ...
460000 reads processed ...
470000 reads processed ...
480000 reads processed ...
490000 reads processed ...
500000 reads processed ...
510000 reads processed ...
520000 reads processed ...
530000 reads processed ...
540000 reads processed ...
550000 reads processed ...
560000 reads processed ...
570000 reads processed ...
580000 reads processed ...
590000 reads processed ...
600000 reads processed ...
610000 reads processed ...
620000 reads processed ...
630000 reads processed ...
640000 reads processed ...
650000 reads processed ...
660000 reads processed ...
670000 reads processed ...
680000 reads processed ...
690000 reads processed ...
700000 reads processed ...
710000 reads processed ...
720000 reads processed ...
730000 reads processed ...
740000 reads processed ...
750000 reads processed ...
760000 reads processed ...
770000 reads processed ...
780000 reads processed ...
790000 reads processed ...
800000 reads processed ...
810000 reads processed ...
820000 reads processed ...
830000 reads processed ...
840000 reads processed ...
850000 reads processed ...
860000 reads processed ...
870000 reads processed ...
880000 reads processed ...
890000 reads processed ...
900000 reads processed ...
910000 reads processed ...
920000 reads processed ...
930000 reads processed ...
940000 reads processed ...
950000 reads processed ...
960000 reads processed ...

2-25
2 High-Throughput Sequence Analysis

970000 reads processed ...

980000 reads processed ...
990000 reads processed ...
1000000 reads processed ...
1010000 reads processed ...
1020000 reads processed ...
1030000 reads processed ...
1040000 reads processed ...
1050000 reads processed ...
1060000 reads processed ...
1070000 reads processed ...
1080000 reads processed ...
1090000 reads processed ...
1100000 reads processed ...
1110000 reads processed ...
1120000 reads processed ...
1130000 reads processed ...
1140000 reads processed ...
1150000 reads processed ...
1160000 reads processed ...
1170000 reads processed ...
1180000 reads processed ...
1190000 reads processed ...
1200000 reads processed ...
1210000 reads processed ...
1220000 reads processed ...
1230000 reads processed ...
1240000 reads processed ...
1250000 reads processed ...
1260000 reads processed ...
1270000 reads processed ...
1280000 reads processed ...
1290000 reads processed ...
1300000 reads processed ...
1310000 reads processed ...
1320000 reads processed ...
1330000 reads processed ...
1340000 reads processed ...
1350000 reads processed ...
1360000 reads processed ...
1370000 reads processed ...
1380000 reads processed ...
1390000 reads processed ...
1400000 reads processed ...
1410000 reads processed ...
1420000 reads processed ...
1430000 reads processed ...
1440000 reads processed ...
1450000 reads processed ...
1460000 reads processed ...
1470000 reads processed ...
1480000 reads processed ...
1490000 reads processed ...
1500000 reads processed ...
1510000 reads processed ...
1520000 reads processed ...
1530000 reads processed ...
1540000 reads processed ...

2-26
 Count Features from NGS Reads

1550000 reads processed ...

1560000 reads processed ...
1570000 reads processed ...
1580000 reads processed ...
1590000 reads processed ...
1600000 reads processed ...
1610000 reads processed ...
1620000 reads processed ...
1630000 reads processed ...
1640000 reads processed ...
1650000 reads processed ...
1660000 reads processed ...
1670000 reads processed ...
1680000 reads processed ...
1690000 reads processed ...
1700000 reads processed ...
1710000 reads processed ...
1720000 reads processed ...
1730000 reads processed ...
1740000 reads processed ...
1750000 reads processed ...
1760000 reads processed ...
1770000 reads processed ...
1780000 reads processed ...
1790000 reads processed ...
1800000 reads processed ...
1810000 reads processed ...
1820000 reads processed ...
1830000 reads processed ...
1840000 reads processed ...
1850000 reads processed ...
1860000 reads processed ...
1870000 reads processed ...
1880000 reads processed ...
1890000 reads processed ...
1900000 reads processed ...
1910000 reads processed ...
1920000 reads processed ...
1930000 reads processed ...
1940000 reads processed ...
1950000 reads processed ...
1960000 reads processed ...
1970000 reads processed ...
1980000 reads processed ...
1990000 reads processed ...
2000000 reads processed ...
2010000 reads processed ...
2020000 reads processed ...
2030000 reads processed ...
2040000 reads processed ...
2050000 reads processed ...
2060000 reads processed ...
2070000 reads processed ...
2080000 reads processed ...
2090000 reads processed ...
2100000 reads processed ...
2110000 reads processed ...
2120000 reads processed ...

2-27
2 High-Throughput Sequence Analysis

2130000 reads processed ...

2140000 reads processed ...
2150000 reads processed ...
2160000 reads processed ...
2170000 reads processed ...
2180000 reads processed ...
2190000 reads processed ...
2200000 reads processed ...
2210000 reads processed ...
2220000 reads processed ...
2230000 reads processed ...
2240000 reads processed ...
2250000 reads processed ...
2260000 reads processed ...
2270000 reads processed ...
2280000 reads processed ...
2290000 reads processed ...
2300000 reads processed ...
2310000 reads processed ...
2320000 reads processed ...
2330000 reads processed ...
2340000 reads processed ...
2350000 reads processed ...
2360000 reads processed ...
2370000 reads processed ...
2380000 reads processed ...
2390000 reads processed ...
2400000 reads processed ...
2410000 reads processed ...
2420000 reads processed ...
2430000 reads processed ...
2440000 reads processed ...
2450000 reads processed ...
2460000 reads processed ...
2470000 reads processed ...
2480000 reads processed ...
2490000 reads processed ...
2500000 reads processed ...
2510000 reads processed ...
2520000 reads processed ...
2530000 reads processed ...
2540000 reads processed ...
2550000 reads processed ...
2560000 reads processed ...
2570000 reads processed ...
2580000 reads processed ...
2590000 reads processed ...
2600000 reads processed ...
2610000 reads processed ...
2620000 reads processed ...
2630000 reads processed ...
2640000 reads processed ...
2650000 reads processed ...
2660000 reads processed ...
2670000 reads processed ...
2680000 reads processed ...
2690000 reads processed ...
2700000 reads processed ...

2-28
 Count Features from NGS Reads

2710000 reads processed ...

2720000 reads processed ...
2730000 reads processed ...
2740000 reads processed ...
2750000 reads processed ...
2760000 reads processed ...
2770000 reads processed ...
2780000 reads processed ...
2790000 reads processed ...
2800000 reads processed ...
2810000 reads processed ...
2820000 reads processed ...
2830000 reads processed ...
2840000 reads processed ...
2850000 reads processed ...
2860000 reads processed ...
2870000 reads processed ...
2880000 reads processed ...
2890000 reads processed ...
2900000 reads processed ...
2910000 reads processed ...
2920000 reads processed ...
2930000 reads processed ...
2940000 reads processed ...
2950000 reads processed ...
2960000 reads processed ...
2970000 reads processed ...
2980000 reads processed ...
2990000 reads processed ...
3000000 reads processed ...
3010000 reads processed ...
3020000 reads processed ...
3030000 reads processed ...
3040000 reads processed ...
3050000 reads processed ...
3060000 reads processed ...
3070000 reads processed ...
3080000 reads processed ...
3090000 reads processed ...
3100000 reads processed ...
3110000 reads processed ...
3120000 reads processed ...
3130000 reads processed ...
3140000 reads processed ...
3150000 reads processed ...
3160000 reads processed ...
3170000 reads processed ...
3180000 reads processed ...
3190000 reads processed ...
3200000 reads processed ...
3210000 reads processed ...
3220000 reads processed ...
3230000 reads processed ...
3240000 reads processed ...
3250000 reads processed ...
3260000 reads processed ...
3270000 reads processed ...
3280000 reads processed ...

2-29
2 High-Throughput Sequence Analysis

3290000 reads processed ...

3300000 reads processed ...
3310000 reads processed ...
3320000 reads processed ...
3330000 reads processed ...
3340000 reads processed ...
3350000 reads processed ...
3360000 reads processed ...
3370000 reads processed ...
3380000 reads processed ...
3390000 reads processed ...
3400000 reads processed ...
3410000 reads processed ...
3420000 reads processed ...
3430000 reads processed ...
3440000 reads processed ...
3450000 reads processed ...
3460000 reads processed ...
3470000 reads processed ...
3480000 reads processed ...
3490000 reads processed ...
3500000 reads processed ...
3510000 reads processed ...
3520000 reads processed ...
3530000 reads processed ...
3540000 reads processed ...
3550000 reads processed ...
3560000 reads processed ...
3570000 reads processed ...
3580000 reads processed ...
3590000 reads processed ...
3600000 reads processed ...
3610000 reads processed ...
3620000 reads processed ...
3630000 reads processed ...
3640000 reads processed ...
3650000 reads processed ...
3660000 reads processed ...
3670000 reads processed ...
3680000 reads processed ...
3690000 reads processed ...
3700000 reads processed ...
3710000 reads processed ...
3720000 reads processed ...
3730000 reads processed ...
3740000 reads processed ...
3750000 reads processed ...
3760000 reads processed ...
3770000 reads processed ...
3780000 reads processed ...
3790000 reads processed ...
3800000 reads processed ...
3810000 reads processed ...
3820000 reads processed ...
3830000 reads processed ...
3840000 reads processed ...
3850000 reads processed ...
3860000 reads processed ...

2-30
 Count Features from NGS Reads

3870000 reads processed ...

3880000 reads processed ...
3890000 reads processed ...
3900000 reads processed ...
3910000 reads processed ...
3920000 reads processed ...
3930000 reads processed ...
3940000 reads processed ...
3950000 reads processed ...
3960000 reads processed ...
3970000 reads processed ...
Done.

See Also
bamsort | samsort | bwamem | bowtie2 | bowtie2build | featurecount | BioMap |
cuffgffread | cufflinks

2-31
2 High-Throughput Sequence Analysis

Identifying Differentially Expressed Genes from RNA-Seq Data

This example shows how to test RNA-Seq data for differentially expressed genes using a negative
binomial model.

Introduction

A typical differential expression analysis of RNA-Seq data consists of normalizing the raw counts and
performing statistical tests to reject or accept the null hypothesis that two groups of samples show no
significant difference in gene expression. This example shows how to inspect the basic statistics of
raw count data, how to determine size factors for count normalization and how to infer the most
differentially expressed genes using a negative binomial model.

The dataset for this example comprises of RNA-Seq data obtained in the experiment described by
Brooks et al. [1]. The authors investigated the effect of siRNA knock-down of pasilla, a gene known to
play an important role in the regulation of splicing in Drosophila melanogaster. The dataset consists
of 2 biological replicates of the control (untreated) samples and 2 biological replicates of the knock-
down (treated) samples.

Inspecting Read Count Tables for Genomic Features

The starting point for this analysis of RNA-Seq data is a count matrix, where the rows correspond to
genomic features of interest, the columns correspond to the given samples and the values represent
the number of reads mapped to each feature in a given sample.

The included file pasilla_count_noMM.mat contains two tables with the count matrices at the
gene level and at the exon level for each of the considered samples. You can obtain similar matrices
using the function featurecount.
load pasilla_count_noMM.mat

% preview the table of read counts for genes

head(geneCountTable)

ID Reference untreated3 untreated4 treated2 treated3

_____________ _________ __________ __________ ________ ________

"FBgn0000003" "3R" 0 1 1 2
"FBgn0000008" "2R" 142 117 138 132
"FBgn0000014" "3R" 20 12 10 19
"FBgn0000015" "3R" 2 4 0 1
"FBgn0000017" "3L" 6591 5127 4809 6027
"FBgn0000018" "2L" 469 530 492 574
"FBgn0000024" "3R" 5 6 10 8
"FBgn0000028" "X" 0 0 2 1

Note that when counting is performed without summarization, the individual features (exons in this
case) are reported with their metafeature assignment (genes in this case) followed by the start and
stop positions.
% preview the table of read counts for exons
head(exonCountTable)

ID Reference untreated3 untreated4 treated2 treat

_______________________________ _________ __________ __________ ________ _____

2-32
 Identifying Differentially Expressed Genes from RNA-Seq Data

"FBgn0000003_2648220_2648518" "3R" 0 0 0 1
"FBgn0000008_18024938_18025756" "2R" 0 1 0 2
"FBgn0000008_18050410_18051199" "2R" 13 9 14 12
"FBgn0000008_18052282_18052494" "2R" 4 2 5 0
"FBgn0000008_18056749_18058222" "2R" 32 27 26 23
"FBgn0000008_18058283_18059490" "2R" 14 18 29 22
"FBgn0000008_18059587_18059757" "2R" 1 4 3 0
"FBgn0000008_18059821_18059938" "2R" 0 0 2 0

You can annotate and group the samples by creating a logical vector as follows:

samples = geneCountTable(:,3:end).Properties.VariableNames;
untreated = strncmp(samples,'untreated',length('untreated'))
treated = strncmp(samples,'treated',length('treated'))

untreated =

1x4 logical array

1 1 0 0

treated =

1x4 logical array

0 0 1 1

Plotting the Feature Assignments

The included file also contains a table geneSummaryTable with the summary of assigned and
unassigned SAM entries. You can plot the basic distribution of the counting results by considering the
number of reads that are assigned to the given genomic features (exons or genes for this example), as
well as the number of reads that are unassigned (i.e. not overlapping any feature) or ambiguous (i.e.
overlapping multiple features).

st = geneSummaryTable({'Assigned','Unassigned_ambiguous','Unassigned_noFeature'},:)
bar(table2array(st)','stacked');
legend(st.Properties.RowNames','Interpreter','none','Location','southeast');
xlabel('Sample')
ylabel('Number of reads')

st =

3x4 table

untreated3 untreated4 treated2 treated3

__________ __________ __________ __________

Assigned 1.5457e+07 1.6302e+07 1.5146e+07 1.8856e+07

Unassigned_ambiguous 1.5708e+05 1.6882e+05 1.6194e+05 1.9977e+05
Unassigned_noFeature 7.5455e+05 5.8309e+05 5.8756e+05 6.8356e+05

2-33
2 High-Throughput Sequence Analysis

Note that a small fraction of the alignment records in the SAM files is not reported in the summary
table. You can notice this in the difference between the total number of records in a SAM file and the
total number of records processed during the counting procedure for that same SAM file. These
unreported records correspond to the records mapped to reference sequences that are not annotated
in the GTF file and therefore are not processed in the counting procedure. If the gene models account
for all the reference sequences used during the read mapping step, then all records are reported in
one of the categories of the summary table.
geneSummaryTable{'TotalEntries',:} - sum(geneSummaryTable{2:end,:})

ans =

89516 95885 98207 104629

Plotting Read Coverage Across a Given Chromosome

When read counting is performed without summarization using the function featurecount, the
default IDs are composed by the attribute or metafeature (by default, gene_id) followed by the start
and the stop positions of the feature (by default, exon). You can use the exon start positions to plot
the read coverage across any chromosome in consideration, for example chromosome arm 2L.
% consider chromosome arm 2L
chr2L = strcmp(exonCountTable.Reference,'2L');
exonCount = exonCountTable{:,3:end};

2-34
 Identifying Differentially Expressed Genes from RNA-Seq Data

% retrieve exon start positions

exonStart = regexp(exonCountTable{chr2L,1},'_(\d+)_','tokens');
exonStart = [exonStart{:}];
exonStart = cellfun(@str2num, [exonStart{:}]');

% sort exon by start positions

[~,idx] = sort(exonStart);

% plot read coverage along the genomic coordinates

figure;
plot(exonStart(idx),exonCount(idx,treated),'.-r',...
exonStart(idx),exonCount(idx,untreated),'.-b');
xlabel('Genomic position');
ylabel('Read count (exon level)');
title('Read count on Chromosome arm 2L');

% plot read coverage for each group separately

figure;
subplot(2,1,1);
plot(exonStart(idx),exonCount(idx,untreated),'.-r');
ylabel('Read count (exon level)');
title('Chromosome arm 2L (treated samples)');
subplot(2,1,2);
plot(exonStart(idx),exonCount(idx,treated),'.-b');
ylabel('Read count (exon level)');
xlabel('Genomic position');
title('Chromosome arm 2L (untreated samples)');

2-35
2 High-Throughput Sequence Analysis

Alternatively, you can plot the read coverage considering the starting position of each gene in a given
chromosome. The file pasilla_geneLength.mat contains a table with the start and stop position of
each gene in the corresponding gene annotation file.
% load gene start and stop position information
load pasilla_geneLength
head(geneLength)

ID Name Reference Start Stop

_____________ _________ _________ _____ _____

"FBgn0037213" "CG12581" 3R 380 10200

"FBgn0000500" "Dsk" 3R 15388 16170
"FBgn0053294" "CR33294" 3R 17136 21871
"FBgn0037215" "CG12582" 3R 23029 30295
"FBgn0037217" "CG14636" 3R 30207 41033
"FBgn0037218" "aux" 3R 37505 53244
"FBgn0051516" "CG31516" 3R 44179 45852
"FBgn0261436" "DhpD" 3R 53106 54971

% consider chromosome 3 ('Reference' is a categorical variable)

chr3 = (geneLength.Reference == '3L') | (geneLength.Reference == '3R');
sum(chr3)

% consider the counts for genes in chromosome 3

counts = geneCountTable{:,3:end};
[~,j,k] = intersect(geneCountTable{:,'ID'},geneLength{chr3,'ID'});

2-36
 Identifying Differentially Expressed Genes from RNA-Seq Data

gstart = geneLength{k,'Start'};
gcounts = counts(j,:);

% sort according to ascending start position

[~,idx] = sort(gstart);

% plot read coverage by genomic position

figure;
plot(gstart(idx), gcounts(idx,treated),'.-r',...
gstart(idx), gcounts(idx,untreated),'.-b');
xlabel('Genomic position')
ylabel('Read count');
title('Read count on Chromosome 3');

ans =

6360

Normalizing Read Counts

The read count in RNA-Seq data has been found to be linearly related to the abundance of transcripts
[2]. However, the read count for a given gene depends not only on the expression level of the gene,
but also on the total number of reads sequenced and the length of the gene transcript. Therefore, in
order to infer the expression level of a gene from the read count, we need to account for the
sequencing depth and the gene transcript length. One common technique to normalize the read count

2-37
2 High-Throughput Sequence Analysis

is to use the RPKM (Read Per Kilobase Mapped) values, where the read count is normalized by the
total number of reads yielded (in millions) and the length of each transcript (in kilobases). This
normalization technique, however, is not always effective since few, very highly expressed genes can
dominate the total lane count and skew the expression analysis.

A better normalization technique consists of computing the effective library size by considering a size
factor for each sample. By dividing each sample's counts by the corresponding size factors, we bring
all the count values to a common scale, making them comparable. Intuitively, if sample A is
sequenced N times deeper than sample B, the read counts of non-differentially expressed genes are
expected to be on average N times higher in sample A than in sample B, even if there is no difference
in expression.

To estimate the size factors, take the median of the ratios of observed counts to those of a pseudo-
reference sample, whose counts can be obtained by considering the geometric mean of each gene
across all samples [3]. Then, to transform the observed counts to a common scale, divide the
observed counts in each sample by the corresponding size factor.

% estimate pseudo-reference with geometric mean row by row

pseudoRefSample = geomean(counts,2);
nz = pseudoRefSample > 0;
ratios = bsxfun(@rdivide,counts(nz,:),pseudoRefSample(nz));
sizeFactors = median(ratios,1)

sizeFactors =

0.9374 0.9725 0.9388 1.1789

% transform to common scale

normCounts = bsxfun(@rdivide,counts,sizeFactors);
normCounts(1:10,:)

ans =

1.0e+03 *

0 0.0010 0.0011 0.0017

0.1515 0.1203 0.1470 0.1120
0.0213 0.0123 0.0107 0.0161
0.0021 0.0041 0 0.0008
7.0315 5.2721 5.1225 5.1124
0.5003 0.5450 0.5241 0.4869
0.0053 0.0062 0.0107 0.0068
0 0 0.0021 0.0008
1.2375 1.1753 1.2122 1.2003
0 0 0 0.0008

You can appreciate the effect of this normalization by using the function boxplot to represent
statistical measures such as median, quartiles, minimum and maximum.

figure;

subplot(2,1,1)
maboxplot(log2(counts),'title','Raw Read Count','orientation','horizontal')

2-38
 Identifying Differentially Expressed Genes from RNA-Seq Data

ylabel('sample')
xlabel('log2(counts)')

subplot(2,1,2)
maboxplot(log2(normCounts),'title','Normalized Read Count','orientation','horizontal')
ylabel('sample')
xlabel('log2(counts)')

Computing Mean, Dispersion and Fold Change

In order to better characterize the data, we consider the mean and the dispersion of the normalized
counts. The variance of read counts is given by the sum of two terms: the variation across samples
(raw variance) and the uncertainty of measuring the expression by counting reads (shot noise or
Poisson). The raw variance term dominates for highly expressed genes, whereas the shot noise
dominates for lowly expressed genes. You can plot the empirical dispersion values against the mean
of the normalized counts in a log scale as shown below.

% consider the mean

meanTreated = mean(normCounts(:,treated),2);
meanUntreated = mean(normCounts(:,untreated),2);

% consider the dispersion

dispTreated = std(normCounts(:,treated),0,2) ./ meanTreated;
dispUntreated = std(normCounts(:,untreated),0,2) ./ meanUntreated;

% plot on a log-log scale

figure;

2-39
2 High-Throughput Sequence Analysis

loglog(meanTreated,dispTreated,'r.');
hold on;
loglog(meanUntreated,dispUntreated,'b.');
xlabel('log2(Mean)');
ylabel('log2(Dispersion)');
legend('Treated','Untreated','Location','southwest');

Given the small number of replicates, it is not surprising to expect that the dispersion values scatter
with some variance around the true value. Some of this variance reflects sampling variance and some
reflects the true variability among the gene expressions of the samples.

You can look at the difference of the gene expression among two conditions, by calculating the fold
change (FC) for each gene, i.e. the ratio between the counts in the treated group over the counts in
the untreated group. Generally these ratios are considered in the log2 scale, so that any change is
symmetric with respect to zero (e.g. a ratio of 1/2 or 2/1 corresponds to -1 or +1 in the log scale).

% compute the mean and the log2FC

meanBase = (meanTreated + meanUntreated) / 2;
foldChange = meanTreated ./ meanUntreated;
log2FC = log2(foldChange);

% plot mean vs. fold change (MA plot)

mairplot(meanTreated, meanUntreated,'Type','MA','Plotonly',true);
set(get(gca,'Xlabel'),'String','mean of normalized counts')
set(get(gca,'Ylabel'),'String','log2(fold change)')

Warning: Zero values are ignored

2-40
 Identifying Differentially Expressed Genes from RNA-Seq Data

It is possible to annotate the values in the plot with the corresponding gene names, interactively
select genes, and export gene lists to the workspace by calling the mairplot function as illustrated
below:

mairplot(meanTreated,meanUntreated,'Labels',geneCountTable.ID,'Type','MA');

Warning: Zero values are ignored

2-41
2 High-Throughput Sequence Analysis

It is convenient to store the information about the mean value and fold change for each gene in a
table. You can then access information about a given gene or a group of genes satisfying specific
criteria by indexing the table by gene names.

% create table with statistics about each gene

geneTable = table(meanBase,meanTreated,meanUntreated,foldChange,log2FC);
geneTable.Properties.RowNames = geneCountTable.ID;

% summary
summary(geneTable)

2-42
 Identifying Differentially Expressed Genes from RNA-Seq Data

Variables:

meanBase: 11609x1 double

Values:

Min 0.21206
Median 201.24
Max 2.6789e+05

meanTreated: 11609x1 double

Values:

Min 0
Median 201.54
Max 2.5676e+05

meanUntreated: 11609x1 double

Values:

Min 0
Median 199.44
Max 2.7903e+05

foldChange: 11609x1 double

Values:

Min 0
Median 0.99903
Max Inf

log2FC: 11609x1 double

Values:

Min -Inf
Median -0.001406
Max Inf

% preview
head(geneTable)

meanBase meanTreated meanUntreated foldChange log2FC

________ ___________ _____________ __________ _________

FBgn0000003 0.9475 1.3808 0.51415 2.6857 1.4253

FBgn0000008 132.69 129.48 135.9 0.95277 -0.069799
FBgn0000014 15.111 13.384 16.838 0.79488 -0.33119
FBgn0000015 1.7738 0.42413 3.1234 0.13579 -2.8806
FBgn0000017 5634.6 5117.4 6151.8 0.83186 -0.26559
FBgn0000018 514.08 505.48 522.67 0.96711 -0.048243
FBgn0000024 7.2354 8.7189 5.752 1.5158 0.60009

2-43
2 High-Throughput Sequence Analysis

FBgn0000028 0.74465 1.4893 0 Inf Inf

% access information about a specific gene

myGene = 'FBgn0261570';
geneTable(myGene,:)
geneTable(myGene,{'meanBase','log2FC'})

% access information about a given gene list

myGeneSet = ["FBgn0261570","FBgn0261573","FBgn0261575","FBgn0261560"];
geneTable(myGeneSet,:)

ans =

1x5 table

meanBase meanTreated meanUntreated foldChange log2FC

________ ___________ _____________ __________ _______

FBgn0261570 4435.5 4939.1 3931.8 1.2562 0.32907

ans =

1x2 table

meanBase log2FC
________ _______

FBgn0261570 4435.5 0.32907

ans =

4x5 table

meanBase meanTreated meanUntreated foldChange log2FC

________ ___________ _____________ __________ _______

FBgn0261570 4435.5 4939.1 3931.8 1.2562 0.32907

FBgn0261573 2936.9 2954.8 2919.1 1.0122 0.01753
FBgn0261575 4.3776 5.6318 3.1234 1.8031 0.85047
FBgn0261560 2041.1 1494.3 2588 0.57738 -0.7924

Inferring Differential Expression with a Negative Binomial Model

Determining whether the gene expressions in two conditions are statistically different consists of
rejecting the null hypothesis that the two data samples come from distributions with equal means.
This analysis assumes the read counts are modeled according to a negative binomial distribution (as
proposed in [3]). The function rnaseqde performs this type of hypothesis testing with three possible
options to specify the type of linkage between the variance and the mean.

By specifying the link between variance and mean as an identity, we assume the variance is equal to
the mean, and the counts are modeled by the Poisson distribution [4]. "IDColumns" specifies columns
from the input table to append to the output table to help keep data organized.

2-44
 Identifying Differentially Expressed Genes from RNA-Seq Data

diffTableIdentity = rnaseqde(geneCountTable,["untreated3","untreated4"],["treated2","treated3"],V

% Preview the results.

head(diffTableIdentity)

ID Mean1 Mean2 Log2FoldChange PValue AdjustedPValue

_____________ _______ _______ ______________ _________ ______________

"FBgn0000003" 0.51415 1.3808 1.4253 0.627 0.75892

"FBgn0000008" 135.9 129.48 -0.069799 0.48628 0.64516
"FBgn0000014" 16.838 13.384 -0.33119 0.44445 0.61806
"FBgn0000015" 3.1234 0.42413 -2.8806 0.05835 0.12584
"FBgn0000017" 6151.8 5117.4 -0.26559 2.864e-42 6.0233e-41
"FBgn0000018" 522.67 505.48 -0.048243 0.39015 0.5616
"FBgn0000024" 5.752 8.7189 0.60009 0.35511 0.52203
"FBgn0000028" 0 1.4893 Inf 0.252 0.39867

Alternatively, by specifying the variance as the sum of the shot noise term (i.e. mean) and a constant
multiplied by the squared mean, the counts are modeled according to a distribution described in [5].
The constant term is estimated using all the rows in the data.

diffTableConstant = rnaseqde(geneCountTable,["untreated3","untreated4"],["treated2","treated3"],V

% Preview the results.

head(diffTableConstant)

ID Mean1 Mean2 Log2FoldChange PValue AdjustedPValue

_____________ _______ _______ ______________ __________ ______________

"FBgn0000003" 0.51415 1.3808 1.4253 0.62769 0.7944

"FBgn0000008" 135.9 129.48 -0.069799 0.53367 0.72053
"FBgn0000014" 16.838 13.384 -0.33119 0.45592 0.68454
"FBgn0000015" 3.1234 0.42413 -2.8806 0.058924 0.16938
"FBgn0000017" 6151.8 5117.4 -0.26559 8.5529e-05 0.00077269
"FBgn0000018" 522.67 505.48 -0.048243 0.54834 0.73346
"FBgn0000024" 5.752 8.7189 0.60009 0.36131 0.58937
"FBgn0000028" 0 1.4893 Inf 0.2527 0.46047

Finally, by considering the variance as the sum of the shot noise term (i.e. mean) and a locally
regressed non-parametric smooth function of the mean, the counts are modeled according to the
distribution proposed in [3].

diffTableLocal = rnaseqde(geneCountTable,["untreated3","untreated4"],["treated2", "treated3"],Var

% Preview the results.

head(diffTableLocal)

ID Mean1 Mean2 Log2FoldChange PValue AdjustedPValue

_____________ _______ _______ ______________ _________ ______________

"FBgn0000003" 0.51415 1.3808 1.4253 1 1

"FBgn0000008" 135.9 129.48 -0.069799 0.67298 0.89231
"FBgn0000014" 16.838 13.384 -0.33119 0.6421 0.87234
"FBgn0000015" 3.1234 0.42413 -2.8806 0.22776 0.57215
"FBgn0000017" 6151.8 5117.4 -0.26559 0.0014429 0.014207
"FBgn0000018" 522.67 505.48 -0.048243 0.65307 0.88136

2-45
2 High-Throughput Sequence Analysis

"FBgn0000024" 5.752 8.7189 0.60009 0.55154 0.81984

"FBgn0000028" 0 1.4893 Inf 0.42929 0.7765

The output of rnaseqde includes a vector of P-values. A P-value indicates the probability that a
change in expression as strong as the one observed (or even stronger) would occur under the null
hypothesis, i.e. the conditions have no effect on gene expression. In the histogram of the P-values we
observe an enrichment of low values (due to differentially expressed genes), whereas other values are
uniformly spread (due to non-differentially expressed genes). The enrichment of values equal to 1 are
due to genes with very low counts.

figure;
histogram(diffTableLocal.PValue,100)
xlabel('P-value')
ylabel('Frequency')
title('P-value enrichment')

Filter out those genes with relatively low count to observe a more uniform spread of non-significant P-
values across the range (0,1]. Note that this does not affect the distribution of significant P-values.

lowCountThreshold = 10;
lowCountGenes = all(counts < lowCountThreshold, 2);
histogram(diffTableLocal.PValue(~lowCountGenes),100)
xlabel('P-value')
ylabel('Frequency')
title('P-value enrichment without low count genes')

2-46
 Identifying Differentially Expressed Genes from RNA-Seq Data

Multiple Testing and Adjusted P-values

Thresholding P-values to determine what fold changes are more significant than others is not
appropriate for this type of data analysis, due to the multiple testing problem. While performing a
large number of simultaneous tests, the probability of getting a significant result simply due to
chance increases with the number of tests. In order to account for multiple testing, perform a
correction (or adjustment) of the P-values so that the probability of observing at least one significant
result due to chance remains below the desired significance level.

The Benjamini-Hochberg adjustment [6] is a statistical method that provides an adjusted P-value
answering the following question: what would be the fraction of false positives if all the genes with
adjusted P-values below a given threshold were considered significant?

The output of rnaseqde includes a vector of adjusted P-values in the "AdjustedPValue" field. By
default, the P-values are adjusted using the Benjamini-Hochberg adjustment. Alternatively, the
"FDRMethod" Name-Value argument in rnaseqde can be set to "storey" to perform Storey's
procedure [7].

Set a threshold of 0.1 for the adjusted P-values, equivalent to consider a 10% false positives as
acceptable, and identify the genes that are significantly expressed by considering all the genes with
adjusted P-values below this threshold.
% create a table with significant genes
sig = diffTableLocal.AdjustedPValue < 0.1;
diffTableLocalSig = diffTableLocal(sig,:);
diffTableLocalSig = sortrows(diffTableLocalSig,'AdjustedPValue');
numberSigGenes = size(diffTableLocalSig,1)

2-47
2 High-Throughput Sequence Analysis

numberSigGenes =

1904

Identifying the Most Up-regulated and Down-regulated Genes

You can now identify the most up-regulated or down-regulated genes by considering an absolute fold
change above a chosen cutoff. For example, a cutoff of 1 in log2 scale yields the list of genes that are
up-regulated with a 2 fold change.

% find up-regulated genes

up = diffTableLocalSig.Log2FoldChange > 1;
upGenes = sortrows(diffTableLocalSig(up,:),'Log2FoldChange','descend');
numberSigGenesUp = sum(up)

% display the top 10 up-regulated genes

top10GenesUp = upGenes(1:10,:)

% find down-regulated genes

down = diffTableLocalSig.Log2FoldChange < -1;
downGenes = sortrows(diffTableLocalSig(down,:),'Log2FoldChange','ascend');
numberSigGenesDown = sum(down)

% find top 10 down-regulated genes

top10GenesDown = downGenes(1:10,:)

numberSigGenesUp =

129

top10GenesUp =

10x6 table

ID Mean1 Mean2 Log2FoldChange PValue AdjustedPValue

_____________ ______ ______ ______________ __________ ______________

"FBgn0030173" 0 6.7957 Inf 0.0063115 0.047764

"FBgn0036822" 0 6.2729 Inf 0.012203 0.079274
"FBgn0052548" 1.0476 15.269 3.8654 0.00016945 0.0022662
"FBgn0050495" 1.0283 12.635 3.6191 0.0018949 0.017972
"FBgn0063667" 3.1042 38.042 3.6153 8.5037e-08 2.3845e-06
"FBgn0033764" 16.324 167.61 3.3601 1.8345e-25 2.9174e-23
"FBgn0037290" 16.228 155.46 3.26 3.5583e-23 4.6941e-21
"FBgn0033733" 1.5424 13.384 3.1172 0.0027276 0.024283
"FBgn0037191" 1.6003 12.753 2.9945 0.0047803 0.038193
"FBgn0033943" 1.581 12.319 2.962 0.0053635 0.041986

numberSigGenesDown =

181

2-48
 Identifying Differentially Expressed Genes from RNA-Seq Data

top10GenesDown =

10x6 table

ID Mean1 Mean2 Log2FoldChange PValue AdjustedPValue

_____________ ______ _______ ______________ __________ ______________

"FBgn0053498" 30.938 0 -Inf 9.8404e-11 4.345e-09

"FBgn0259236" 13.618 0 -Inf 1.5526e-05 0.00027393
"FBgn0052500" 8.7405 0 -Inf 0.00066783 0.0075343
"FBgn0039331" 7.3908 0 -Inf 0.0019558 0.018474
"FBgn0040697" 6.8381 0 -Inf 0.0027378 0.024336
"FBgn0034972" 5.8291 0 -Inf 0.0068564 0.05073
"FBgn0040967" 5.2764 0 -Inf 0.0096039 0.065972
"FBgn0031923" 4.7429 0 -Inf 0.016164 0.098762
"FBgn0085359" 121.97 2.9786 -5.3557 5.5813e-33 1.5068e-30
"FBgn0004854" 14.402 0.53259 -4.7571 8.1587e-05 0.0012034

A good visualization of the gene expressions and their significance is given by plotting the fold
change versus the mean in log scale and coloring the data points according to the adjusted P-values.

figure
scatter(log2(geneTable.meanBase),diffTableLocal.Log2FoldChange,3,diffTableLocal.PValue,'o')
colormap(flipud(cool(256)))
colorbar;
ylabel('log2(Fold Change)')
xlabel('log2(Mean of normalized counts)')
title('Fold change by FDR')

2-49
2 High-Throughput Sequence Analysis

You can see here that for weakly expressed genes (i.e. those with low means), the FDR is generally
high because low read counts are dominated by Poisson noise and consequently any biological
variability is drowned in the uncertainties from the read counting.

References

[1] Brooks et al. Conservation of an RNA regulatory map between Drosophila and mammals. Genome
Research 2011. 21:193-202.

[2] Mortazavi et al. Mapping and quantifying mammalian transcriptomes by RNA-Seq. Nature
Methods 2008. 5:621-628.

[3] Anders et al. Differential expression analysis for sequence count data. Genome Biology 2010.
11:R106.

[4] Marioni et al. RNA-Seq: An assessment of technical reproducibility and comparison with gene
expression arrays. Genome Research 2008. 18:1509-1517.

[5] Robinson et al. Moderated statistical test for assessing differences in tag abundance.
Bioinformatics 2007. 23(21):2881-2887.

[6] Benjamini et al. Controlling the false discovery rate: a practical and powerful approach to multiple
testing. 1995. Journal of the Royal Statistical Society, Series B57 (1):289-300.

2-50
 Identifying Differentially Expressed Genes from RNA-Seq Data

[7] J.D. Storey. "A direct approach to false discovery rates", Journal of the Royal Statistical Society, B
(2002), 64(3), pp.479-498.

See Also
featurecount | nbintest | mairplot | plotVarianceLink

More About
• “High-Throughput Sequencing”

2-51
2 High-Throughput Sequence Analysis

Visualize NGS Data Using Genomics Viewer App

The Genomics Viewer app lets you view and explore integrated genomic data with an embedded
version of the Integrative Genomics Viewer (IGV) [1][2]. The genomic data include NGS read
alignments, genome variants, and segmented copy number data.

The first part of this example gives a brief overview of the app and supported file formats. The second
part of the example explores a single nucleotide variation in the cytochrome p450 gene (CYP2C19).

Open the App

At the command line, type genomicsViewer. Alternatively, click the app icon on the Apps tab. The
app requires an internet connection.

By default, the app loads Human (GRCh38/hg38) as the reference sequence and Refseq Genes as the
annotation file. There are two main panels in the app. The left panel is the Tracks panel and the right
panel is the embedded IGV web application. The Tracks panel is a read-only area displaying the track
names, source file names, and track types. The Tracks panel updates accordingly as you configure
the tracks in the embedded IGV app.

The Reset button restores the app to the default view with two tracks (HG38 with Refseq Genes) and
removes any other existing tracks. Before resetting, you can save the current view as a session
(.json) file and restore it later.

Add Tracks by Importing Data

Import Reference Sequence

You can import a single reference sequence. The reference sequence must be in a FASTA file. Select
Import Reference on the Home tab. You can also import a corresponding cytoband file that contains

2-52
 Visualize NGS Data Using Genomics Viewer App

cytogenetic G-banding data. You can add local files or specify external URLs. The URL must start with
either https or gs. Other file transfer protocols, such as ftp, are not supported.

Import Sequence Read Alignment Data

You can import multiple data sets of sequence read alignment data. The alignment data must be a
BAM or CRAM file. It is not required that you have the corresponding index file (.BAI or .CRAI) in
the same location as your BAM or CRAM file. However, the absence of the index file will make the
app slower.

You can add read alignment files using Add tracks from file and Add tracks from URL options
from the Add tracks button. If you are specifying a URL, the URL must start with either https or gs.
Other file transfer protocols, such as ftp, are not supported.

Import Feature Annotations and Other Genomic Data

You can import multiple sets of feature annotations from several files that contain data for a single
reference sequence. The supported annotation files are: .BED, .GFF, .GFF3, and .GTF.

You can also import structural variants (.VCF) and visualize genetic alterations, such as insertions
and deletions.

You can view segmented copy number data (.SEG) and quantitative genomic data (.WIG, .BIGWIG,
and .BEDGRAPH), such as ChIP peaks and alignment coverage.

You can add annotation and genomic data files using Add tracks from file and Add tracks from
URL options from the Add tracks button. If you are specifying a URL, the URL must start with either
https or gs. Other file transfer protocols, such as FTP, are not supported.

Visualize Single Nucleotide Variation in Cytochrome P450

The CYP2C19 gene is a member of the cytochrome P450 gene family. Enzymes produced from
cytochrome P450 genes are involved in the metabolism of various molecules and chemicals within
cells. The CYP2C19 enzyme plays a role in the metabolizing of at least 10 percent of commonly
prescribed drugs [3]. Polymorphisms in the cytochrome p450 family may cause adverse drug
responses in individuals. One example of single nucleotide variation is rs4986893 at position
chr10:94,780,653 where G is replaced by A. This allelic variant is also known as CYP2C19*3. The
following steps show how to visualize such variation in the app using both low coverage and high
coverage data.

Load Session File

For the purposes of this example, start with a session file that has some preloaded tracks. To load the
file, click Open. Navigate to matlabroot\examples\bioinfo\data, where matlabroot is the
folder where you have installed MATLAB. Select rs4986893.json.

2-53
2 High-Throughput Sequence Analysis

The session contains three tracks:

• Human (GRCh38/hg38) as a reference

• NA18564 as low coverage alignment data
• Refseq Genes

The low coverage alignment data comes from a female Han Chinese from Beijing, China. The sample
ID is NA18564 and the sample has been identified with the CYP2C19*3 mutation [4].

Explore Low Coverage Data

In this session file, the alignment data has been centered around the location of the mutation on the
CYP2C19 gene.

1 Click the orange bar in the coverage area to look at the position and allele distribution
information.

2-54
 Visualize NGS Data Using Genomics Viewer App

It shows that 71% of the reads have G while 29% have A at the location chr10:94,780,653. This
data is a low coverage data and may not show all the occurrences of this mutation. A high
coverage data will be explored later in the example.

Close the data tip window.

2 You can customize the various aspects of the data display in the app. For example, you can
change the track height to make more room for later tracks. Click the second gear icon. Select
Set track height. Enter 200.

2-55
2 High-Throughput Sequence Analysis

For details on the embedded IGV app and its available options, visit here.

Explore High Coverage Data

You can look at the high coverage data from the same sample to see the occurrences of this mutation.

1 Go to The International Genome Sample Resource website.

2 Search for the sample NA18564.
3 Download the Exome alignment file that is in the .CRAM format.
4 Also download the corresponding index file that is in the .CRAI format. Save the file in the same
location as the source .CRAM file.
5 Click the (+) icon on the Home tab. Select the downloaded .CRAM file and click Open.

The high coverage data appears as track3. You can now see many occurrences of the mutation in
several reads.
6 Click the orange bar in the coverage area to see the allele distribution. It shows that G is
replaced by A in almost 50% of the time.

2-56
 Visualize NGS Data Using Genomics Viewer App

References
[1] Robinson, J., H. Thorvaldsdóttir, W. Winckler, M. Guttman, E. Lander, G. Getz, J. Mesirov. 2011.
Integrative Genomics Viewer. Nature Biotechnology. 29:24–26.

[2] Thorvaldsdóttir, H., J. Robinson, J. Mesirov. 2013. Integrative Genomics Viewer (IGV): High-
performance genomics data visualization and exploration. Briefings in Bioinformatics.
14:178–192.

[3] https://medlineplus.gov/genetics/gene/cyp2c19/

[4] https://www.coriell.org/0/Sections/Search/Sample_Detail.aspx?Ref=NA18564&Product=DNA

See Also
Genomics Viewer | Sequence Alignment | Sequence Viewer

2-57
2 High-Throughput Sequence Analysis

Exploring Genome-Wide Differences in DNA Methylation

Profiles

This example shows how to perform a genome-wide analysis of DNA methylation in the human by
using genome sequencing.

Note: For enhanced performance, MathWorks recommends that you run this example on a 64-bit
platform, because the memory footprint is close to 2 GB. On a 32-bit platform, if you receive "Out of
memory" errors when running this example, try increasing the virtual memory (or swap space) of
your operating system or try setting the 3GB switch (32-bit Windows® XP only). For details, see
“Resolve “Out of Memory” Errors”.

Introduction

DNA methylation is an epigenetic modification that modulates gene expression and the maintenance
of genomic organization in normal and disease processes. DNA methylation can define different
states of the cell, and it is inheritable during cell replication. Aberrant DNA methylation patterns
have been associated with cancer and tumor suppressor genes.

In this example you will explore the DNA methylation profiles of two human cancer cells: parental
HCT116 colon cancer cells and DICERex5 cells. DICERex5 cells are derived from HCT116 cells after
the truncation of the DICER1 alleles. Serre et al. in [1] proposed to study DNA methylation profiles by
using the MBD2 protein as a methyl CpG binding domain and subsequently used high-throughput
sequencing (HTseq). This technique is commonly known as MBD-Seq. Short reads for two replicates
of the two samples have been submitted to NCBI's SRA archive by the authors of [1]. There are other
technologies available to interrogate DNA methylation status of CpG sites in combination with HTseq,
for example MeDIP-seq or the use of restriction enzymes. You can also analyze this type of data sets
following the approach presented in this example.

Data Sets

You can obtain the unmapped single-end reads for four sequencing experiments from NCBI. Short
reads were produced using Illumina®'s Genome Analyzer II. Average insert size is 120 bp, and the
length of short reads is 36 bp.

This example assumes that you:

(1) downloaded the files SRR030222.sra, SRR030223.sra, SRR030224.sra and SRR030225.sra

containing the unmapped short reads for two replicates of from the DICERex5 sample and two
replicates from the HCT116 sample respectively, from NCBI SRA Run Selector and converted them to
FASTQ-formatted files using the NCBI SRA Toolkit.

(2) produced SAM-formatted files by mapping the short reads to the reference human genome (NCBI
Build 37.5) using the Bowtie [2] algorithm. Only uniquely mapped reads are reported.

(3) compressed the SAM formatted files to BAM and ordered them by reference name first, then by
genomic position by using SAMtools [3].

This example also assumes that you downloaded the reference human genome (GRCh37.p5). You can
use the bowtie-inspect command to reconstruct the human reference directly from the bowtie
indices. Or you may download the reference from the NCBI repository by uncommenting the
following line:

2-58
 Exploring Genome-Wide Differences in DNA Methylation Profiles

% getgenbank('NC_000009','FileFormat','fasta','tofile','hsch9.fasta');

Creating a MATLAB® Interface to the BAM-Formatted Files

To explore the signal coverage of the HCT116 samples you need to construct a BioMap. BioMap has
an interface that provides direct access to the mapped short reads stored in the BAM-formatted file,
thus minimizing the amount of data that is actually loaded into memory. Use the function baminfo to
obtain a list of the existing references and the actual number of short reads mapped to each one.

info = baminfo('SRR030224.bam','ScanDictionary',true);
fprintf('%-35s%s\n','Reference','Number of Reads');
for i = 1:numel(info.ScannedDictionary)
fprintf('%-35s%d\n',info.ScannedDictionary{i},...
info.ScannedDictionaryCount(i));
end

Reference Number of Reads

gi|224589800|ref|NC_000001.10| 205065
gi|224589811|ref|NC_000002.11| 187019
gi|224589815|ref|NC_000003.11| 73986
gi|224589816|ref|NC_000004.11| 84033
gi|224589817|ref|NC_000005.9| 96898
gi|224589818|ref|NC_000006.11| 87990
gi|224589819|ref|NC_000007.13| 120816
gi|224589820|ref|NC_000008.10| 111229
gi|224589821|ref|NC_000009.11| 106189
gi|224589801|ref|NC_000010.10| 112279
gi|224589802|ref|NC_000011.9| 104466
gi|224589803|ref|NC_000012.11| 87091
gi|224589804|ref|NC_000013.10| 53638
gi|224589805|ref|NC_000014.8| 64049
gi|224589806|ref|NC_000015.9| 60183
gi|224589807|ref|NC_000016.9| 146868
gi|224589808|ref|NC_000017.10| 195893
gi|224589809|ref|NC_000018.9| 60344
gi|224589810|ref|NC_000019.9| 166420
gi|224589812|ref|NC_000020.10| 148950
gi|224589813|ref|NC_000021.8| 310048
gi|224589814|ref|NC_000022.10| 76037
gi|224589822|ref|NC_000023.10| 32421
gi|224589823|ref|NC_000024.9| 18870
gi|17981852|ref|NC_001807.4| 1015
Unmapped 6805842

In this example you will focus on the analysis of chromosome 9. Create a BioMap for the two HCT116
sample replicates.

bm_hct116_1 = BioMap('SRR030224.bam','SelectRef','gi|224589821|ref|NC_000009.11|')
bm_hct116_2 = BioMap('SRR030225.bam','SelectRef','gi|224589821|ref|NC_000009.11|')

bm_hct116_1 =

BioMap with properties:

SequenceDictionary: 'gi|224589821|ref|NC_000009.11|'
Reference: [106189x1 File indexed property]
Signature: [106189x1 File indexed property]

2-59
2 High-Throughput Sequence Analysis

Start: [106189x1 File indexed property]

MappingQuality: [106189x1 File indexed property]
Flag: [106189x1 File indexed property]
MatePosition: [106189x1 File indexed property]
Quality: [106189x1 File indexed property]
Sequence: [106189x1 File indexed property]
Header: [106189x1 File indexed property]
NSeqs: 106189
Name: ''

bm_hct116_2 =

BioMap with properties:

SequenceDictionary: 'gi|224589821|ref|NC_000009.11|'
Reference: [107586x1 File indexed property]
Signature: [107586x1 File indexed property]
Start: [107586x1 File indexed property]
MappingQuality: [107586x1 File indexed property]
Flag: [107586x1 File indexed property]
MatePosition: [107586x1 File indexed property]
Quality: [107586x1 File indexed property]
Sequence: [107586x1 File indexed property]
Header: [107586x1 File indexed property]
NSeqs: 107586
Name: ''

Using a binning algorithm provided by the getBaseCoverage method, you can plot the coverage of
both replicates for an initial inspection. For reference, you can also add the ideogram for the human
chromosome 9 to the plot using the chromosomeplot function.

figure
ha = gca;
hold on
n = 141213431; % length of chromosome 9
[cov,bin] = getBaseCoverage(bm_hct116_1,1,n,'binWidth',100);
h1 = plot(bin,cov,'b'); % plots the binned coverage of bm_hct116_1
[cov,bin] = getBaseCoverage(bm_hct116_2,1,n,'binWidth',100);
h2 = plot(bin,cov,'g'); % plots the binned coverage of bm_hct116_2
chromosomeplot('hs_cytoBand.txt', 9, 'AddToPlot', ha) % plots an ideogram along the x-axis
axis(ha,[1 n 0 100]) % zooms-in the y-axis
fixGenomicPositionLabels(ha) % formats tick labels and adds datacursors
legend([h1 h2],'HCT116-1','HCT116-2','Location','NorthEast')
ylabel('Coverage')
title('Coverage for two replicates of the HCT116 sample')
fig = gcf;
fig.Position = max(fig.Position,[0 0 900 0]); % resize window

2-60
 Exploring Genome-Wide Differences in DNA Methylation Profiles

Because short reads represent the methylated regions of the DNA, there is a correlation between
aligned coverage and DNA methylation. Observe the increased DNA methylation close to the
chromosome telomeres; it is known that there is an association between DNA methylation and the
role of telomeres for maintaining the integrity of the chromosomes. In the coverage plot you can also
see a long gap over the chromosome centromere. This is due to the repetitive sequences present in
the centromere, which prevent us from aligning short reads to a unique position in this region. In the
data sets used in this example only about 30% of the short reads were uniquely mapped to the
reference genome.

Correlating CpG Islands and DNA Methylation

DNA methylation normally occurs in CpG dinucleotides. Alteration of the DNA methylation patterns
can lead to transcriptional silencing, especially in the gene promoter CpG islands. But, it is also
known that DNA methylation can block CTCF binding and can silence miRNA transcription among
other relevant functions. In general, it is expected that mapped reads should preferably align to CpG
rich regions.

Load the human chromosome 9 from the reference file hs37.fasta. For this example, it is assumed
that you recovered the reference from the Bowtie indices using the bowtie-inspect command;
therefore hs37.fasta contains all the human chromosomes. To load only the chromosome 9 you can
use the option nave-value pair BLOCKREAD with the fastaread function.

chr9 = fastaread('hs37.fasta','blockread',9);
chr9.Header

ans =

'gi|224589821|ref|NC_000009.11| Homo sapiens chromosome 9, GRCh37 primary reference assembly'

2-61
2 High-Throughput Sequence Analysis

Use the cpgisland function to find the CpG clusters. Using the standard definition for CpG islands
[4], 200 or more bp islands with 60% or greater CpGobserved/CpGexpected ratio, leads to 1682 GpG
islands found in chromosome 9.

cpgi = cpgisland(chr9.Sequence)

cpgi =

struct with fields:

Starts: [10783 29188 73049 73686 113309 114488 116877 117469 117987 … ]
Stops: [11319 29409 73624 73893 114336 114809 117105 117985 118203 … ]

Use the getCounts method to calculate the ratio of aligned bases that are inside CpG islands. For
the first replicate of the sample HCT116, the ratio is close to 45%.

aligned_bases_in_CpG_islands = getCounts(bm_hct116_1,cpgi.Starts,cpgi.Stops,'method','sum')
aligned_bases_total = getCounts(bm_hct116_1,1,n,'method','sum')
ratio = aligned_bases_in_CpG_islands ./ aligned_bases_total

aligned_bases_in_CpG_islands =

1724363

aligned_bases_total =

3822804

ratio =

0.4511

You can explore high resolution coverage plots of the two sample replicates and observe how the
signal correlates with the CpG islands. For example, explore the region between 23,820,000 and
23,830,000 bp. This is the 5' region of the human gene ELAVL2.

r1 = 23820001; % set the region limits

r2 = 23830000;
fhELAVL2 = figure; % keep the figure handle to use it later
hold on
% plot high-resolution coverage of bm_hct116_1
h1 = plot(r1:r2,getBaseCoverage(bm_hct116_1,r1,r2,'binWidth',1),'b');
% plot high-resolution coverage of bm_hct116_2
h2 = plot(r1:r2,getBaseCoverage(bm_hct116_2,r1,r2,'binWidth',1),'g');

% mark the CpG islands within the [r1 r2] region

for i = 1:numel(cpgi.Starts)
if cpgi.Starts(i)>r1 && cpgi.Stops(i)<r2 % is CpG island inside [r1 r2]?
px = [cpgi.Starts([i i]) cpgi.Stops([i i])]; % x-coordinates for patch
py = [0 max(ylim) max(ylim) 0]; % y-coordinates for patch
hp = patch(px,py,'r','FaceAlpha',.1,'EdgeColor','r','Tag','cpgi');
end

2-62
 Exploring Genome-Wide Differences in DNA Methylation Profiles

end

axis([r1 r2 0 20]) % zooms-in the y-axis

fixGenomicPositionLabels(gca) % formats tick labels and adds datacursors
legend([h1 h2 hp],'HCT116-1','HCT116-2','CpG Islands')
ylabel('Coverage')
xlabel('Chromosome 9 position')
title('Coverage for two replicates of the HCT116 sample')

Statistical Modelling of Count Data

To find regions that contain more mapped reads than would be expected by chance, you can follow a
similar approach to the one described by Serre et al. [1]. The number of counts for non-overlapping
contiguous 100 bp windows is statistically modeled.

First, use the getCounts method to count the number of mapped reads that start at each window. In
this example you use a binning approach that considers only the start position of every mapped read,
following the approach of Serre et al. However, you may also use the OVERLAP and METHOD name-
value pairs in getCounts to compute more accurate statistics. For instance, to obtain the maximum
coverage for each window considering base pair resolution, set OVERLAP to 1 and METHOD to MAX.
n = numel(chr9.Sequence); % length of chromosome
w = 1:100:n; % windows of 100 bp

counts_1 = getCounts(bm_hct116_1,w,w+99,'independent',true,'overlap','start');
counts_2 = getCounts(bm_hct116_2,w,w+99,'independent',true,'overlap','start');

First, try to model the counts assuming that all the windows with counts are biologically significant
and therefore from the same distribution. Use the negative bionomial distribution to fit a model the
count data.
nbp = nbinfit(counts_1);

Plot the fitted model over a histogram of the empirical data.

2-63
2 High-Throughput Sequence Analysis

figure
hold on
emphist = histc(counts_1,0:100); % calculate the empirical distribution
bar(0:100,emphist./sum(emphist),'c','grouped') % plot histogram
plot(0:100,nbinpdf(0:100,nbp(1),nbp(2)),'b','linewidth',2); % plot fitted model
axis([0 50 0 .001])
legend('Empirical Distribution','Negative Binomial Fit')
ylabel('Frequency')
xlabel('Counts')
title('Frequency of counts, 100bp windows (HCT116-1)')

The poor fitting indicates that the observed distribution may be due to the mixture of two models, one
that represents the background and one that represents the count data in methylated DNA windows.

A more realistic scenario would be to assume that windows with a small number of mapped reads are
mainly the background (or null model). Serre et al. assumed that 100-bp windows containing four or
more reads are unlikely to be generated by chance. To estimate a good approximation to the null
model, you can fit the left body of the empirical distribution to a truncated negative binomial
distribution. To fit a truncated distribution use the mle function. First you need to define an
anonymous function that defines the right-truncated version of nbinpdf.

rtnbinpdf = @(x,p1,p2,t) nbinpdf(x,p1,p2) ./ nbincdf(t-1,p1,p2);

Define the fitting function using another anonymous function.

rtnbinfit = @(x2,t) mle(x2,'pdf',@(x3,p1,p2) rtnbinpdf(x3,p1,p2,t),'start',nbinfit(x2),'lower',[0

2-64
 Exploring Genome-Wide Differences in DNA Methylation Profiles

Before fitting the real data, let us assess the fitting procedure with some sampled data from a known
distribution.

nbp = [0.5 0.2]; % Known coefficients

x = nbinrnd(nbp(1),nbp(2),10000,1); % Random sample
trun = 6; % Set a truncation threshold

nbphat1 = nbinfit(x); % Fit non-truncated model to all data

nbphat2 = nbinfit(x(x<trun)); % Fit non-truncated model to truncated data (wrong)
nbphat3 = rtnbinfit(x(x<trun),trun); % Fit truncated model to truncated data

figure
hold on
emphist = histc(x,0:100); % Calculate the empirical distribution
bar(0:100,emphist./sum(emphist),'c','grouped') % plot histogram
h1 = plot(0:100,nbinpdf(0:100,nbphat1(1),nbphat1(2)),'b-o','linewidth',2);
h2 = plot(0:100,nbinpdf(0:100,nbphat2(1),nbphat2(2)),'r','linewidth',2);
h3 = plot(0:100,nbinpdf(0:100,nbphat3(1),nbphat3(2)),'g','linewidth',2);
axis([0 25 0 .2])
legend([h1 h2 h3],'Neg-binomial fitted to all data',...
'Neg-binomial fitted to truncated data',...
'Truncated neg-binomial fitted to truncated data')
ylabel('Frequency')
xlabel('Counts')

2-65
2 High-Throughput Sequence Analysis

Identifying Significant Methylated Regions

For the two replicates of the HCT116 sample, fit a right-truncated negative binomial distribution to
the observed null model using the rtnbinfit anonymous function previously defined.

trun = 4; % Set a truncation threshold (as in [1])

pn1 = rtnbinfit(counts_1(counts_1<trun),trun); % Fit to HCT116-1 counts
pn2 = rtnbinfit(counts_2(counts_2<trun),trun); % Fit to HCT116-2 counts

Calculate the p-value for each window to the null distribution.

pval1 = 1 - nbincdf(counts_1,pn1(1),pn1(2));
pval2 = 1 - nbincdf(counts_2,pn2(1),pn2(2));

Calculate the false discovery rate using the mafdr function. Use the name-value pair BHFDR to use
the linear-step up (LSU) procedure ([6]) to calculate the FDR adjusted p-values. Setting the FDR <
0.01 permits you to identify the 100-bp windows that are significantly methylated.

fdr1 = mafdr(pval1,'bhfdr',true);
fdr2 = mafdr(pval2,'bhfdr',true);

w1 = fdr1<.01; % logical vector indicating significant windows in HCT116-1

w2 = fdr2<.01; % logical vector indicating significant windows in HCT116-2
w12 = w1 & w2; % logical vector indicating significant windows in both replicates

Number_of_sig_windows_HCT116_1 = sum(w1)
Number_of_sig_windows_HCT116_2 = sum(w2)
Number_of_sig_windows_HCT116 = sum(w12)

Number_of_sig_windows_HCT116_1 =

1662

Number_of_sig_windows_HCT116_2 =

1674

Number_of_sig_windows_HCT116 =

1346

Overall, you identified 1662 and 1674 non-overlapping 100-bp windows in the two replicates of the
HCT116 samples, which indicates there is significant evidence of DNA methylation. There are 1346
windows that are significant in both replicates.

For example, looking again in the promoter region of the ELAVL2 human gene you can observe that
in both sample replicates, multiple 100-bp windows have been marked significant.

figure(fhELAVL2) % bring back to focus the previously plotted figure

plot(w(w1)+50,counts_1(w1),'bs', 'HandleVisibility','off') % plot significant windows in HCT116-1
plot(w(w2)+50,counts_2(w2),'gs', 'HandleVisibility','off') % plot significant windows in HCT116-2
axis([r1 r2 0 100])
title('Significant 100-bp windows in both replicates of the HCT116 sample')

2-66
 Exploring Genome-Wide Differences in DNA Methylation Profiles

Finding Genes With Significant Methylated Promoter Regions

DNA methylation is involved in the modulation of gene expression. For instance, it is well known that
hypermethylation is associated with the inactivation of several tumor suppressor genes. You can
study in this data set the methylation of gene promoter regions.

First, download from Ensembl a tab-separated-value (TSV) table with all protein encoding genes to a
text file, ensemblmart_genes_hum37.txt. For this example, we are using Ensembl release 64.
Using Ensembl's BioMart service, you can select a table with the following attributes: chromosome
name, gene biotype, gene name, gene start/end, and strand direction.

Use the provided helper function ensemblmart2gff to convert the downloaded TSV file to a GFF
formatted file. Then use GFFAnnotation to load the file into MATLAB and create a subset with the
genes present in chromosome 9 only. This results 800 annotated protein-coding genes in the Ensembl
database.
GFFfilename = ensemblmart2gff('ensemblmart_genes_hum37.txt');
a = GFFAnnotation(GFFfilename)
a9 = getSubset(a,'reference','9')
numGenes = a9.NumEntries

a =

GFFAnnotation with properties:

FieldNames: {1×9 cell}

NumEntries: 21184

a9 =

GFFAnnotation with properties:

2-67
2 High-Throughput Sequence Analysis

FieldNames: {1×9 cell}

NumEntries: 800

numGenes =

800

Find the promoter regions for each gene. In this example we consider the proximal promoter as the
-500/100 upstream region.

downstream = 500;
upstream = 100;

geneDir = strcmp(a9.Strand,'+'); % logical vector indicating strands in the forward direction

% calculate promoter's start position for genes in the forward direction

promoterStart(geneDir) = a9.Start(geneDir) - downstream;
% calculate promoter's end position for genes in the forward direction
promoterStop(geneDir) = a9.Start(geneDir) + upstream;
% calculate promoter's start position for genes in the reverse direction
promoterStart(~geneDir) = a9.Stop(~geneDir) - upstream;
% calculate promoter's end position for genes in the reverse direction
promoterStop(~geneDir) = a9.Stop(~geneDir) + downstream;

Use a dataset as a container for the promoter information, as we can later add new columns to
store gene counts and p-values.

promoters = dataset({a9.Feature,'Gene'});
promoters.Strand = char(a9.Strand);
promoters.Start = promoterStart';
promoters.Stop = promoterStop';

Find genes with significant DNA methylation in the promoter region by looking at the number of
mapped short reads that overlap at least one base pair in the defined promoter region.

promoters.Counts_1 = getCounts(bm_hct116_1,promoters.Start,promoters.Stop,...
'overlap',1,'independent',true);
promoters.Counts_2 = getCounts(bm_hct116_2,promoters.Start,promoters.Stop,...
'overlap',1,'independent',true);

Fit a null distribution for each sample replicate and compute the p-values:

trun = 5; % Set a truncation threshold

pn1 = rtnbinfit(promoters.Counts_1(promoters.Counts_1<trun),trun); % Fit to HCT116-1 promoter cou
pn2 = rtnbinfit(promoters.Counts_2(promoters.Counts_2<trun),trun); % Fit to HCT116-2 promoter cou
promoters.pval_1 = 1 - nbincdf(promoters.Counts_1,pn1(1),pn1(2)); % p-value for every promoter in
promoters.pval_2 = 1 - nbincdf(promoters.Counts_2,pn2(1),pn2(2)); % p-value for every promoter in

Number_of_sig_promoters = sum(promoters.pval_1<.01 & promoters.pval_2<.01)

Ratio_of_sig_methylated_promoters = Number_of_sig_promoters./numGenes

Number_of_sig_promoters =

74

2-68
 Exploring Genome-Wide Differences in DNA Methylation Profiles

Ratio_of_sig_methylated_promoters =

0.0925

Observe that only 74 (out of 800) genes in chromosome 9 have significantly DNA methylated regions
(pval<0.01 in both replicates). Display a report of the 30 genes with the most significant methylated
promoter regions.

[~,order] = sort(promoters.pval_1.*promoters.pval_2);
promoters(order(1:30),[1 2 3 4 5 7 6 8])

ans =

Gene Strand Start Stop Counts_1

{'DMRT3' } + 976464 977064 223
{'CNTFR' } - 34590021 34590621 219
{'GABBR2' } - 101471379 101471979 404
{'CACNA1B' } + 140771741 140772341 454
{'BARX1' } - 96717554 96718154 264
{'FAM78A' } - 134151834 134152434 497
{'FOXB2' } + 79634071 79634671 163
{'TLE4' } + 82186188 82186788 157
{'ASTN2' } - 120177248 120177848 141
{'FOXE1' } + 100615036 100615636 149
{'MPDZ' } - 13279489 13280089 129
{'PTPRD' } - 10612623 10613223 145
{'PALM2-AKAP2'} + 112542089 112542689 134
{'FAM69B' } + 139606522 139607122 112
{'WNK2' } + 95946698 95947298 108
{'IGFBPL1' } - 38424344 38424944 110
{'AKAP2' } + 112542269 112542869 107
{'C9orf4' } - 111929471 111930071 102
{'COL5A1' } + 137533120 137533720 84
{'LHX3' } - 139096855 139097455 74
{'OLFM1' } + 137966768 137967368 75
{'NPR2' } + 35791651 35792251 68
{'DBC1' } - 122131645 122132245 61
{'SOHLH1' } - 138591274 138591874 56
{'PIP5K1B' } + 71320075 71320675 59
{'PRDM12' } + 133539481 133540081 53
{'ELAVL2' } - 23826235 23826835 50
{'ZFP37' } - 115818939 115819539 59
{'RP11-35N6.1'} + 103790491 103791091 60
{'DMRT2' } + 1049854 1050454 54

pval_1 Counts_2 pval_2

6.6613e-16 253 5.5511e-16
6.6613e-16 226 5.5511e-16
6.6613e-16 400 5.5511e-16
6.6613e-16 408 5.5511e-16
6.6613e-16 286 5.5511e-16
6.6613e-16 499 5.5511e-16
1.4e-13 165 6.0352e-13

2-69
2 High-Throughput Sequence Analysis

3.5649e-13 151 4.7347e-12

4.3566e-12 163 8.0969e-13
1.2447e-12 133 6.7598e-11
2.8679e-11 148 7.3682e-12
2.3279e-12 127 1.6448e-10
1.3068e-11 135 5.0276e-11
4.1911e-10 144 1.3295e-11
7.897e-10 125 2.2131e-10
5.7523e-10 114 1.1364e-09
9.2538e-10 106 3.7513e-09
2.0467e-09 96 1.6795e-08
3.6266e-08 97 1.4452e-08
1.8171e-07 91 3.5644e-08
1.5457e-07 69 1.0074e-06
4.8093e-07 73 5.4629e-07
1.5082e-06 62 2.9575e-06
3.4322e-06 67 1.3692e-06
2.0943e-06 63 2.5345e-06
5.6364e-06 61 3.4518e-06
9.2778e-06 62 2.9575e-06
2.0943e-06 47 3.0746e-05
1.7771e-06 42 6.8037e-05
4.7762e-06 46 3.6016e-05

Finding Intergenic Regions that are Significantly Methylated

Serre et al. [1] reported that, in these data sets, approximately 90% of the uniquely mapped reads fall
outside the 5' gene promoter regions. Using a similar approach as before, you can find genes that
have intergenic methylated regions. To compensate for the varying lengths of the genes, you can use
the maximum coverage, computed base-by-base, instead of the raw number of mapped short reads.
Another alternative approach to normalize the counts by the gene length is to set the METHOD name-
value pair to rpkm in the getCounts function.

intergenic = dataset({a9.Feature,'Gene'});
intergenic.Strand = char(a9.Strand);
intergenic.Start = a9.Start;
intergenic.Stop = a9.Stop;

intergenic.Counts_1 = getCounts(bm_hct116_1,intergenic.Start,intergenic.Stop,...
'overlap','full','method','max','independent',true);
intergenic.Counts_2 = getCounts(bm_hct116_2,intergenic.Start,intergenic.Stop,...
'overlap','full','method','max','independent',true);
trun = 10; % Set a truncation threshold
pn1 = rtnbinfit(intergenic.Counts_1(intergenic.Counts_1<trun),trun); % Fit to HCT116-1 intergenic
pn2 = rtnbinfit(intergenic.Counts_2(intergenic.Counts_2<trun),trun); % Fit to HCT116-2 intergenic
intergenic.pval_1 = 1 - nbincdf(intergenic.Counts_1,pn1(1),pn1(2)); % p-value for every intergeni
intergenic.pval_2 = 1 - nbincdf(intergenic.Counts_2,pn2(1),pn2(2)); % p-value for every intergeni

Number_of_sig_genes = sum(intergenic.pval_1<.01 & intergenic.pval_2<.01)

Ratio_of_sig_methylated_genes = Number_of_sig_genes./numGenes

[~,order] = sort(intergenic.pval_1.*intergenic.pval_2);

intergenic(order(1:30),[1 2 3 4 5 7 6 8])

2-70
 Exploring Genome-Wide Differences in DNA Methylation Profiles

Number_of_sig_genes =

62

Ratio_of_sig_methylated_genes =

0.0775

ans =

Gene Strand Start Stop Counts_1

{'AL772363.1'} - 140762377 140787022 106
{'CACNA1B' } + 140772241 141019076 106
{'SUSD1' } - 114803065 114937688 88
{'C9orf172' } + 139738867 139741797 99
{'NR5A1' } - 127243516 127269709 86
{'BARX1' } - 96713628 96717654 77
{'KCNT1' } + 138594031 138684992 58
{'GABBR2' } - 101050391 101471479 65
{'FOXB2' } + 79634571 79635869 51
{'NDOR1' } + 140100119 140113813 54
{'KIAA1045' } + 34957484 34984679 50
{'ADAMTSL2' } + 136397286 136440641 55
{'PAX5' } - 36833272 37034476 48
{'OLFM1' } + 137967268 138013025 55
{'PBX3' } + 128508551 128729656 45
{'FOXE1' } + 100615536 100618986 49
{'MPDZ' } - 13105703 13279589 51
{'ASTN2' } - 119187504 120177348 43
{'ARRDC1' } + 140500106 140509812 49
{'IGFBPL1' } - 38408991 38424444 45
{'LHX3' } - 139088096 139096955 44
{'PAPPA' } + 118916083 119164601 44
{'CNTFR' } - 34551430 34590121 41
{'DMRT3' } + 976964 991731 40
{'TUSC1' } - 25676396 25678856 46
{'ELAVL2' } - 23690102 23826335 35
{'SMARCA2' } + 2015342 2193624 36
{'GAS1' } - 89559279 89562104 34
{'GRIN1' } + 140032842 140063207 36
{'TLE4' } + 82186688 82341658 36

pval_1 Counts_2 pval_2

8.6597e-15 98 1.8763e-14
8.6597e-15 98 1.8763e-14
2.2904e-12 112 7.7716e-16
7.4718e-14 96 3.5749e-14
4.268e-12 90 2.5457e-13
7.0112e-11 62 2.569e-09
2.5424e-08 73 6.9019e-11
2.9078e-09 58 9.5469e-09
2.2131e-07 58 9.5469e-09
8.7601e-08 55 2.5525e-08
3.0134e-07 55 2.5525e-08

2-71
2 High-Throughput Sequence Analysis

6.4307e-08 45 6.7163e-07
5.585e-07 49 1.8188e-07
6.4307e-08 42 1.7861e-06
1.4079e-06 51 9.4566e-08
4.1027e-07 46 4.8461e-07
2.2131e-07 42 1.7861e-06
2.6058e-06 43 1.2894e-06
4.1027e-07 36 1.2564e-05
1.4079e-06 39 4.7417e-06
1.9155e-06 36 1.2564e-05
1.9155e-06 35 1.7377e-05
4.8199e-06 37 9.0815e-06
6.5537e-06 37 9.0815e-06
1.0346e-06 31 6.3417e-05
3.0371e-05 41 2.4736e-06
2.2358e-05 40 3.4251e-06
4.1245e-05 41 2.4736e-06
2.2358e-05 38 6.5629e-06
2.2358e-05 37 9.0815e-06

For instance, explore the methylation profile of the BARX1 gene, the sixth significant gene with
intergenic methylation in the previous list. The GTF formatted file ensemblmart_barx1.gtf
contains structural information for this gene obtained from Ensembl using the BioMart service.

Use GTFAnnotation to load the structural information into MATLAB. There are two annotated
transcripts for this gene.
barx1 = GTFAnnotation('ensemblmart_barx1.gtf')
transcripts = getTranscriptNames(barx1)

barx1 =

GTFAnnotation with properties:

FieldNames: {1×11 cell}

NumEntries: 18

transcripts =

2×1 cell array

{'ENST00000253968'}
{'ENST00000401724'}

Plot the DNA methylation profile for both HCT116 sample replicates with base-pair resolution.
Overlay the CpG islands and plot the exons for each of the two transcripts along the bottom of the
plot.
range = barx1.getRange;
r1 = range(1)-1000; % set the region limits
r2 = range(2)+1000;
figure
hold on
% plot high-resolution coverage of bm_hct116_1

2-72
 Exploring Genome-Wide Differences in DNA Methylation Profiles

h1 = plot(r1:r2,getBaseCoverage(bm_hct116_1,r1,r2,'binWidth',1),'b');
% plot high-resolution coverage of bm_hct116_2
h2 = plot(r1:r2,getBaseCoverage(bm_hct116_2,r1,r2,'binWidth',1),'g');

% mark the CpG islands within the [r1 r2] region

for i = 1:numel(cpgi.Starts)
if cpgi.Starts(i)>r1 && cpgi.Stops(i)<r2 % is CpG island inside [r1 r2]?
px = [cpgi.Starts([i i]) cpgi.Stops([i i])]; % x-coordinates for patch
py = [0 max(ylim) max(ylim) 0]; % y-coordinates for patch
hp = patch(px,py,'r','FaceAlpha',.1,'EdgeColor','r','Tag','cpgi');
end
end

% mark the exons at the bottom of the axes

for i = 1:numel(transcripts)
exons = getSubset(barx1,'Transcript',transcripts{i},'Feature','exon');
for j = 1:exons.NumEntries
px = [exons.Start([j j]);exons.Stop([j j])]'; % x-coordinates for patch
py = [0 1 1 0]-i*2-1; % y-coordinates for patch
hq = patch(px,py,'b','FaceAlpha',.1,'EdgeColor','b','Tag','exon');
end
end

axis([r1 r2 -numel(transcripts)*2-2 80]) % zooms-in the y-axis

fixGenomicPositionLabels(gca) % formats tick labels and adds datacursors
ylabel('Coverage')
xlabel('Chromosome 9 position')
title('High resolution coverage in the BARX1 gene')
legend([h1 h2 hp hq],'HCT116-1','HCT116-2','CpG Islands','Exons','Location','NorthWest')

Observe the highly methylated region in the 5' promoter region (right-most CpG island). Recall that
for this gene transcription occurs in the reverse strand. More interesting, observe the highly
methylated regions that overlap the initiation of each of the two annotated transcripts (two middle
CpG islands).

2-73
2 High-Throughput Sequence Analysis

Differential Analysis of Methylation Patterns

In the study by Serre et al. another cell line is also analyzed. New cells (DICERex5) are derived from
the same HCT116 colon cancer cells after truncating the DICER1 alleles. It has been reported in
literature [5] that there is a localized change of DNA methylation at small number of gene promoters.
In this example, you will find significant 100-bp windows in two sample replicates of the DICERex5
cells following the same approach as the parental HCT116 cells, and then you will search statistically
significant differences between the two cell lines.

The helper function getWindowCounts captures the similar steps to find windows with significant
coverage as before. getWindowCounts returns vectors with counts, p-values, and false discovery
rates for each new replicate.

bm_dicer_1 = BioMap('SRR030222.bam','SelectRef','gi|224589821|ref|NC_000009.11|');
bm_dicer_2 = BioMap('SRR030223.bam','SelectRef','gi|224589821|ref|NC_000009.11|');
[counts_3,pval3,fdr3] = getWindowCounts(bm_dicer_1,4,w,100);
[counts_4,pval4,fdr4] = getWindowCounts(bm_dicer_2,4,w,100);
w3 = fdr3<.01; % logical vector indicating significant windows in DICERex5_1
w4 = fdr4<.01; % logical vector indicating significant windows in DICERex5-2
w34 = w3 & w4; % logical vector indicating significant windows in both replicates
Number_of_sig_windows_DICERex5_1 = sum(w3)
Number_of_sig_windows_DICERex5_2 = sum(w4)
Number_of_sig_windows_DICERex5 = sum(w34)

Number_of_sig_windows_DICERex5_1 =

908

Number_of_sig_windows_DICERex5_2 =

1041

Number_of_sig_windows_DICERex5 =

759

To perform a differential analysis you use the 100-bp windows that are significant in at least one of
the samples (either HCT116 or DICERex5).

wd = w34 | w12; % logical vector indicating windows included in the diff. analysis

counts = [counts_1(wd) counts_2(wd) counts_3(wd) counts_4(wd)];

ws = w(wd); % window start for each row in counts

Use the function manorm to normalize the data. The PERCENTILE name-value pair lets you filter out
windows with very large number of counts while normalizing, since these windows are mainly due to
artifacts, such as repetitive regions in the reference chromosome.

counts_norm = round(manorm(counts,'percentile',90).*100);

Use the function mattest to perform a two-sample t-test to identify differentially covered windows
from the two different cell lines.

2-74
 Exploring Genome-Wide Differences in DNA Methylation Profiles

pval = mattest(counts_norm(:,[1 2]),counts_norm(:,[3 4]),'bootstrap',true,...

'showhist',true,'showplot',true);

2-75
2 High-Throughput Sequence Analysis

Create a report with the 25 most significant differentially covered windows. While creating the report
use the helper function findClosestGene to determine if the window is intergenic, intragenic, or if
it is in a proximal promoter region.
[~,ord] = sort(pval);
fprintf('Window Pos Type p-value HCT116 DICERex5\n\n');
for i = 1:25
j = ord(i);
[~,msg] = findClosestGene(a9,[ws(j) ws(j)+99]);
fprintf('%10d %-25s %7.6f%5d%5d %5d%5d\n', ...
ws(j),msg,pval(j),counts_norm(j,:));
end

Window Pos Type p-value HCT116 DICERex5

140311701 Intergenic (EXD3) 0.000026 13 13 104 105

139546501 Intragenic 0.001827 21 21 91 93
10901 Intragenic 0.002671 258 257 434 428
120176801 Intergenic (ASTN2) 0.002733 266 270 155 155
139914801 Intergenic (ABCA2) 0.002980 64 63 26 25
126128501 Intergenic (CRB2) 0.003193 94 93 129 130
71939501 Prox. Promoter (FAM189A2) 0.005549 107 101 0 0
124461001 Intergenic (DAB2IP) 0.005618 77 76 39 37
140086501 Intergenic (TPRN) 0.006520 47 42 123 124
79637201 Intragenic 0.007512 52 51 32 31
136470801 Intragenic 0.007512 52 51 32 31
140918001 Intergenic (CACNA1B) 0.008115 176 169 71 68
100615901 Intergenic (FOXE1) 0.008346 262 253 123 118

2-76
 Exploring Genome-Wide Differences in DNA Methylation Profiles

98221901 Intergenic (PTCH1) 0.009934 26 30 104 99

138730601 Intergenic (CAMSAP1) 0.010273 26 21 97 93
89561701 Intergenic (GAS1) 0.010336 77 76 6 12
977401 Intergenic (DMRT3) 0.010369 236 245 129 124
37002601 Intergenic (PAX5) 0.010559 133 127 207 211
139744401 Intergenic (PHPT1) 0.010869 47 46 32 31
126771301 Intragenic 0.011458 43 46 97 93
93922501 Intragenic 0.011486 34 34 149 161
94187101 Intragenic 0.011507 73 80 6 6
136044401 Intragenic 0.011567 39 34 110 105
139611201 Intergenic (FAM69B) 0.011567 39 34 110 105
139716201 Intergenic (C9orf86) 0.011832 73 72 136 130

Plot the DNA methylation profile for the promoter region of gene FAM189A2, the most significant
differentially covered promoter region from the previous list. Overlay the CpG islands and the
FAM189A2 gene.

range = getRange(getSubset(a9,'Feature','FAM189A2'));
r1 = range(1)-1000;
r2 = range(2)+1000;
figure
hold on

% plot high-resolution coverage of all replicates

h1 = plot(r1:r2,getBaseCoverage(bm_hct116_1,r1,r2,'binWidth',1),'b');
h2 = plot(r1:r2,getBaseCoverage(bm_hct116_2,r1,r2,'binWidth',1),'g');
h3 = plot(r1:r2,getBaseCoverage(bm_dicer_1,r1,r2,'binWidth',1),'r');
h4 = plot(r1:r2,getBaseCoverage(bm_dicer_2,r1,r2,'binWidth',1),'m');

% mark the CpG islands within the [r1 r2] region

for i = 1:numel(cpgi.Starts)
if cpgi.Starts(i)>r1 && cpgi.Stops(i)<r2 % is CpG island inside [r1 r2]?
px = [cpgi.Starts([i i]) cpgi.Stops([i i])]; % x-coordinates for patch
py = [0 max(ylim) max(ylim) 0]; % y-coordinates for patch
hp = patch(px,py,'r','FaceAlpha',.1,'EdgeColor','r','Tag','cpgi');
end
end

% mark the gene at the bottom of the axes

px = range([1 1 2 2]);
py = [0 1 1 0]-2;
hq = patch(px,py,'b','FaceAlpha',.1,'EdgeColor','b','Tag','gene');

axis([r1 r1+4000 -4 30]) % zooms-in

fixGenomicPositionLabels(gca) % formats tick labels and adds datacursors
ylabel('Coverage')
xlabel('Chromosome 9 position')
title('DNA Methylation profiles along the promoter region of the FAM189A2 gene')
legend([h1 h2 h3 h4 hp hq],...
'HCT116-1','HCT116-2','DICERex5-1','DICERex5-2','CpG Islands','FAM189A2 Gene',...
'Location','NorthEast')

2-77
2 High-Throughput Sequence Analysis

Observe that the CpG islands are clearly unmethylated for both of the DICERex5 replicates.

References

[1] Serre, D., Lee, B.H., and Ting A.H., "MBD-isolated Genome Sequencing provides a high-
throughput and comprehensive survey of DNA methylation in the human genome", Nucleic Acids
Research, 38(2):391-9, 2010.

[2] Langmead, B., Trapnell, C., Pop, M., and Salzberg, S.L., "Ultrafast and Memory-efficient
Alignment of Short DNA Sequences to the Human Genome", Genome Biology, 10(3):R25, 2009.

[3] Li, H., et al., "The Sequence Alignment/map (SAM) Format and SAMtools", Bioinformatics,
25(16):2078-9, 2009.

[4] Gardiner-Garden, M. and Frommer, M., "CpG islands in vertebrate genomes", Journal of Molecular
Biology, 196(2):261-82, 1987.

[5] Ting, A.H., et al., "A Requirement for DICER to Maintain Full Promoter CpG Island
Hypermethylation in Human Cancer Cells", Cancer Research, 68(8):2570-5, 2008.

[6] Benjamini, Y. and Hochberg, Y., "Controlling the false discovery rate: a practical and powerful
approach to multiple testing", Journal of the Royal Statistical Society, 57(1):289-300, 1995.

2-78
 Exploring Protein-DNA Binding Sites from Paired-End ChIP-Seq Data

Exploring Protein-DNA Binding Sites from Paired-End ChIP-Seq

Data

This example shows how to perform a genome-wide analysis of a transcription factor in the
Arabidopsis Thaliana (Thale Cress) model organism.

For enhanced performance, it is recommended that you run this example on a 64-bit platform,
because the memory footprint is close to 2 Gb. On a 32-bit platform, if you receive "Out of
memory" errors when running this example, try increasing the virtual memory (or swap space) of
your operating system or try setting the 3GB switch (32-bit Windows® XP only). For details, see
“Resolve “Out of Memory” Errors”.

Introduction

ChIP-Seq is a technology that is used to identify transcription factors that interact with specific DNA
sites. First chromatin immunoprecipitation enriches DNA-protein complexes using an antibody that
binds to a particular protein of interest. Then, all the resulting fragments are processed using high-
throughput sequencing. Sequencing fragments are mapped back to the reference genome. By
inspecting over-represented regions it is possible to mark the genomic location of DNA-protein
interactions.

In this example, short reads are produced by the paired-end Illumina® platform. Each fragment is
reconstructed from two short reads successfully mapped, with this the exact length of the fragment
can be computed. Using paired-end information from sequence reads maximizes the accuracy of
predicting DNA-protein binding sites.

Data Set

This example explores the paired-end ChIP-Seq data generated by Wang et.al. [1] using the Illumina®
platform. The data set has been courteously submitted to the Gene Expression Omnibus repository
with accession number GSM424618. The unmapped paired-end reads can be obtained from the NCBI
FTP site.

This example assumes that you:

(1) downloaded the data containing the unmapped short read and converted it to FASTQ formatted
files using the NCBI SRA Toolkit.

(2) produced a SAM formatted file by mapping the short reads to the Thale Cress reference genome,
using a mapper such as BWA [2], Bowtie, or SSAHA2 (which is the mapper used by authors of [1]),
and,

(3) ordered the SAM formatted file by reference name first, then by genomic position.

For the published version of this example, 8,655,859 paired-end short reads are mapped using the
BWA mapper [2]. BWA produced a SAM formatted file (aratha.sam) with 17,311,718 records
(8,655,859 x 2). Repetitive hits were randomly chosen, and only one hit is reported, but with lower
mapping quality. The SAM file was ordered and converted to a BAM formatted file using SAMtools [3]
before being loaded into MATLAB.

The last part of the example also assumes that you downloaded the reference genome for the Thale
Cress model organism (which includes five chromosomes). Uncomment the following lines of code to
download the reference from the NCBI repository:

2-79
2 High-Throughput Sequence Analysis

% getgenbank('NC_003070','FileFormat','fasta','tofile','ach1.fasta');
% getgenbank('NC_003071','FileFormat','fasta','tofile','ach2.fasta');
% getgenbank('NC_003074','FileFormat','fasta','tofile','ach3.fasta');
% getgenbank('NC_003075','FileFormat','fasta','tofile','ach4.fasta');
% getgenbank('NC_003076','FileFormat','fasta','tofile','ach5.fasta');

Creating a MATLAB® Interface to a BAM Formatted File

To create local alignments and look at the coverage we need to construct a BioMap. BioMap has an
interface that provides direct access to the mapped short reads stored in the BAM formatted file, thus
minimizing the amount of data that is actually loaded to the workspace. Create a BioMap to access all
the short reads mapped in the BAM formatted file.

bm = BioMap('aratha.bam')

bm =

BioMap with properties:

SequenceDictionary: {5x1 cell}

Reference: [14637324x1 File indexed property]
Signature: [14637324x1 File indexed property]
Start: [14637324x1 File indexed property]
MappingQuality: [14637324x1 File indexed property]
Flag: [14637324x1 File indexed property]
MatePosition: [14637324x1 File indexed property]
Quality: [14637324x1 File indexed property]
Sequence: [14637324x1 File indexed property]
Header: [14637324x1 File indexed property]
NSeqs: 14637324
Name: ''

Use the getSummary method to obtain a list of the existing references and the actual number of
short read mapped to each one.

getSummary(bm)

BioMap summary:
Name: ''
Container_Type: 'Data is file indexed.'
Total_Number_of_Sequences: 14637324
Number_of_References_in_Dictionary: 5

Number_of_Sequences Genomic_Range
Chr1 3151847 1 30427671
Chr2 3080417 1000 19698292
Chr3 3062917 94 23459782
Chr4 2218868 1029 18585050
Chr5 3123275 11 26975502

The remainder of this example focuses on the analysis of one of the five chromosomes, Chr1. Create a
new BioMap to access the short reads mapped to the first chromosome by subsetting the first one.

bm1 = getSubset(bm,'SelectReference','Chr1')

2-80
 Exploring Protein-DNA Binding Sites from Paired-End ChIP-Seq Data

bm1 =

BioMap with properties:

SequenceDictionary: 'Chr1'
Reference: [3151847x1 File indexed property]
Signature: [3151847x1 File indexed property]
Start: [3151847x1 File indexed property]
MappingQuality: [3151847x1 File indexed property]
Flag: [3151847x1 File indexed property]
MatePosition: [3151847x1 File indexed property]
Quality: [3151847x1 File indexed property]
Sequence: [3151847x1 File indexed property]
Header: [3151847x1 File indexed property]
NSeqs: 3151847
Name: ''

By accessing the Start and Stop positions of the mapped short read you can obtain the genomic
range.

x1 = min(getStart(bm1))
x2 = max(getStop(bm1))

x1 =

uint32

x2 =

uint32

30427671

Exploring the Coverage at Different Resolutions

To explore the coverage for the whole range of the chromosome, a binning algorithm is required. The
getBaseCoverage method produces a coverage signal based on effective alignments. It also allows
you to specify a bin width to control the size (or resolution) of the output signal. However internal
computations are still performed at the base pair (bp) resolution. This means that despite setting a
large bin size, narrow peaks in the coverage signal can still be observed. Once the coverage signal is
plotted you can program the figure's data cursor to display the genomic position when using the
tooltip. You can zoom and pan the figure to determine the position and height of the ChIP-Seq peaks.

[cov,bin] = getBaseCoverage(bm1,x1,x2,'binWidth',1000,'binType','max');
figure
plot(bin,cov)
axis([x1,x2,0,100]) % sets the axis limits
fixGenomicPositionLabels % formats tick labels and adds datacursors
xlabel('Base position')

2-81
2 High-Throughput Sequence Analysis

ylabel('Depth')
title('Coverage in Chromosome 1')

It is also possible to explore the coverage signal at the bp resolution (also referred to as the pile-up
profile). Explore one of the large peaks observed in the data at position 4598837.

p1 = 4598837-1000;
p2 = 4598837+1000;

figure
plot(p1:p2,getBaseCoverage(bm1,p1,p2))
xlim([p1,p2]) % sets the x-axis limits
fixGenomicPositionLabels % formats tick labels and adds datacursors
xlabel('Base position')
ylabel('Depth')
title('Coverage in Chromosome 1')

2-82
 Exploring Protein-DNA Binding Sites from Paired-End ChIP-Seq Data

Identifying and Filtering Regions with Artifacts

Observe the large peak with coverage depth of 800+ between positions 4599029 and 4599145.
Investigate how these reads are aligning to the reference chromosome. You can retrieve a subset of
these reads enough to satisfy a coverage depth of 25, since this is sufficient to understand what is
happening in this region. Use getIndex to obtain indices to this subset. Then use
getCompactAlignment to display the corresponding multiple alignment of the short-reads.

i = getIndex(bm1,4599029,4599145,'depth',25);
bmx = getSubset(bm1,i,'inmemory',false)
getCompactAlignment(bmx,4599029,4599145)

bmx =

BioMap with properties:

SequenceDictionary: 'Chr1'
Reference: [62x1 File indexed property]
Signature: [62x1 File indexed property]
Start: [62x1 File indexed property]
MappingQuality: [62x1 File indexed property]
Flag: [62x1 File indexed property]
MatePosition: [62x1 File indexed property]
Quality: [62x1 File indexed property]
Sequence: [62x1 File indexed property]
Header: [62x1 File indexed property]
NSeqs: 62
Name: ''

ans =

2-83
2 High-Throughput Sequence Analysis

35x117 char array

'AGTT AATCAAATAGAAAGCCCCGAGGGCGCCATATCCTAGGCGC AAACTATGTGATTGAATAAATCCTCCTCTATCTGTTGCGG GA

'AGTGC TCAAATAGAAAGCCCCGAGGGCGCCATATTCTAGGAGCCC GAATAAATCCTCCTCTATCTGTTGCGGGTCGA
'AGTTCAA CCCGAGGGCGCCATATTCTAGGAGCCCAAACTATGTGATT TATCTGTTGCGGGTCGA
'AGTTCAATCAAATAGAAAGC TTCTAGGAGCCCAAACTATGTGATTGAATAAATCCTCCTC A
'AGTT AAGGAGCCCAAAATATGTGATTGAATAAATCCACCTCTAT
'AGTACAATCAAATAGAAAGCCCCGAGGGCGCCATA TAGGAGCCCAAACTATGTGATTGAATAAATCCTCCTCTAT
'CGTACAATCAAATAGAAAGCCCCGAGGGCGCCATATTC GGAGCCCAAACTATGTGATTGAATAAATCCTCCTCTATCT
'CGTACAATCAAATAGAAAGCCCCGAGGGCGCCATATTC GGAGCCCAAACTATGTGATTGAATAAATCCTCCTCTATCT
'CGTACAATCAAATAGAAAGCCCCGAGGGCGCCATATTC GGAGCCCAAGCTATGTGATTGAATAAATCCTCCTCTATCT
'CGTACAATCAAATAGAAAGCCCCGAGGGCGCCATATTC GGAGCCCAAACTATGTGATTGAATAAATCCTCCTCTATCT
'AGTTCAATCAAATAGAAAGCCCCGAGGGCGCCATATTCTA GAGCCCAAACTATGTGATTGAATAAATCCTCCTCTATCTG
'GATACAATCAAATAGAAAGCCCCGAGGGCGCCATATTCTA GAGCCCAAACTATGTGATTGAATAAATCTTCCTCTATCTG
'GATACAATCAAATAGAAAGCCCCGAGGGCGCCATATTCTA GAGCCCAAACTATGTGATTGAATAAATCCTCCTCTATCTG
'GATACAATCAAATAGAAAGCCCCGAGGGCGCCATATTCTA GAGCCCAAACTATGTGATTGAATAAATCCTCCTCTATCTG
'GATACAATCAAATAGAAAGCCCCGAGGGCGCCATATTCTA GAGCCCAAATTATGTGATTGAATAAATCCTCCTCTATCTG
' ATACAATCAAATAGAAAGCCCCGAGGGCGCCATATTCTAG CCCAAACTATGTGATTGAATAAATCCTCCTCTATCTGTTG
' ATACAATCAAATAGAAAGCCCCGAGGGCGCCATATTCTAG CACAAACTATGTGATTGAATAAATCCTCCTCTATCTGTTG
' ATACAATCAAATAGAAAGCCCCGAGGGCGCCATATTCTAG CCAAACTATGTGATTGAATAAATCCTCCTCTATCTGTTGC
' ATACAATCAAATAGAAAGCCCCGAGGGCGCCATATTCTAG
' ATACAATCAAATAGAAAGCCCCGAGGGCGCCATATTCTCG
' ATACAATCAAATAGAAAGCCCCGGGGGCGCCATATTCTAG
' ATTGAGTCAAATAGAAAGCCCCGAGGGCGCCATATTCTAG
' ATACAATCAAATAGAAAGCCCCGAGGGCGCCATATTCTAG
' ATACAATCAAATAGAAAGCCCCGAGGGCGCCATATTCTAG
' ATACAATCAAATAGAAAGCCCCGAGGGCGCCATATTCTAG
' CAATCAAATAGAAAGCCCCGAGGGCGCCATATTCTAGGAG
' CAATCAAATAGAAAGCCCCGAGGGCGCCATATTCTAGGAG
' TAGGAGCCCAAACTATGTGATTGAATAAATCCTCCTCTAT
' TAGGAGCCCAAACTATGCCATTGAATAAATCCTCCGCTAT
' GGAGCCCAAGCTATGTGATTGAATAAATCCTCCTCTATCT
' GAGCCCAAACTATGTGATTGAATAAATCCTCCTCTATCTG
' GAGCCCAAACTATGTGATTGAATAAATCCTCCTCTATCTG
' GAGCCCAAACTATGTGATTGAATAAATCCTCCTCTATCTG
' GAGCCCAAACTATGTGATTGAATAAATCCTCCTCTATCTG
' GAGCCCAAACTATGTGATTGAATAAATCCTCCTCTATCTG

In addition to visually confirming the alignment, you can also explore the mapping quality for all the
short reads in this region, as this may hint to a potential problem. In this case, less than one percent
of the short reads have a Phred quality of 60, indicating that the mapper most likely found multiple
hits within the reference genome, hence assigning a lower mapping quality.

figure
i = getIndex(bm1,4599029,4599145);
hist(double(getMappingQuality(bm1,i)))
title('Mapping Quality of the reads between 4599029 and 4599145')
xlabel('Phred Quality Score')
ylabel('Number of Reads')

2-84
 Exploring Protein-DNA Binding Sites from Paired-End ChIP-Seq Data

Most of the large peaks in this data set occur due to satellite repeat regions or due to its closeness to
the centromere [4], and show characteristics similar to the example just explored. You may explore
other regions with large peaks using the same procedure.

To prevent these problematic regions, two techniques are used. First, given that the provided data set
uses paired-end sequencing, by removing the reads that are not aligned in a proper pair reduces the
number of potential aligner errors or ambiguities. You can achieve this by exploring the flag field of
the SAM formatted file, in which the second less significant bit is used to indicate if the short read is
mapped in a proper pair.

i = find(bitget(getFlag(bm1),2));
bm1_filtered = getSubset(bm1,i)

bm1_filtered =

BioMap with properties:

SequenceDictionary: 'Chr1'
Reference: [3040724x1 File indexed property]
Signature: [3040724x1 File indexed property]
Start: [3040724x1 File indexed property]
MappingQuality: [3040724x1 File indexed property]
Flag: [3040724x1 File indexed property]
MatePosition: [3040724x1 File indexed property]
Quality: [3040724x1 File indexed property]
Sequence: [3040724x1 File indexed property]

2-85
2 High-Throughput Sequence Analysis

Header: [3040724x1 File indexed property]

NSeqs: 3040724
Name: ''

Second, consider only uniquely mapped reads. You can detect reads that are equally mapped to
different regions of the reference sequence by looking at the mapping quality, because BWA assigns a
lower mapping quality (less than 60) to this type of short read.

i = find(getMappingQuality(bm1_filtered)==60);
bm1_filtered = getSubset(bm1_filtered,i)

bm1_filtered =

BioMap with properties:

SequenceDictionary: 'Chr1'
Reference: [2313252x1 File indexed property]
Signature: [2313252x1 File indexed property]
Start: [2313252x1 File indexed property]
MappingQuality: [2313252x1 File indexed property]
Flag: [2313252x1 File indexed property]
MatePosition: [2313252x1 File indexed property]
Quality: [2313252x1 File indexed property]
Sequence: [2313252x1 File indexed property]
Header: [2313252x1 File indexed property]
NSeqs: 2313252
Name: ''

Visualize again the filtered data set using both, a coarse resolution with 1000 bp bins for the whole
chromosome, and a fine resolution for a small region of 20,000 bp. Most of the large peaks due to
artifacts have been removed.

[cov,bin] = getBaseCoverage(bm1_filtered,x1,x2,'binWidth',1000,'binType','max');
figure
plot(bin,cov)
axis([x1,x2,0,100]) % sets the axis limits
fixGenomicPositionLabels % formats tick labels and adds datacursors
xlabel('Base Position')
ylabel('Depth')
title('Coverage in Chromosome 1 after Filtering')

p1 = 24275801-10000;
p2 = 24275801+10000;

figure
plot(p1:p2,getBaseCoverage(bm1_filtered,p1,p2))
xlim([p1,p2]) % sets the x-axis limits
fixGenomicPositionLabels % formats tick labels and adds datacursors
xlabel('Base Position')
ylabel('Depth')
title('Coverage in Chromosome 1 after Filtering')

2-86
 Exploring Protein-DNA Binding Sites from Paired-End ChIP-Seq Data

Recovering Sequencing Fragments from the Paired-End Reads

In Wang's paper [1] it is hypothesized that paired-end sequencing data has the potential to increase
the accuracy of the identification of chromosome binding sites of DNA associated proteins because
the fragment length can be derived accurately, while when using single-end sequencing it is
necessary to resort to a statistical approximation of the fragment length, and use it indistinctly for all
putative binding sites.

2-87
2 High-Throughput Sequence Analysis

Use the paired-end reads to reconstruct the sequencing fragments. First, get the indices for the
forward and the reverse reads in each pair. This information is captured in the fifth bit of the flag
field, according to the SAM file format.

fow_idx = find(~bitget(getFlag(bm1_filtered),5));
rev_idx = find(bitget(getFlag(bm1_filtered),5));

SAM-formatted files use the same header strings to identify pair mates. By pairing the header strings
you can determine how the short reads in BioMap are paired. To pair the header strings, simply order
them in ascending order and use the sorting indices (hf and hr) to link the unsorted header strings.

[~,hf] = sort(getHeader(bm1_filtered,fow_idx));
[~,hr] = sort(getHeader(bm1_filtered,rev_idx));
mate_idx = zeros(numel(fow_idx),1);
mate_idx(hf) = rev_idx(hr);

Use the resulting fow_idx and mate_idx variables to retrieve pair mates. For example, retrieve the
paired-end reads for the first 10 fragments.

for j = 1:10
disp(getInfo(bm1_filtered, fow_idx(j)))
disp(getInfo(bm1_filtered, mate_idx(j)))
end

SRR054715.sra.6849385 163 20 60 40M AACCCTAAACCTCTGAATCCTTAATCCCTAAATCCCTAAA BB

SRR054715.sra.6849385 83 229 60 40M CCTATTTCTTGTGGTTTTCTTTCCTTCACTTAGCTATGGA 06
SRR054715.sra.6992346 99 20 60 40M AACCCTAAACCTCTGAATCCTTAATCCCTAAATCCCTAAA =B?
SRR054715.sra.6992346 147 239 60 40M GTGGTTTTCTTTCCTTCACTTAGCTATGGATGGTTTATCT B
SRR054715.sra.8438570 163 47 60 40M CTAAATCCCTAAATCTTTAAATCCTACATCCATGAATCCC BC
SRR054715.sra.8438570 83 274 60 40M TATCTTCATTTGTTATATTGGATACAAGCTTTGCTACGAT BB
SRR054715.sra.1676744 163 67 60 40M ATCCTACATCCATGAATCCCTAAATACCTAATCCCCTAAA BB
SRR054715.sra.1676744 83 283 60 40M TTGTTATATTGGATACAAGCTTTGCTACGATCTACATTTG CC
SRR054715.sra.6820328 163 73 60 40M CATCCATGAATCCCTAAATACCTAATTCCCTAAACCCGAA BB
SRR054715.sra.6820328 83 267 60 40M GTTGGTGTATCTTCATTTGTTATATTGGATACGAGCTTTG BB
SRR054715.sra.1559757 163 103 60 40M TAAACCCGAAACCGGTTTCTCTGGTTGAAACTCATTGTGT B
SRR054715.sra.1559757 83 311 60 40M GATCTACATTTGGGAATGTGAGTCTCTTATTGTAACCTTA <?
SRR054715.sra.5658991 163 103 60 40M CAAACCCGAAACCGGTTTCTCTGGTTGAAACTCATTGTGT 7
SRR054715.sra.5658991 83 311 60 40M GATCTACATTTGGGAATGTGAGTCTCTTATTGTAACCTTA 3,
SRR054715.sra.4625439 163 143 60 40M ATATAATGATAATTTTAGCGTTTTTATGCAATTGCTTATT B
SRR054715.sra.4625439 83 347 60 40M CTTAGTGTTGGTTTATCTCAAGAATCTTATTAATTGTTTG +B
SRR054715.sra.1007474 163 210 60 40M ATTTGAGGTCAATACAAATCCTATTTCTTGTGGTTTGCTT B
SRR054715.sra.1007474 83 408 60 40M TATTGTCATTCTTACTCCTTTGTGGAAATGTTTGTTCTAT BB
SRR054715.sra.7345693 99 213 60 40M TGAGGTCAATACAAATCCTATTTCTTGTGGTTTTCTTTCT B>
SRR054715.sra.7345693 147 393 60 40M TTATTTTTGGACATTTATTGTCATTCTTACTCCTTTGGGG B

Use the paired-end indices to construct a new BioMap with the minimal information needed to
represent the sequencing fragments. First, calculate the insert sizes.

J = getStop(bm1_filtered, fow_idx);
K = getStart(bm1_filtered, mate_idx);
L = K - J - 1;

Obtain the new signature (or CIGAR string) for each fragment by using the short read original
signatures separated by the appropriate number of skip CIGAR symbols (N).

n = numel(L);
cigars = cell(n,1);
for i = 1:n

2-88
 Exploring Protein-DNA Binding Sites from Paired-End ChIP-Seq Data

cigars{i} = sprintf('%dN' ,L(i));

end
cigars = strcat( getSignature(bm1_filtered, fow_idx),...
cigars,...
getSignature(bm1_filtered, mate_idx));

Reconstruct the sequences for the fragments by concatenating the respective sequences of the
paired-end short reads.

seqs = strcat( getSequence(bm1_filtered, fow_idx),...

getSequence(bm1_filtered, mate_idx));

Calculate and plot the fragment size distribution.

J = getStart(bm1_filtered,fow_idx);
K = getStop(bm1_filtered,mate_idx);
L = K - J + 1;
figure
hist(double(L),100)
title(sprintf('Fragment Size Distribution\n %d Paired-end Fragments Mapped to Chromosome 1',n))
xlabel('Fragment Size')
ylabel('Count')

Construct a new BioMap to represent the sequencing fragments. With this, you will be able explore
the coverage signals as well as local alignments of the fragments.

bm1_fragments = BioMap('Sequence',seqs,'Signature',cigars,'Start',J)

2-89
2 High-Throughput Sequence Analysis

bm1_fragments =

BioMap with properties:

SequenceDictionary: {0x1 cell}

Reference: {0x1 cell}
Signature: {1156626x1 cell}
Start: [1156626x1 uint32]
MappingQuality: [0x1 uint8]
Flag: [0x1 uint16]
MatePosition: [0x1 uint32]
Quality: {0x1 cell}
Sequence: {1156626x1 cell}
Header: {0x1 cell}
NSeqs: 1156626
Name: ''

Exploring the Coverage Using Fragment Alignments

Compare the coverage signal obtained by using the reconstructed fragments with the coverage signal
obtained by using individual paired-end reads. Notice that enriched binding sites, represented by
peaks, can be better discriminated from the background signal.

cov_reads = getBaseCoverage(bm1_filtered,x1,x2,'binWidth',1000,'binType','max');
[cov_fragments,bin] = getBaseCoverage(bm1_fragments,x1,x2,'binWidth',1000,'binType','max');

figure
plot(bin,cov_reads,bin,cov_fragments)
xlim([x1,x2]) % sets the x-axis limits
fixGenomicPositionLabels % formats tick labels and adds datacursors
xlabel('Base position')
ylabel('Depth')
title('Coverage Comparison')
legend('Short Reads','Fragments')

2-90
 Exploring Protein-DNA Binding Sites from Paired-End ChIP-Seq Data

Perform the same comparison at the bp resolution. In this dataset, Wang et.al. [1] investigated a basic
helix-loop-helix (bHLH) transcription factor. bHLH proteins typically bind to a consensus sequence
called an E-box (with a CANNTG motif). Use fastaread to load the reference chromosome, search for
the E-box motif in the 3' and 5' directions, and then overlay the motif positions on the coverage
signals. This example works over the region 1-200,000, however the figure limits are narrowed to a
3000 bp region in order to better depict the details.

p1 = 1;
p2 = 200000;

cov_reads = getBaseCoverage(bm1_filtered,p1,p2);
[cov_fragments,bin] = getBaseCoverage(bm1_fragments,p1,p2);

chr1 = fastaread('ach1.fasta');
mp1 = regexp(chr1.Sequence(p1:p2),'CA..TG')+3+p1;
mp2 = regexp(chr1.Sequence(p1:p2),'GT..AC')+3+p1;
motifs = [mp1 mp2];

figure
plot(bin,cov_reads,bin,cov_fragments)
hold on
plot([1;1;1]*motifs,[0;max(ylim);NaN],'r')
xlim([111000 114000]) % sets the x-axis limits
fixGenomicPositionLabels % formats tick labels and adds datacursors
xlabel('Base position')
ylabel('Depth')
title('Coverage Comparison')
legend('Short Reads','Fragments','E-box motif')

2-91
2 High-Throughput Sequence Analysis

Observe that it is not possible to associate each peak in the coverage signals with an E-box motif. This
is because the length of the sequencing fragments is comparable to the average motif distance,
blurring peaks that are close. Plot the distribution of the distances between the E-box motif sites.

motif_sep = diff(sort(motifs));
figure
hist(motif_sep(motif_sep<500),50)
title('Distance (bp) between adjacent E-box motifs')
xlabel('Distance (bp)')
ylabel('Counts')

2-92
 Exploring Protein-DNA Binding Sites from Paired-End ChIP-Seq Data

Finding Significant Peaks in the Coverage Signal

Use the function mspeaks to perform peak detection with Wavelets denoising on the coverage signal
of the fragment alignments. Filter putative ChIP peaks using a height filter to remove peaks that are
not enriched by the binding process under consideration.

putative_peaks = mspeaks(bin,cov_fragments,'noiseestimator',20,...
'heightfilter',10,'showplot',true);
hold on
legend('off')
plot([1;1;1]*motifs(motifs>p1 & motifs<p2),[0;max(ylim);NaN],'r')
xlim([111000 114000]) % sets the x-axis limits
fixGenomicPositionLabels % formats tick labels and adds datacursors
legend('Coverage from Fragments','Wavelet Denoised Coverage','Putative ChIP peaks','E-box Motifs'
xlabel('Base position')
ylabel('Depth')
title('ChIP-Seq Peak Detection')

2-93
2 High-Throughput Sequence Analysis

Use the knnsearch function to find the closest motif to each one of the putative peaks. As expected,
most of the enriched ChIP peaks are close to an E-box motif [1]. This reinforces the importance of
performing peak detection at the finest resolution possible (bp resolution) when the expected density
of binding sites is high, as it is in the case of the E-box motif. This example also illustrates that for
this type of analysis, paired-end sequencing should be considered over single-end sequencing [1].

h = knnsearch(motifs',putative_peaks(:,1));
distance = putative_peaks(:,1)-motifs(h(:))';
figure
hist(distance(abs(distance)<200),50)
title('Distance to Closest E-box Motif for Each Detected Peak')
xlabel('Distance (bp)')
ylabel('Counts')

2-94
 Exploring Protein-DNA Binding Sites from Paired-End ChIP-Seq Data

References
[1] Wang, Congmao, Jie Xu, Dasheng Zhang, Zoe A Wilson, and Dabing Zhang. “An Effective Approach
for Identification of in Vivo Protein-DNA Binding Sites from Paired-End ChIP-Seq Data.” BMC
Bioinformatics 11, no. 1 (2010): 81.

[2] Li, H., and R. Durbin. “Fast and Accurate Short Read Alignment with Burrows-Wheeler
Transform.” Bioinformatics 25, no. 14 (July 15, 2009): 1754–60.

[3] Li, H., B. Handsaker, A. Wysoker, T. Fennell, J. Ruan, N. Homer, G. Marth, G. Abecasis, R. Durbin,
and 1000 Genome Project Data Processing Subgroup. “The Sequence Alignment/Map Format
and SAMtools.” Bioinformatics 25, no. 16 (August 15, 2009): 2078–79.

[4] Jothi, R., S. Cuddapah, A. Barski, K. Cui, and K. Zhao. “Genome-Wide Identification of in Vivo
Protein-DNA Binding Sites from ChIP-Seq Data.” Nucleic Acids Research 36, no. 16 (August 1,
2008): 5221–31.

[5] Hoffman, Brad G, and Steven J M Jones. “Genome-Wide Identification of DNA–Protein Interactions
Using Chromatin Immunoprecipitation Coupled with Flow Cell Sequencing.” Journal of
Endocrinology 201, no. 1 (April 2009): 1–13.

[6] Ramsey, Stephen A., Theo A. Knijnenburg, Kathleen A. Kennedy, Daniel E. Zak, Mark Gilchrist,
Elizabeth S. Gold, Carrie D. Johnson, et al. “Genome-Wide Histone Acetylation Data Improve
Prediction of Mammalian Transcription Factor Binding Sites.” Bioinformatics 26, no. 17
(September 1, 2010): 2071–75.

2-95
2 High-Throughput Sequence Analysis

See Also
BioMap | getBaseCoverage | getgenbank | getSummary

Related Examples
• “Identifying Differentially Expressed Genes from RNA-Seq Data” on page 2-32
• “Count Features from NGS Reads” on page 2-23
• “Exploring Genome-Wide Differences in DNA Methylation Profiles” on page 2-58

2-96
 Working with Illumina/Solexa Next-Generation Sequencing Data

Working with Illumina/Solexa Next-Generation Sequencing

Data

This example shows how to read and perform basic operations with data produced by the Illumina®/
Solexa Genome Analyzer®.

Introduction

During an analysis run with the Genome Analyzer Pipeline software, several intermediate files are
produced. In this example, you will learn how to read and manipulate the information contained in
sequence files (_sequence.txt).

Reading _sequence.txt (FASTQ) Files

The _sequence.txt files are FASTQ-formatted files that contain the sequence reads and their
quality scores, after quality trimming and filtering. You can use the fastqinfo function to display a
summary of the contents of a _sequence.txt file, and the fastqread function to read the contents
of the file. The output, reads, is a cell array of structures containing the Header, Sequence and
Quality fields.

filename = 'ilmnsolexa_sequence.txt';
info = fastqinfo(filename)
reads = fastqread(filename)

info =

struct with fields:

Filename: 'ilmnsolexa_sequence.txt'
FilePath: 'C:\TEMP\Bdoc22b_2134332_7304\ibD166FA\17\tp9f51422f\bioinfo-ex25447385'
FileModDate: '06-May-2009 16:02:48'
FileSize: 30124
NumberOfEntries: 260

reads =

1x260 struct array with fields:

Header
Sequence
Quality

Because there is one sequence file per tile, it is not uncommon to have a collection of over 1,000 files
in total. You can read the entire collection of files associated with a given analysis run by
concatenating the _sequence.txt files into a single file. However, because this operation usually
produces a large file that requires ample memory to be stored and processed, it is advisable to read
the content in chunks using the blockread option of the fastqread function. For example, you can
read the first M sequences, or the last M sequences, or any M sequences in the file.

M = 150;
N = info.NumberOfEntries;

2-97
2 High-Throughput Sequence Analysis

readsFirst = fastqread(filename, 'blockread', [1 M])

readsLast = fastqread(filename, 'blockread', [N-M+1, N])

readsFirst =

1x150 struct array with fields:

Header
Sequence
Quality

readsLast =

1x150 struct array with fields:

Header
Sequence
Quality

Surveying the Length Distribution of Sequence Reads

Once you load the sequence information into your workspace, you can determine the number and
length of the sequence reads and plot their distribution as follows:

seqs = {reads.Sequence};
readsLen = cellfun(@length, seqs);

figure(); hist(readsLen);
xlabel('Number of bases'); ylabel('Number of sequence reads');
title('Length distribution of sequence reads')

2-98
 Working with Illumina/Solexa Next-Generation Sequencing Data

As expected, in this example all sequence reads are 36 bp long.

Surveying the Base Composition of the Sequence Reads

You can also examine the nucleotide composition by surveying the number of occurrences of each
base type in each sequence read, as shown below:

nt = {'A', 'C', 'G', 'T'};

pos = cell(4,N);

for i = 1:4
pos(i,:) = strfind(seqs, nt{i});
end

count = zeros(4,N);
for i = 1:4
count(i,:) = cellfun(@length, pos(i,:));
end

%=== plot nucleotide distribution

figure();
subplot(2,2,1); hist(count(1,:)); title('A'); ylabel('Number of sequence reads');
subplot(2,2,2); hist(count(2,:)); title('C');
subplot(2,2,3); hist(count(3,:)); title('G'); xlabel('Occurrences'); ylabel('Number of sequence r
subplot(2,2,4); hist(count(4,:)); title('T'); xlabel('Occurrences');

figure(); hist(count');

2-99
2 High-Throughput Sequence Analysis

xlabel('Occurrences');
ylabel('Number of sequence reads');
legend('A', 'C', 'G', 'T');
title('Base distribution by nucleotide type');

2-100
 Working with Illumina/Solexa Next-Generation Sequencing Data

Surveying the Quality Score Distribution

Each sequence read in the _sequence.txt file is associated with a score. The score is defined as SQ
= -10 * log10 (p / (1-p)), where p is the probability error of a base. You can examine the quality scores
associated with the base calls by converting the ASCII format into a numeric representation, and then
plotting their distribution, as shown below:

sq = {reads.Quality}; % in ASCII format

SQ = cellfun(@(x) double(x)-64, {reads.Quality}, 'UniformOutput', false); % in integer format

%=== average, median and standard deviation

avgSQ = cellfun(@mean, SQ);
medSQ = cellfun(@median, SQ);
stdSQ = cellfun(@std, SQ);

%=== plot distribution of median and average quality

figure();
subplot(1,2,1); hist(medSQ);
xlabel('Median Score SQ'); ylabel('Number of sequence reads');
subplot(1,2,2); boxplot(avgSQ); ylabel('Average Score SQ');

2-101
2 High-Throughput Sequence Analysis

Converting Quality Scores Between Standards

The quality scores found in Solexa/Illumina files are asymptotic, but not identical, to the quality
scores used in the Sanger standard (Phred-like scores, Q). Q is defined as -10 * log10 (p), where p is
the error probability of a base. For example, if the quality score of a base is Q = 20, then p = 10
^(-20/10) = .01. This means that there is one wrong base call every 100 base calls with a score of20.

While Phred quality scores are positive integers, Solexa/Illumina quality scores can be negative. We
can convert Solexa quality scores into Phred quality scores using the following code:

%=== convert from Solexa to Sanger standard

Q = cellfun(@(x) floor(.499 + 10 * log10(1+ 10 .^ (x/10))), SQ, ...
'UniformOutput', false); % in integer format
q = cellfun(@(x) char(x+33), Q, 'UniformOutput', false); % in ascii format

sanger = q(1:3)'
solexa = sq(1:3)'

sanger =

3x1 cell array

{'>>>>>>>>>>>>:>:>>>>>>>>>>>>7&*7.1-%4'}
{'>>>>>>>>>>>>:>>>>>>>>>:17>5><1;1+&&,'}
{'>>>>:>>>>>7>5>>>>>5>>>>>7>5.+'69'(-%'}

2-102
 Working with Illumina/Solexa Next-Generation Sequencing Data

solexa =

3x1 cell array

{']]]]]]]]]]]]Y]Y]]]]]]]]]]]]VCHVMPLAS'}
{']]]]]]]]]]]]Y]]]]]]]]]YPV]T][PZPICCK'}
{']]]]Y]]]]]V]T]]]]]T]]]]]V]TMJEUXEFLA'}

Filtering and Masking According to Quality Scores

Signal purity filtering has already been applied to the sequences in the _sequence.txt files. You
can perform additional filtering, for example by considering only those sequence reads whose bases
have all quality scores above a specific threshold:

%=== find sequence reads whose bases all have quality above threshold
len = 36;
qt = 10; % minimum quality threshold
a = cellfun(@(x) x > qt, SQ, 'UniformOutput', false);
b = cellfun(@sum, a);
c1 = find(b == len);
n1= numel(c1); % number of sequence reads passing the filter

disp([num2str(n1) ' sequence reads have all bases above threshold ' num2str(qt)]);

30 sequence reads have all bases above threshold 10

Alternatively, you can consider only those sequence reads that have less than a given number of bases
with quality scores below threshold:

%=== find sequence reads having less than M bases with quality below threshold
M = 5; % max number of bases with poor quality
a = cellfun(@(x) x <= qt, SQ, 'UniformOutput', false);
b = cellfun(@sum, a);
c2 = find(b <= M);
n2 = numel(c2); % number of sequence reads passing the filter

disp([num2str(n2) ' sequence reads have less than ' num2str(M) ' bases below threshold ' num2str(

235 sequence reads have less than 5 bases below threshold 10

Finally, you can apply a lower case mask to those bases that have quality scores below threshold:

seq = reads(1).Sequence
mseq = seq;
qt2 = 20; % quality threshold
mask = SQ{1} < qt;
mseq(mask) = lower(seq(mask))

seq =

'GGACTTTGTAGGATACCCTCGCTTTCCTTCTCCTGT'

mseq =

2-103
2 High-Throughput Sequence Analysis

'GGACTTTGTAGGATACCCTCGCTTTCCTtcTCCTgT'

Summarizing Read Occurrences

To summarize read occurrences, you can determine the number of unique read sequences and their
distribution across the data set. You can also identify those sequence reads that occur multiple times,
often because they correspond to adapters or primers used in the sequencing process.

%=== determine read frequency

[uReads,~,n] = unique({reads.Sequence});
numUnique = numel(uReads)
readFreq = accumarray(n(:),1);
figure(); hist(readFreq, unique(readFreq));
xlabel('Occurrences'); ylabel('Number of sequence reads');
title('Read occurrences');

%=== identify multiply-occurring sequence reads

d = readFreq > 1;
dupReads = uReads(d)'
dupFreq = readFreq(d)'

numUnique =

250

dupReads =

9x1 cell array

{'AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA'}
{'GATTTTATTGGTATCAGGGTTAATCGTGCCAAGAAA'}
{'GCATGGGTGATGCTGGTATTAAATCTGCCATTCAAG'}
{'GGGATGAACATAATAAGCAATGACGGCAGCAATAAA'}
{'GGGGGAGCACATTGTAGCATTGTGCCAATTCATCCA'}
{'GGTTATTAAAGAGATTATTTGTCTCCAGCCACTTAA'}
{'GTTCTCACTTCTGTTACTCCAGCTTCTTCGGCACCT'}
{'GTTGCTGCCATCTCAAAAACATTTGGACTGCTCCGC'}
{'GTTGGTTTCTATGTGGCTAAATACGTTAACAAAAAG'}

dupFreq =

2 2 2 2 2 2 3 2 2

2-104
 Working with Illumina/Solexa Next-Generation Sequencing Data

Identifying Homopolymers Artifacts

Illumina/Solexa sequencing may produce false polyA at the edges of a tile. To identify these artifacts,
you need to identify homopolymers, that is, sequence reads composed of one type of nucleotide only.
In the data set under consideration, there are two homopolymers, both of which are polyA.

%=== find homopolymers

pc = (count ./ len) * 100;
[homopolType,homopolIndex] = find(pc == 100);

homopolIndex
homopol = {reads(homopolIndex).Sequence}'

homopolIndex =

251
257

homopol =

2x1 cell array

{'AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA'}
{'AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA'}

2-105
2 High-Throughput Sequence Analysis

Similarly, you can identify sequence reads that are near-matches to homopolymers, that is, sequence
reads that are composed almost exclusively of one nucleotide type.

%=== find near-homopolymers

[nearhomopolType, nearhomopolIndex] = find(pc < 100 & pc > 85); % more than 85% same base
nearhomopolIndex
nearHomopol = {reads(nearhomopolIndex).Sequence}'

nearhomopolIndex =

4
243

nearHomopol =

2x1 cell array

{'AAAAACATAAAAAAAAAAATAAAAAAACAAAAAAAA'}
{'AAAAAAATAAAAAAAAAAATAAAAAAAAATTAAAAA'}

Writing Data to FASTQ Format

Once you have processed and analyzed your data, it might be convenient to save a subset of
sequences in a separate FASTQ file for future consideration. For this purpose you can use the
fastqwrite function.

2-106
 3

Sequence Analysis

Sequence analysis is the process you use to find information about a nucleotide or amino acid
sequence using computational methods. Common tasks in sequence analysis are identifying genes,
determining the similarity of two genes, determining the protein coded by a gene, and determining
the function of a gene by finding a similar gene in another organism with a known function.

• “Exploring a Nucleotide Sequence Using Command Line” on page 3-2

• “Exploring a Nucleotide Sequence Using the Sequence Viewer App” on page 3-15
• “Explore a Protein Sequence Using the Sequence Viewer App” on page 3-26
• “Compare Sequences Using Sequence Alignment Algorithms” on page 3-30
• “View and Align Multiple Sequences” on page 3-41
• “Analyzing Synonymous and Nonsynonymous Substitution Rates” on page 3-55
• “Investigating the Bird Flu Virus” on page 3-65
• “Exploring Primer Design” on page 3-81
• “Identifying Over-Represented Regulatory Motifs” on page 3-91
• “Predicting and Visualizing the Secondary Structure of RNA Sequences” on page 3-102
• “Using HMMs for Profile Analysis of a Protein Family” on page 3-114
• “Predicting Protein Secondary Structure Using a Neural Network” on page 3-131
• “Visualizing the Three-Dimensional Structure of a Molecule” on page 3-148
• “Calculating and Visualizing Sequence Statistics” on page 3-163
• “Aligning Pairs of Sequences” on page 3-177
• “Assessing the Significance of an Alignment” on page 3-185
• “Using Scoring Matrices to Measure Evolutionary Distance” on page 3-194
• “Calling Bioperl Functions from MATLAB” on page 3-198
• “Accessing NCBI Entrez Databases with E-Utilities” on page 3-210
3 Sequence Analysis

Exploring a Nucleotide Sequence Using Command Line

In this section...
“Overview of Example” on page 3-2
“Searching the Web for Sequence Information” on page 3-2
“Reading Sequence Information from the Web” on page 3-4
“Determining Nucleotide Composition” on page 3-5
“Determining Codon Composition” on page 3-8
“Open Reading Frames” on page 3-11
“Amino Acid Conversion and Composition” on page 3-13

Overview of Example
After sequencing a piece of DNA, one of the first tasks is to investigate the nucleotide content in the
sequence. Starting with a DNA sequence, this example uses sequence statistics functions to
determine mono-, di-, and trinucleotide content, and to locate open reading frames.

Searching the Web for Sequence Information

The following procedure illustrates how to use the MATLAB Help browser to search the Web for
information. In this example you are interested in studying the human mitochondrial genome. While
many genes that code for mitochondrial proteins are found in the cell nucleus, the mitochondrial has
genes that code for proteins used to produce energy.

First research information about the human mitochondria and find the nucleotide sequence for the
genome. Next, look at the nucleotide content for the entire sequence. And finally, determine open
reading frames and extract specific gene sequences.

1 Use the MATLAB Help browser to explore the Web. In the MATLAB Command Window, type

web('http://www.ncbi.nlm.nih.gov/')

A separate browser window opens with the home page for the NCBI Web site.
2 Search the NCBI Web site for information. For example, to search for the human mitochondrion
genome, from the Search list, select Genome , and in the Search list, enter mitochondrion
homo sapiens.

The NCBI Web search returns a list of links to relevant pages.

3-2
 Exploring a Nucleotide Sequence Using Command Line

3 Select a result page. For example, click the link labeled NC_012920.

The MATLAB Help browser displays the NCBI page for the human mitochondrial genome.

3-3
3 Sequence Analysis

Reading Sequence Information from the Web

The following procedure illustrates how to find a nucleotide sequence in a public database and read
the sequence information into the MATLAB environment. Many public databases for nucleotide
sequences are accessible from the Web. The MATLAB Command Window provides an integrated
environment for bringing sequence information into the MATLAB environment.

The consensus sequence for the human mitochondrial genome has the GenBank accession number
NC_012920. Since the whole GenBank entry is quite large and you might only be interested in the
sequence, you can get just the sequence information.

3-4
 Exploring a Nucleotide Sequence Using Command Line

1 Get sequence information from a Web database. For example, to retrieve sequence information
for the human mitochondrial genome, in the MATLAB Command Window, type

mitochondria = getgenbank('NC_012920','SequenceOnly',true)

The getgenbank function retrieves the nucleotide sequence from the GenBank database and
creates a character array.

mitochondria =
GATCACAGGTCTATCACCCTATTAACCACTCACGGGAGCTCTCCATGCAT
TTGGTATTTTCGTCTGGGGGGTGTGCACGCGATAGCATTGCGAGACGCTG
GAGCCGGAGCACCCTATGTCGCAGTATCTGTCTTTGATTCCTGCCTCATT
CTATTATTTATCGCACCTACGTTCAATATTACAGGCGAACATACCTACTA
AAGT . . .
2 If you don't have a Web connection, you can load the data from a MAT file included with the
Bioinformatics Toolbox software, using the command

load mitochondria

The load function loads the sequence mitochondria into the MATLAB Workspace.
3 Get information about the sequence. Type

whos mitochondria

Information about the size of the sequence displays in the MATLAB Command Window.

Name Size Bytes Class Attributes

mitochondria 1x16569 33138 char

Determining Nucleotide Composition

The following procedure illustrates how to determine the monomers and dimers, and then visualize
data in graphs and bar plots. Sections of a DNA sequence with a high percent of A+T nucleotides
usually indicate intergenic parts of the sequence, while low A+T and higher G+C nucleotide
percentages indicate possible genes. Many times high CG dinucleotide content is located before a
gene.

After you read a sequence into the MATLAB environment, you can use the sequence statistics
functions to determine if your sequence has the characteristics of a protein-coding region. This
procedure uses the human mitochondrial genome as an example. See “Reading Sequence Information
from the Web” on page 3-4.

1 Plot monomer densities and combined monomer densities in a graph. In the MATLAB Command
Window, type

ntdensity(mitochondria)

This graph shows that the genome is A+T rich.

3-5
3 Sequence Analysis

2 Count the nucleotides using the basecount function.

basecount(mitochondria)

A list of nucleotide counts is shown for the 5'-3' strand.

ans =
A: 5124
C: 5181
G: 2169
T: 4094

3 Count the nucleotides in the reverse complement of a sequence using the seqrcomplement
function.
basecount(seqrcomplement(mitochondria))

As expected, the nucleotide counts on the reverse complement strand are complementary to the
5'-3' strand.
ans =
A: 4094
C: 2169
G: 5181
T: 5124
4 Use the function basecount with the chart option to visualize the nucleotide distribution.
figure
basecount(mitochondria,'chart','pie');

A pie chart displays in the MATLAB Figure window.

3-6
 Exploring a Nucleotide Sequence Using Command Line

5 Count the dimers in a sequence and display the information in a bar chart.

figure
dimercount(mitochondria,'chart','bar')

ans =

AA: 1604
AC: 1495
AG: 795
AT: 1230
CA: 1534
CC: 1771
CG: 435
CT: 1440
GA: 613
GC: 711
GG: 425
GT: 419
TA: 1373
TC: 1204
TG: 513
TT: 1004

3-7
3 Sequence Analysis

Determining Codon Composition

The following procedure illustrates how to look at codons for the six reading frames. Trinucleotides
(codon) code for an amino acid, and there are 64 possible codons in a nucleotide sequence. Knowing
the percent of codons in your sequence can be helpful when you are comparing with tables for
expected codon usage.

After you read a sequence into the MATLAB environment, you can analyze the sequence for codon
composition. This procedure uses the human mitochondria genome as an example. See “Reading
Sequence Information from the Web” on page 3-4.

1 Count codons in a nucleotide sequence. In the MATLAB Command Window, type

codoncount(mitochondria)

The codon counts for the first reading frame displays.

AAA - 167 AAC - 171 AAG - 71 AAT - 130

ACA - 137 ACC - 191 ACG - 42 ACT - 153
AGA - 59 AGC - 87 AGG - 51 AGT - 54
ATA - 126 ATC - 131 ATG - 55 ATT - 113
CAA - 146 CAC - 145 CAG - 68 CAT - 148
CCA - 141 CCC - 205 CCG - 49 CCT - 173
CGA - 40 CGC - 54 CGG - 29 CGT - 27
CTA - 175 CTC - 142 CTG - 74 CTT - 101
GAA - 67 GAC - 53 GAG - 49 GAT - 35
GCA - 81 GCC - 101 GCG - 16 GCT - 59
GGA - 36 GGC - 47 GGG - 23 GGT - 28
GTA - 43 GTC - 26 GTG - 18 GTT - 41

3-8
 Exploring a Nucleotide Sequence Using Command Line

TAA - 157 TAC - 118 TAG - 94 TAT - 107

TCA - 125 TCC - 116 TCG - 37 TCT - 103
TGA - 64 TGC - 40 TGG - 29 TGT - 26
TTA - 96 TTC - 107 TTG - 47 TTT - 78
2 Count the codons in all six reading frames and plot the results in heat maps.
for frame = 1:3
figure
subplot(2,1,1);
codoncount(mitochondria,'frame',frame,'figure',true,...
'geneticcode','Vertebrate Mitochondrial');
title(sprintf('Codons for frame %d',frame));
subplot(2,1,2);
codoncount(mitochondria,'reverse',true,'frame',frame,...
'figure',true,'geneticcode','Vertebrate Mitochondrial');
title(sprintf('Codons for reverse frame %d',frame));
end

Heat maps display all 64 codons in the 6 reading frames.

3-9
3 Sequence Analysis

3-10
 Exploring a Nucleotide Sequence Using Command Line

Open Reading Frames

The following procedure illustrates how to locate the open reading frames using a specific genetic
code. Determining the protein-coding sequence for a eukaryotic gene can be a difficult task because
introns (noncoding sections) are mixed with exons. However, prokaryotic genes generally do not have
introns and mRNA sequences have the introns removed. Identifying the start and stop codons for
translation determines the protein-coding section, or open reading frame (ORF), in a sequence. Once
you know the ORF for a gene or mRNA, you can translate a nucleotide sequence to its corresponding
amino acid sequence.

After you read a sequence into the MATLAB environment, you can analyze the sequence for open
reading frames. This procedure uses the human mitochondria genome as an example. See “Reading
Sequence Information from the Web” on page 3-4.
1 Display open reading frames (ORFs) in a nucleotide sequence. In the MATLAB Command
Window, type:

3-11
3 Sequence Analysis

seqshoworfs(mitochondria);

If you compare this output to the genes shown on the NCBI page for NC_012920, there are fewer
genes than expected. This is because vertebrate mitochondria use a genetic code slightly
different from the standard genetic code. For a list of genetic codes, see the Genetic Code table
in the aa2nt reference page.
2 Display ORFs using the Vertebrate Mitochondrial code.

orfs= seqshoworfs(mitochondria,...
'GeneticCode','Vertebrate Mitochondrial',...
'alternativestart',true);

Notice that there are now two large ORFs on the third reading frame. One starts at position 4470
and the other starts at 5904. These correspond to the genes ND2 (NADH dehydrogenase subunit
2 [Homo sapiens] ) and COX1 (cytochrome c oxidase subunit I) genes.
3 Find the corresponding stop codon. The start and stop positions for ORFs have the same indices
as the start positions in the fields Start and Stop.

ND2Start = 4470;
StartIndex = find(orfs(3).Start == ND2Start)
ND2Stop = orfs(3).Stop(StartIndex)

The stop position displays.

ND2Stop =

5511
4 Using the sequence indices for the start and stop of the gene, extract the subsequence from the
sequence.

ND2Seq = mitochondria(ND2Start:ND2Stop)

The subsequence (protein-coding region) is stored in ND2Seq and displayed on the screen.

attaatcccctggcccaacccgtcatctactctaccatctttgcaggcac
actcatcacagcgctaagctcgcactgattttttacctgagtaggcctag
aaataaacatgctagcttttattccagttctaaccaaaaaaataaaccct
cgttccacagaagctgccatcaagtatttcctcacgcaagcaaccgcatc
cataatccttc . . .
5 Determine the codon distribution.

codoncount (ND2Seq)

The codon count shows a high amount of ACC, ATA, CTA, and ATC.

AAA - 10 AAC - 14 AAG - 2 AAT - 6

ACA - 11 ACC - 24 ACG - 3 ACT - 5
AGA - 0 AGC - 4 AGG - 0 AGT - 1
ATA - 23 ATC - 24 ATG - 1 ATT - 8
CAA - 8 CAC - 3 CAG - 2 CAT - 1
CCA - 4 CCC - 12 CCG - 2 CCT - 5
CGA - 0 CGC - 3 CGG - 0 CGT - 1
CTA - 26 CTC - 18 CTG - 4 CTT - 7
GAA - 5 GAC - 0 GAG - 1 GAT - 0
GCA - 8 GCC - 7 GCG - 1 GCT - 4
GGA - 5 GGC - 7 GGG - 0 GGT - 1

3-12
 Exploring a Nucleotide Sequence Using Command Line

GTA - 3 GTC - 2 GTG - 0 GTT - 3

TAA - 0 TAC - 8 TAG - 0 TAT - 2
TCA - 7 TCC - 11 TCG - 1 TCT - 4
TGA - 10 TGC - 0 TGG - 1 TGT - 0
TTA - 8 TTC - 7 TTG - 1 TTT - 8
6 Look up the amino acids for codons ATA, CTA, ACC, and ATC.

aminolookup('code',nt2aa('ATA'))
aminolookup('code',nt2aa('CTA'))
aminolookup('code',nt2aa('ACC'))
aminolookup('code',nt2aa('ATC'))

The following displays:

Ile isoleucine
Leu leucine
Thr threonine
Ile isoleucine

Amino Acid Conversion and Composition

The following procedure illustrates how to extract the protein-coding sequence from a gene sequence
and convert it to the amino acid sequence for the protein. Determining the relative amino acid
composition of a protein will give you a characteristic profile for the protein. Often, this profile is
enough information to identify a protein. Using the amino acid composition, atomic composition, and
molecular weight, you can also search public databases for similar proteins.

After you locate an open reading frame (ORF) in a gene, you can convert it to an amino sequence and
determine its amino acid composition. This procedure uses the human mitochondria genome as an
example. See “Open Reading Frames” on page 3-11.

1 Convert a nucleotide sequence to an amino acid sequence. In this example, only the protein-
coding sequence between the start and stop codons is converted.

ND2AASeq = nt2aa(ND2Seq,'geneticcode',...
'Vertebrate Mitochondrial')

The sequence is converted using the Vertebrate Mitochondrial genetic code. Because the
property AlternativeStartCodons is set to 'true' by default, the first codon att is
converted to M instead of I.

MNPLAQPVIYSTIFAGTLITALSSHWFFTWVGLEMNMLAFIPVLTKKMNP
RSTEAAIKYFLTQATASMILLMAILFNNMLSGQWTMTNTTNQYSSLMIMM
AMAMKLGMAPFHFWVPEVTQGTPLTSGLLLLTWQKLAPISIMYQISPSLN
VSLLLTLSILSIMAGSWGGLNQTQLRKILAYSSITHMGWMMAVLPYNPNM
TILNLTIYIILTTTAFLLLNLNSSTTTLLLSRTWNKLTWLTPLIPSTLLS
LGGLPPLTGFLPKWAIIEEFTKNNSLIIPTIMATITLLNLYFYLRLIYST
SITLLPMSNNVKMKWQFEHTKPTPFLPTLIALTTLLLPISPFMLMIL
2 Compare your conversion with the published conversion in the GenPept database.

ND2protein = getgenpept('YP_003024027','sequenceonly',true)

The getgenpept function retrieves the published conversion from the NCBI database and reads
it into the MATLAB Workspace.
3 Count the amino acids in the protein sequence.

3-13
3 Sequence Analysis

aacount(ND2AASeq, 'chart','bar')

A bar graph displays. Notice the high content for leucine, threonine and isoleucine, and also
notice the lack of cysteine and aspartic acid.

4 Determine the atomic composition and molecular weight of the protein.

atomiccomp(ND2AASeq)
molweight (ND2AASeq)

The following displays in the MATLAB Workspace:

ans =

C: 1818
H: 2882
N: 420
O: 471
S: 25

ans =

3.8960e+004

If this sequence was unknown, you could use this information to identify the protein by
comparing it with the atomic composition of other proteins in a database.

3-14
 Exploring a Nucleotide Sequence Using the Sequence Viewer App

Exploring a Nucleotide Sequence Using the Sequence Viewer

App
In this section...
“Overview of the Sequence Viewer” on page 3-15
“Importing a Sequence into the Sequence Viewer” on page 3-15
“Viewing Nucleotide Sequence Information” on page 3-17
“Searching for Words” on page 3-19
“Exploring Open Reading Frames” on page 3-22
“Closing the Sequence Viewer” on page 3-25

Overview of the Sequence Viewer

The Sequence Viewer integrates many of the sequence functions in the Bioinformatics Toolbox
toolbox. Instead of entering commands in the MATLAB Command Window, you can select and enter
options using the app.

Importing a Sequence into the Sequence Viewer

The first step when analyzing a nucleotide or amino acid sequence is to import sequence information
into the MATLAB environment. The Sequence Viewer can connect to Web databases such as NCBI
and EMBL and read information into the MATLAB environment.

The following procedure illustrates how to retrieve sequence information from the NCBI database on
the Web. This example uses the GenBank accession number NM_000520, which is the human gene
HEXA that is associated with Tay-Sachs disease.

Note Data in public repositories is frequently curated and updated; therefore, the results of this
example might be slightly different when you use up-to-date sequences.

1 In the MATLAB Command Window, type

seqviewer

Alternatively, click Sequence Viewer on the Apps tab.

The Sequence Viewer opens without a sequence loaded. Notice that the panes to the right and
bottom are blank.
2 To retrieve a sequence from the NCBI database, select File > Download Sequence from >
NCBI.

The Download Sequence from NCBI dialog box opens.

3-15
3 Sequence Analysis

3 In the Enter Sequence box, type an accession number for an NCBI database entry, for example,
NM_000520. Click the Nucleotide option button, and then click OK.

The MATLAB software accesses the NCBI database on the Web, loads nucleotide sequence
information for the accession number you entered, and calculates some basic statistics.

3-16
 Exploring a Nucleotide Sequence Using the Sequence Viewer App

Viewing Nucleotide Sequence Information

After you import a sequence into the Sequence Viewer app, you can read information stored with
the sequence, or you can view graphic representations for ORFs and CDSs.

1 In the left pane tree, click Comments. The right pane displays general information about the
sequence.

3-17
3 Sequence Analysis

2 Now click Features. The right pane displays NCBI feature information, including index numbers
for a gene and any CDS sequences.
3 Click ORF to show the search results for ORFs in the six reading frames.

3-18
 Exploring a Nucleotide Sequence Using the Sequence Viewer App

4 Click Annotated CDS to show the protein coding part of a nucleotide sequence.

Searching for Words

You can also search for characteristic words or sequence patterns using regular expressions. You can
enter the IUB/IUPAC nucleotide and amino acid symbols that are automatically converted to
corresponding nucleotides and amino acids accordingly. For details about how symbols are
interpreted, see the Nucleotide Conversion and Amino Acid Conversion tables of seq2regexp.

3-19
3 Sequence Analysis

For instance, if you search for the word 'TAR' with the Regular Expression box checked, the app
highlights all the occurrences of 'TAA' and 'TAG' in the sequence since R = [AG].

1 Select Sequence > Find Word.

2 In the Find Word dialog box, type a sequence word or pattern, for example, atg, and then click
Find.

The Sequence Viewer searches and displays the location of the selected word.

3-20
 Exploring a Nucleotide Sequence Using the Sequence Viewer App

3
Clear the display by clicking the Clear Word Selection button on the toolbar.

3-21
3 Sequence Analysis

Exploring Open Reading Frames

The following procedure illustrates how to identify the protein coding part of a nucleotide sequence
and copy it into a new view. Identifying coding sections of a nucleotide sequence is a common
bioinformatics task. After locating the coding part of a sequence, you can copy it to a new view,
translate it to an amino acid sequence, and continue with your analysis.

1 In the left pane, click ORF.

The Sequence Viewer displays the ORFs for the six reading frames in the lower-right pane.
Hover the cursor over a frame to display information about it.

2 Click the longest ORF on reading frame 2.

The ORF is highlighted to indicate the part of the sequence that is selected.

3 Right-click the selected ORF and then select Export to Workspace. In the Export to MATLAB
Workspace dialog box, type a variable name, for example, NM_000520_ORF_2, then click
Export.

3-22
 Exploring a Nucleotide Sequence Using the Sequence Viewer App

The NM_000520_ORF_2 variable is added to the MATLAB Workspace.

4 Select File > Import from Workspace. Type the name of a variable with an exported ORF, for
example, NM_000520_ORF_2, and then click Import.

The Sequence Viewer adds a tab at the bottom for the new sequence while leaving the original
sequence open.

3-23
3 Sequence Analysis

5 In the left pane, click Full Translation. Select Display > Amino Acid Residue Display > One
Letter Code.

The Sequence Viewer displays the amino acid sequence below the nucleotide sequence.

3-24
 Exploring a Nucleotide Sequence Using the Sequence Viewer App

Closing the Sequence Viewer

Close the Sequence Viewer from the MATLAB command line using the following syntax:

seqviewer('close')

3-25
3 Sequence Analysis

Explore a Protein Sequence Using the Sequence Viewer App

In this section...
“Overview of the Sequence Viewer” on page 3-26
“Viewing Amino Acid Sequence Statistics” on page 3-26
“Closing the Sequence Viewer” on page 3-28
“References” on page 3-29

Overview of the Sequence Viewer

The Sequence Viewer app integrates many of the sequence functions in the Bioinformatics Toolbox
toolbox. Instead of entering commands in the MATLAB Command Window, you can select and enter
options using the app.

Viewing Amino Acid Sequence Statistics

The following procedure illustrates how to view an amino acid sequence for an ORF located in a
nucleotide sequence. You can import your own amino acid sequence, or you can get a protein
sequence from the GenBank database. This example uses the GenBank accession number
NP_000511, which is the alpha subunit for a human enzyme associated with Tay-Sachs disease.

1 Select File > Download Sequence from > NCBI.

The Download Sequence from NCBI dialog box opens.

2 In the dialog box, type an accession number for an NCBI database entry, for example,
NP_000511. Click the Protein option button, and then click OK.

The Sequence Viewer accesses the NCBI database on the Web and loads amino acid sequence
information for the accession number you entered.

3-26
 Explore a Protein Sequence Using the Sequence Viewer App

3 Select Display > Amino Acid Color Scheme, and then select Charge, Function,
Hydrophobicity, Structure, or Taylor. For example, select Function.

The display colors change to highlight charge information about the amino acid residues. The
following table shows color legends for the amino acid color schemes.

3-27
3 Sequence Analysis

Amino Acid Color Scheme Color Legend

Charge • Acidic — Red
• Basic — Light Blue
• Neutral — Black
Function • Acidic — Red
• Basic — Light Blue
• Hydropobic, nonpolar — Black
• Polar, uncharged — Green
Hydrophobicity • Hydrophilic — Light Blue
• Hydrophobic — Black
Structure • Ambivalent — Dark Green
• External — Light Blue
• Internal — Orange
Taylor Each amino acid is assigned its own color, based on the
colors proposed by W.R. Taylor on page 3-29.

Closing the Sequence Viewer

Close the Sequence Viewer from the MATLAB command line using the following syntax:

seqviewer('close')

3-28
 Explore a Protein Sequence Using the Sequence Viewer App

References

[1] Taylor, W.R. (1997). Residual colours: a proposal for aminochromography. Protein Engineering 10,
7, 743–746.

3-29
3 Sequence Analysis

Compare Sequences Using Sequence Alignment Algorithms

Determining the similarity between two sequences is a common task in computational biology.
Starting with a nucleotide sequence for a human gene, this example uses alignment algorithms to
locate and verify a corresponding gene in a model organism.

In this example, you are interested in studying Tay-Sachs disease. Tay-Sachs is an autosomal
recessive disease caused by the absence of the enzyme beta-hexosaminidase A (Hex A). This enzyme
is responsible for the breakdown of gangliosides (GM2) in brain and nerve cells.

First, research information about Tay-Sachs and the enzyme that is associated with this disease, then
find the nucleotide sequence for the human gene that codes for the enzyme, and finally find a
corresponding gene in another organism to use as a model for study.

In the MATLAB Command window, enter:

web('https://www.ncbi.nlm.nih.gov/books/NBK22250/')

Your help browser opens with the Tay-Sachs disease page in the Genes and Diseases section of the
NCBI web site. This section provides a comprehensive introduction to medical genetics. In particular,
this page contains an introduction and pictorial representation of the enzyme Hex A and its role in
the metabolism of the lipid GM2 ganglioside.

After completing your research, you have concluded the following:

The gene HEXA codes for the alpha subunit of the dimer enzyme hexosaminidase A (Hex A), while the
gene HEXB codes for the beta subunit of the enzyme. A third gene, GM2A, codes for the activator
protein GM2. However, it is a mutation in the gene HEXA that causes Tay-Sachs.

Retrieve Sequence Information from a Public Database

The following procedure illustrates how to find the nucleotide sequence for a human gene in a public
database and read the sequence information into the MATLAB environment. Many public databases
for nucleotide sequences (for example, GenBank®, EMBL-EBI) are accessible from the Web. The
MATLAB Command Window with the MATLAB Help browser provide an integrated environment for
searching the Web and bringing sequence information into the MATLAB environment.

After you locate a sequence, you need to move the sequence data into the MATLAB Workspace.

Open the MATLAB Help browser to the NCBI Web site. In the MATLAB Command Window, enter:
web('https://www.ncbi.nlm.nih.gov/')

Search for the gene you are interested in studying. For example, from the Search list, select
Nucleotide, and in the for box enter Tay-Sachs. Look for the genes that code the alpha and beta
subunits of the enzyme hexosaminidase A (Hex A), and the gene that codes the activator enzyme. The
NCBI reference for the human gene HEXA has accession number NM_000520.

Get sequence data into the MATLAB environment. For example, to get sequence information for the
human gene HEXA, enter:
humanHEXA = getgenbank('NM_000520')

humanHEXA = struct with fields:

LocusName: 'NM_000520'

3-30
 Compare Sequences Using Sequence Alignment Algorithms

LocusSequenceLength: '4785'
LocusNumberofStrands: ''
LocusTopology: 'linear'
LocusMoleculeType: 'mRNA'
LocusGenBankDivision: 'PRI'
LocusModificationDate: '18-JAN-2021'
Definition: 'Homo sapiens hexosaminidase subunit alpha (HEXA), transcript variant
Accession: 'NM_000520'
Version: 'NM_000520.6'
GI: ''
Project: []
DBLink: []
Keywords: 'RefSeq; MANE Select.'
Segment: []
Source: 'Homo sapiens (human)'
SourceOrganism: [4×65 char]
Reference: {[1×1 struct] [1×1 struct] [1×1 struct] [1×1 struct] [1×1 struct]
Comment: [48×66 char]
Features: [160×74 char]
CDS: [1×1 struct]
Sequence: 'ctcacgtggccagccccctccgagaggggagaccagcgggccatgacaagctccaggctttggttttcg
SearchURL: 'https://www.ncbi.nlm.nih.gov/entrez/viewer.fcgi?db=nuccore&id=NM_0005
RetrieveURL: 'https://eutils.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi?db=nuccore&

Search a Public Database for Related Genes

The following procedure illustrates how to find the nucleotide sequence for a mouse gene related to a
human gene, and read the sequence information into the MATLAB environment. The sequence and
function of many genes is conserved during the evolution of species through homologous genes.
Homologous genes are genes that have a common ancestor and similar sequences. One goal of
searching a public database is to find similar genes. If you are able to locate a sequence in a database
that is similar to your unknown gene or protein, it is likely that the function and characteristics of the
known and unknown genes are the same.

After finding the nucleotide sequence for a human gene, you can do a BLAST search or search in the
genome of another organism for the corresponding gene. This procedure uses the mouse genome as
an example.

In the MATLAB Command window, enter:

web('http://www.ncbi.nlm.nih.gov')

Search the nucleotide database for the gene or protein you are interested in studying. For example,
from the Search list, select Nucleotide, and in the for box enter hexosaminidase A.

The search returns entries for the mouse and human genomes. For the purposes of this example, use
the accession number AK080777 for the mouse gene HEXA.

Get sequence information for the mouse gene into the MATLAB environment.

mouseHEXA = getgenbank('AK080777')

Locate Protein Coding Sequences

The following procedure illustrates how to convert a sequence from nucleotides to amino acids and
identify the open reading frames. A nucleotide sequence includes regulatory sequences before and

3-31
3 Sequence Analysis

after the protein coding section. By analyzing this sequence, you can determine the nucleotides that
code for the amino acids in the final protein.

After you have a list of genes you are interested in studying, you can determine the protein coding
sequences. This procedure uses the human gene HEXA and mouse gene HEXA as an example.

If you did not retrieve gene data from the Web, you can load example data from a MAT-file included
with the Bioinformatics Toolbox™ software. In the MATLAB Command window, enter:

load hexosaminidase

Locate open reading frames (ORFs) in the human gene. For example, for the human gene HEXA,
enter:

humanORFs = seqshoworfs(humanHEXA.Sequence)

humanORFs=1×3 struct array with fields:

Start
Stop

seqshoworfs creates the output structure humanORFs. This structure contains the position of the
start and stop codons for all open reading frames (ORFs) on each reading frame. The figure displays

3-32
 Compare Sequences Using Sequence Alignment Algorithms

the three reading frames with the ORFs colored blue, red, and green. Notice that the longest ORF is
in the first reading frame.

Locate open reading frames (ORFs) in the mouse gene. Enter:

mouseORFs = seqshoworfs(mouseHEXA.Sequence)

mouseORFs=1×3 struct array with fields:

Start
Stop

The mouse gene shows the longest ORF on the first reading frame.

Compare Amino Acid Sequences

The following procedure illustrates how to use global and local alignment functions to compare two
amino acid sequences. You could use alignment functions to look for similarities between two
nucleotide sequences, but alignment functions return more biologically meaningful results when you
are using amino acid sequences.

After you have located the open reading frames on your nucleotide sequences, you can convert the
protein coding sections of the nucleotide sequences to their corresponding amino acid sequences,
and then you can compare them for similarities.

3-33
3 Sequence Analysis

Using the open reading frames identified previously, convert the human and mouse DNA sequences to
the amino acid sequences. Because both the human and mouse HEXA genes were in the first reading
frames (default), you do not need to indicate which frame.
humanProtein = nt2aa(humanHEXA.Sequence);
mouseProtein = nt2aa(mouseHEXA.Sequence);

Draw a dot plot comparing the human and mouse amino acid sequences. Dot plots are one of the
easiest ways to look for similarity between sequences. The diagonal line shown below indicates that
there may be a good alignment between the two sequences.
warning('off','bioinfo:seqdotplot:imageTooBigForScreen');
seqdotplot(mouseProtein,humanProtein,4,3);
ylabel('Mouse hexosaminidase A (alpha subunit)')
xlabel('Human hexosaminidase A (alpha subunit)')
uif = gcf;
uif.Position(:) = [100 100 1280 800]; % Resize the figure.

warning('on','bioinfo:seqdotplot:imageTooBigForScreen');

Globally align the two amino acid sequences, using the Needleman-Wunsch algorithm.
[GlobalScore, GlobalAlignment] = nwalign(humanProtein,mouseProtein)

GlobalScore = 634.3333

GlobalAlignment = 3×812 char array

'SCRRPAQSAARSRSLRSRPEVKGQGVGPPGVAGAEPPLVT*FADKSRGRRSPDQGLTWPAPSERGDQRAMTSSRLWFSLLLAAAFAGRATAL

3-34
 Compare Sequences Using Sequence Alignment Algorithms

' || |: | | | | | ||:: ||| |||||||:| ||||

'--------AA------------GR--------G---------A----G-R-------W----------AMAGCRLWVSLLLAAALACLATAL

You can also visualize the alignment in the Sequence Alignment app. The alignment is very good
between amino acid position 69 and 599, after which the two sequences appear to be unrelated.
Notice that there is a stop (*) in the sequence at this point. If you shorten the sequences to include
only the amino acids that are in the protein you might get a better alignment. Include the amino acid
positions from the first methionine (M) to the first stop (*) that occurs after the first methionine.

seqalignviewer(GlobalAlignment);

Trim the sequence from the first start amino acid (usually M) to the first stop (*) and then try
alignment again. Find the indices for the stops in the sequences.

humanStops = find(humanProtein == '*')

humanStops = 1×6

41 599 611 713 722 730

mouseStops = find(mouseProtein == '*')

3-35
3 Sequence Analysis

mouseStops = 1×4

539 557 574 606

Looking at the amino acid sequence for humanProtein, the first M is at position 70, and the first stop
after that position is actually the second stop in the sequence (position 599). Looking at the amino
acid sequence for mouseProtein, the first M is at position 11, and the first stop after that position is
the first stop in the sequence (position 557).

Truncate the sequences to include only amino acids in the protein and the stop.
humanProteinORF = humanProtein(70:humanStops(2))

humanProteinORF =
'MTSSRLWFSLLLAAAFAGRATALWPWPQNFQTSDQRYVLYPNNFQFQYDVSSAAQPGCSVLDEAFQRYRDLLFGSGSWPRPYLTGKRHTLEKNVLV

mouseProteinORF = mouseProtein(11:mouseStops(1))

mouseProteinORF =
'MAGCRLWVSLLLAAALACLATALWPWPQYIQTYHRRYTLYPNNFQFRYHVSSAAQAGCVVLDEAFRRYRNLLFGSGSWPRPSFSNKQQTLGKNILV

Globally align the trimmed amino acid sequences.

[GlobalScore_trim, GlobalAlignment_trim] = nwalign(humanProteinORF,mouseProteinORF);
seqalignviewer(GlobalAlignment_trim);

3-36
 Compare Sequences Using Sequence Alignment Algorithms

Another way to truncate an amino acid sequence to only those amino acids in the protein is to first
truncate the nucleotide sequence with indices from the seqshoworfs function. Remember that the
ORF for the human HEXA gene and the ORF for the mouse HEXA were both on the first reading
frame.

humanORFs = seqshoworfs(humanHEXA.Sequence)

humanORFs=1×3 struct array with fields:

Start
Stop

mouseORFs = seqshoworfs(mouseHEXA.Sequence)

3-37
3 Sequence Analysis

mouseORFs=1×3 struct array with fields:

Start
Stop

humanPORF = nt2aa(humanHEXA.Sequence(humanORFs(1).Start(1):humanORFs(1).Stop(1)));
mousePORF = nt2aa(mouseHEXA.Sequence(mouseORFs(1).Start(1):mouseORFs(1).Stop(1)));
[GlobalScore2, GlobalAlignment2] = nwalign(humanPORF, mousePORF);
seqalignviewer(GlobalAlignment2);

3-38
 Compare Sequences Using Sequence Alignment Algorithms

The result from first truncating a nucleotide sequence before converting it to an amino acid sequence
is the same as the result from truncating the amino acid sequence after conversion. An alternative
method to working with subsequences is to use a local alignment function with the nontruncated
sequences.

Locally align the two amino acid sequences using a Smith-Waterman algorithm.

[LocalScore, LocalAlignment] = swalign(humanProtein,mouseProtein);

seqalignviewer(LocalAlignment);

3-39
3 Sequence Analysis

close all;

See Also
swalign | nwalign

3-40
 View and Align Multiple Sequences

View and Align Multiple Sequences

In this section...
“Overview of the Sequence Alignment App” on page 3-41
“Visualize Multiple Sequence Alignment” on page 3-41
“Adjust Sequence Alignments Manually” on page 3-42
“Rearrange Rows” on page 3-50
“Generate Phylogenetic Tree from Aligned Sequences” on page 3-52

Overview of the Sequence Alignment App

The Sequence Alignment app integrates many sequence and multiple alignment functions in the
toolbox. Instead of entering commands in the MATLAB Command Window, you can use this app to
visually inspect a multiple alignment and make manual adjustments.

Visualize Multiple Sequence Alignment

1 Read a multiple sequence alignment file of the gag polyprotein for several HIV strains.

gagaa = multialignread('aagag.aln')
2 View the aligned sequences in the Sequence Alignment app.

seqalignviewer(gagaa);

3-41
3 Sequence Analysis

Adjust Sequence Alignments Manually

Algorithms for aligning multiple sequences do not always produce an optimal result. By visually
inspecting the alignment, you can identify areas whose alignment can be improved by a manual
adjustment.

1 To better visualize the sequence alignments, you can zoom in by selecting Display > Zoom in.
Select this option multiple times until you achieve the zoom level you want.
2 Identify an area where you could improve the alignment.

3-42
 View and Align Multiple Sequences

3 Click a letter or a region. The selected region is the center block. You can then drag the
sequence(s) to the left or right of the center block.

3-43
3 Sequence Analysis

4 To move a single letter (T in this example), click and drag the letter T (center block) to the right
to insert a gap.

3-44
 View and Align Multiple Sequences

5 Close the gap by dragging the letter back to the left.

3-45
3 Sequence Analysis

6 You can also move multiple residues (a subsequence). Suppose you want to move a subsequence
to available gaps. First select the gap region that you want to fill in.

3-46
 View and Align Multiple Sequences

7 Drag the subsequence(s) from the right or left of the gap region into the gap area.

3-47
3 Sequence Analysis

8 Suppose you want to remove one or more of the aligned sequences. First select the sequence(s)
to be removed. Then select Edit > Delete Sequences.

3-48
 View and Align Multiple Sequences

9 Remove empty columns by selecting Edit > Remove Empty Columns.

3-49
3 Sequence Analysis

10 After the edit, you can export the aligned sequences or consensus sequence to a FASTA file or
MATLAB Workspace from the File menu.

Rearrange Rows
You can move the rows (sequences) up or down by one row. You can also move selected rows to the
top or bottom of the list.

3-50
 View and Align Multiple Sequences

The selected sequence moves to the bottom of the list.

3-51
3 Sequence Analysis

Generate Phylogenetic Tree from Aligned Sequences

You can generate a phylogenetic tree using the aligned sequences from within the app. You can select
a subset of sequences or use all the sequences to generate a tree.

Select Display > View Tree > Selected... to generate a tree from selected sequences.

3-52
 View and Align Multiple Sequences

A phylogenetic tree for the sequences is displayed in the Phylogenetic Tree app. For details on the
app, see “Using the Phylogenetic Tree App” on page 5-2.

3-53
3 Sequence Analysis

See Also
seqalignviewer | Sequence Alignment | Sequence Viewer | Genomics Viewer

More About
• “Sequence Alignments” on page 1-7
• “Aligning Pairs of Sequences” on page 3-177

3-54
 Analyzing Synonymous and Nonsynonymous Substitution Rates

Analyzing Synonymous and Nonsynonymous Substitution

Rates

This example shows how the analysis of synonymous and nonsynonymous mutations at the nucleotide
level can suggest patterns of molecular adaptation in the genome of HIV-1. This example is based on
the discussion of natural selection at the molecular level presented in Chapter 6 of "Introduction to
Computational Genomics. A Case Studies Approach" [1].

Introduction

The human immunodeficiency virus 1 (HIV-1) is the more geographically widespread of the two viral
strains that cause Acquired Immunodeficiency Syndrome (AIDS) in humans. Because the virus rapidly
and constantly evolves, at the moment there is no cure nor vaccine against HIV infection. The HIV
virus presents a very high mutation rate that allows it to evade the response of our immune system as
well as the action of specific drugs. At the same time, however, the rapid evolution of HIV provides a
powerful mechanism that reveals important insights into its function and resistance to drugs. By
estimating the force of selective pressures (positive and purifying selections) across various regions
of the viral genome, we can gain a general understanding of how the virus evolves. In particular, we
can determine which genes evolve in response to the action of the targeted immune system and
which genes are conserved because they are involved in some of the virus essential functions.

Nonsynonymous mutations to a DNA sequence cause a change in the translated amino acid sequence,
whereas synonymous mutations do not. The comparison between the number of nonsynonymous
mutations (dn or Ka), and the number of synonymous mutations (ds or Ks), can suggest whether, at
the molecular level, natural selection is acting to promote the fixation of advantageous mutations
(positive selection) or to remove deleterious mutations (purifying selection). In general, when positive
selection dominates, the Ka/Ks ratio is greater than 1; in this case, diversity at the amino acid level is
favored, likely due to the fitness advantage provided by the mutations. Conversely, when negative
selection dominates, the Ka/Ks ratio is less than 1; in this case, most amino acid changes are
deleterious and, therefore, are selected against. When the positive and negative selection forces
balance each other, the Ka/Ks ratio is close to 1.

Extracting Sequence Information for Two HIV-1 Genomes

Download two genomic sequences of HIV-1 (GenBank® accession numbers AF033819 and M27323).
For each encoded gene we extract relevant information, such as nucleotide sequence, translated
sequence and gene product name.

hiv1(1) = getgenbank('AF033819');
hiv1(2) = getgenbank('M27323');

For your convenience, previously downloaded sequences are included in a MAT-file. Note that data in
public repositories is frequently curated and updated; therefore the results of this example might be
slightly different when you use up-to-date datasets.

load hiv1.mat

Extract the gene sequence information using the featureparse function.

genes1 = featureparse(hiv1(1),'feature','CDS','Sequence','true');
genes2 = featureparse(hiv1(2),'feature','CDS','Sequence','true');

3-55
3 Sequence Analysis

Calculating the Ka/Ks Ratio for HIV-1 Genes

Align the corresponding protein sequences in the two HIV-1 genomes and use the resulting alignment
as a guide to insert the appropriate gaps in the nucleotide sequences. Then calculate the Ka/Ks ratio
for each individual gene and plot the results.

KaKs = zeros(1,numel(genes1));
for iCDS = 1:numel(genes1)
% align aa sequences of corresponding genes
[score,alignment] = nwalign(genes1(iCDS).translation,genes2(iCDS).translation);
seq1 = seqinsertgaps(genes1(iCDS).Sequence,alignment(1,:));
seq2 = seqinsertgaps(genes2(iCDS).Sequence,alignment(3,:));

% Calculate synonymous and nonsynonymous substitution rates

[dn,ds] = dnds(seq1,seq2);
KaKs(iCDS) = dn/ds;
end

% plot Ka/Ks ratio for each gene

bar(KaKs);
ylabel('Ka / Ks')
xlabel('genes')
ax = gca;
ax.XTickLabel = {genes1.product};
% plot dotted line at threshold 1
hold on
line([0 numel(KaKs)+1],[1 1],'LineStyle', ':');
KaKs

KaKs =

Columns 1 through 7

0.2560 0.1359 0.3013 0.1128 1.1686 0.4179 0.5150

Columns 8 through 9

0.5115 0.3338

3-56
 Analyzing Synonymous and Nonsynonymous Substitution Rates

All the considered genes, with the exception of TAT, have a total Ka/Ks less than 1. This is in
accordance with the fact that most protein-coding genes are considered to be under the effect of
purifying selection. Indeed, the majority of observed mutations are synonymous and do not affect the
integrity of the encoded proteins. As a result, the number of synonymous mutations generally exceeds
the number of nonsynonymous mutations. The case of TAT represents a well known exception; at the
amino acid level, the TAT protein is one of the least conserved among the viral proteins.

Calculating the Ka/Ks Ratio Using Sliding Windows

Oftentimes, different regions of a single gene can be exposed to different selective pressures. In these
cases, calculating Ka/Ks over the entire length of the gene does not provide a detailed picture of the
evolutionary constraints associated with the gene. For example, the total Ka/Ks associated with the
ENV gene is 0.5155. However, the ENV gene encodes for the envelope glycoprotein GP160, which in
turn is the precursor of two proteins: GP120 (residues 31-511 in AF033819) and GP41 (residues
512-856 in AF033819). GP120 is exposed on the surface of the viral envelope and performs the first
step of HIV infection; GP41 is non-covalently bonded to GP120 and is involved in the second step of
HIV infection. Thus, we can expect these two proteins to respond to different selective pressures, and
a global analysis on the entire ENV gene can obscure diversified behavior. For this reason, we
conduct a finer analysis by using sliding windows of different sizes.

Align ENV genes of the two genomes and measure the Ka/Ks ratio over sliding windows of size equal
to 5, 45, and 200 codons.
env = 8; % ORF number corresponding to gene ENV

% align the two ENV genes

3-57
3 Sequence Analysis

[score,alignment] = nwalign(genes1(env).translation,genes2(env).translation);
env_1 = seqinsertgaps(genes1(env).Sequence,alignment(1,:));
env_2 = seqinsertgaps(genes2(env).Sequence,alignment(3,:));

% compute Ka/Ks using sliding windows of different sizes

[dn1, ds1, vardn1, vards1] = dnds(env_1, env_2, 'window', 200);
[dn2, ds2, vardn2, vards2] = dnds(env_1, env_2, 'window', 45);
[dn3, ds3, vardn3, vards3] = dnds(env_1, env_2, 'window', 5);

% plot the Ka/Ks trends for the different window sizes

figure()
hold on
plot(dn1./ds1, 'r');
plot(dn2./ds2, 'b');
plot(dn3./ds3, 'g');
line([0 numel(dn3)],[1 1],'LineStyle',':');
legend('window size = 200', 'window size = 45', 'window size = 5');
ylim([0 10])
ylabel('Ka / Ks')
xlabel('sliding window (starting codon)')
title 'Env';

The choice of the sliding window size can be problematic: windows that are too long (in this example,
200 codons) average across long regions of a single gene, thus hiding segments where Ka/Ks is
potentially behaving in a peculiar manner. Too short windows (in this example, 5 codons) are likely to
produce results that are very noisy and therefore not very meaningful. In the case of the ENV gene, a
sliding window of 45 codons seems to be appropriate. In the plot, although the general trend is below

3-58
 Analyzing Synonymous and Nonsynonymous Substitution Rates

the threshold of 1, we observe several peaks over the threshold of 1. These regions appear to
undergo positive selection that favors amino acid diversity, as it provides some fitness advantage.

Using Sliding Window Analyses for GAG, POL and ENV Genes

You can perform similar analyses on other genes that display a global Ka/Ks ratio less than 1.
Compute the global Ka/Ks ratio for the GAG, POL and ENV genes. Then repeat the calculation using a
sliding window.

gene_index = [1;2;8]; % ORF corresponding to the GAG, POL, ENV genes

windowSize = 45;

% display the global Ka/Ks for the GAG, POL and ENV genes
KaKs(gene_index)

for i = 1:numel(gene_index)
ID = gene_index(i);
[score,alignment] = nwalign(genes1(ID).translation,genes2(ID).translation);
s1 = seqinsertgaps(genes1(ID).Sequence,alignment(1,:));
s2 = seqinsertgaps(genes2(ID).Sequence,alignment(3,:));

% plot Ka/Ks ratio obtained with the sliding window

[dn, ds, vardn, vards] = dnds(s1, s2, 'window', windowSize);
figure()
plot(dn./ds, 'b')
line([0 numel(dn)],[1 1], 'LineStyle', ':')
ylabel('Ka / Ks')
xlabel('sliding window (starting codon)')
title(genes1(ID).product);
end

ans =

0.2560 0.1359 0.5115

3-59
3 Sequence Analysis

3-60
Analyzing Synonymous and Nonsynonymous Substitution Rates

3-61
3 Sequence Analysis

The GAG (Group-specific Antigen) gene provides the basic physical infrastructure of the virus. It
codes for p24 (the viral capsid), p6 and p7 (the nucleocapsid proteins), and p17 (a matrix protein).
Since this gene encodes for many fundamental proteins that are structurally important for the
survival of the virus, the number of synonymous mutations exceeds the number of nonsynonymous
mutations (i.e., Ka/Ks <1). Thus, this protein is expected to be constrained by purifying selection to
maintain viral infectivity.

The POL gene codes for viral enzymes, such as reverse transcriptase, integrase, and protease. These
enzymes are essential to the virus survival and, therefore, the selective pressure to preserve their
function and structural integrity is quite high. Consequently, this gene appears to be under purifying
selection and we observe Ka/Ks ratio values less than 1 for the majority of the gene length.

The ENV gene codes for the precursor to GP120 and GP41, proteins embedded in the viral envelope,
which enable the virus to attach to and fuse with target cells. GP120 infects any target cell by binding
to the CD4 receptor. As a consequence, GP120 has to maintain the mechanism of recognition of the
host cell and at the same time avoid the detection by the immune system. These two roles are carried
out by different parts of the protein, as shown by the trend in the Ka/Ks ratio. This viral protein is
undergoing purifying (Ka/Ks < 1) and positive selection (Ka/Ks >1) in different regions. A similar
trend is observed in GP41.

Analyzing the Ka/Ks Ratio and Epitopes in GP120

The glycoprotein GP120 binds to the CD4 receptor of any target cell, particularly the helper T-cell.
This represents the first step of HIV infection and, therefore, GP120 was among the first proteins
studied with the intent of finding a HIV vaccine. It is interesting to determine which regions of GP120

3-62
 Analyzing Synonymous and Nonsynonymous Substitution Rates

appear to undergo purifying selection, as indicators of protein regions that are functionally or
structurally important for the virus survival, and could potentially represent drug targets.

From ENV genes, extract the sequences coding for GP120. Compute the Ka/Ks over sliding window of
size equal to 45 codons. Plot and overlap the trend of Ka/Ks with the location of four T cell epitopes
for GP120.

% GP120 protein boundaries in genome1 and genome2 respectively

gp120_start = [31; 30]; % protein boundaries
gp120_stop = [511; 501];
gp120_startnt = gp120_start*3-2; % nt boundaries
gp120_stopnt = gp120_stop*3;

% align GP120 proteins and insert appropriate gaps in nt sequence

[score,alignment] = nwalign(genes1(env).translation(gp120_start(1):gp120_stop(1)), ...
genes2(env).translation(gp120_start(2):gp120_stop(2)));
gp120_1 = seqinsertgaps(genes1(env).Sequence(gp120_startnt(1):gp120_stopnt(1)),alignment(1,:));
gp120_2 = seqinsertgaps(genes2(env).Sequence(gp120_startnt(2):gp120_stopnt(2)),alignment(3,:));

% Compute and plot Ka/Ks ratio using the sliding window

[dn120, ds120, vardn120, vards120] = dnds(gp120_1, gp120_2, 'window', windowSize);

% Epitopes for GP120 identified by cellular methods (see reference [2])

epitopes = {'TVYYGVPVWK','HEDIISLWQSLKPCVKLTPL',...
'EVVIRSANFTNDAKATIIVQLNQSVEINCT','QIASKLREQFGNNK',...
'QSSGGDPEIVTHSFNCGGEFF','KQFINMWQEVGKAMYAPP',...
'DMRDNWRSELYKYKVVKIEPLGVAP'};

% Find location of the epitopes in the aligned sequences:

epiLoc = zeros(numel(epitopes),2);
for i = 1:numel(epitopes)
[sco,ali,ind] = swalign(alignment(1,:),epitopes{i});
epiLoc(i,:) = ind(1) + [0 length(ali)-1];
end

figure
hold on
% plot Ka/Ks relatively to the middle codon of the sliding window
plot(windowSize/2+(1:numel(dn120)),dn120./ds120)
plot(epiLoc,[1 1],'linewidth',5)
line([0 numel(dn120)+windowSize/2],[1 1],'LineStyle',':')
title('GP120, Ka / Ks and epitopes');
ylabel('Ka / Ks');
xlabel('sliding window (middle codon)');

3-63
3 Sequence Analysis

Although the general trend of the Ka/Ks ratio is less than 1, there are some regions where the ratio is
greater than one, indicating that these regions are likely to be under positive selection. Interestingly,
the location of some of these regions corresponds to the presence of T cell epitopes, identified by
cellular methods. These segments display high amino acid variability because amino acid diversity in
these regions allows the virus to evade the host immune system recognition. Thus, we can conclude
that the source of variability in this regions is likely to be the host immune response.

References

[1] Cristianini, N. and Hahn, M.W., "Introduction to Computational Genomics: A Case Studies
Approach", Cambridge University Press, 2007.

[2] Siebert, S.A., et al., "Natural Selection on the gag, pol, and env Genes of Human
Immunodeficiency Virus 1 (HIV-1)", Molecular Biology and Evolution, 12(5):803-813, 1995.

3-64
 Investigating the Bird Flu Virus

Investigating the Bird Flu Virus

This example shows how to calculate Ka/Ks ratios for eight genes in the H5N1 and H2N3 virus
genomes, and perform a phylogenetic analysis on the HA gene from H5N1 virus isolated from
chickens across Africa and Asia. For the phylogenetic analysis, you will reconstruct a neighbor-joining
tree and create a 3-D plot of sequence distances using multidimensional scaling. Finally, you will map
the geographic locations where each HA sequence was found on a regional map. Sequences used in
this example were selected from the bird flu case study on the Computational Genomics Website [1].
Note: The final section in this example requires the Mapping Toolbox™.

Introduction

There are three types of influenza virus: Type A, B and C. All influenza genomes are comprised of
eight segments or genes that code for polymerase B2 (PB2), polymerase B1 (PB1), polymerase A (PA),
hemagglutinin (HA), nucleoprotein (NP), neuraminidase (NA), matrix (M1), and non-structural (NS1)
proteins. Note: Type C virus has hemagglutinin-esterase (HE), a homolog to HA.

Of the three types of influenza, Type A has the potential to be the most devastating. It affects birds
(its natural reservoir), humans and other mammals and has been the major cause of global influenza
epidemics. Type B affects only humans causing local epidemics, and Type C does not tend to cause
serious illness.

Type A influenzas are further classified into different subtypes according to variations in the amino
acid sequences of HA (H1-16) and NA (N1-9) proteins. Both proteins are located on the outside of the
virus. HA attaches the virus to the host cell then aids in the process of the virus being fused in to the
cell. NA clips the newly created virus from the host cell so it can move on to a healthy new cell.
Difference in amino acid composition within a protein and recombination of the various HA and NA
proteins contribute to Type A influenzas' ability to jump host species (i.e. bird to humans) and wide
range of severity. Many new drugs are being designed to target HA and NA proteins [2,3,4].

In 1997, H5N1 subtype of the avian influenza virus, a Type A influenza virus, made an unexpected
jump to humans in Hong Kong causing the deaths of six people. To control the rapidly spreading
disease, all poultry in Hong Kong was destroyed. Sequence analysis of the H5N1 virus is shown here
[2,4].

Calculating Ka/Ks Ratio For Each H5N1 Gene

An investigation of the Ka/Ks ratios for each gene segment of the H5N1 virus will provide some
insight into how each is changing over time. Ka/Ks is the ratio of non-synonymous changes to
synonymous in a sequence. For a more detailed explanation of Ka/Ks ratios, see “Analyzing
Synonymous and Nonsynonymous Substitution Rates” on page 3-55. To calculate Ka/Ks, you need a
copy of the gene from two time points. You can use H5N1 virus isolated from chickens in Hong Kong
in 1997 and 2001. For comparison, you can include H2N3 virus isolated from mallard ducks in
Alberta in 1977 and 1985 [1].

For the purpose of this example, sequence data is provided in four MATLAB® structures that were
created by genbankread.

Load H5N1 and H2N3 sequence data.

load('birdflu.mat','chicken1997','chicken2001','mallard1977','mallard1985')

3-65
3 Sequence Analysis

Data in public repositories is frequently curated and updated. You can retrieve the up-to-date
datasets by using the getgenbank function. Note that if data has indeed changed, the results of this
example might be slightly different when you use up-to-date datasets.

chicken1997 = arrayfun(@(x)getgenbank(x{:}),{chicken1997.Accession});
chicken2001 = arrayfun(@(x)getgenbank(x{:}),{chicken2001.Accession});
mallard1977 = arrayfun(@(x)getgenbank(x{:}),{mallard1977.Accession});
mallard1985 = arrayfun(@(x)getgenbank(x{:}),{mallard1985.Accession});

You can extract just the coding portion of the nucleotide sequences using the featureparse
function. The featureparse function returns a structure with fields containing information from the
Features section in a GenBank file including with a Sequence field that contains just the coding
sequence.

for ii = 1:numel(chicken1997)
ntSeq97{ii} = featureparse(chicken1997(ii),'feature','cds','sequence',true);
ntSeq01{ii} = featureparse(chicken2001(ii),'feature','cds','sequence',true);
ntSeq77{ii} = featureparse(mallard1977(ii),'feature','cds','sequence',true);
ntSeq85{ii} = featureparse(mallard1985(ii),'feature','cds','sequence',true);
end

ntSeq97{1}

ans =

struct with fields:

Location: '<1..>2273'
Indices: [1 2273]
UnknownFeatureBoundaries: 1
gene: 'PB2'
codon_start: '1'
product: 'PB2 protein'
protein_id: 'AAF02361.1'
db_xref: 'GI:6048850'
translation: 'RIKELRDLMSQSRTREILTKTTVDHMAIIKKYTSGRQEKNPALRMKWMMAMKYPITADKRIMEMIP
Sequence: 'agaataaaagaactaagagatttgatgtcgcaatctcgcacacgcgagatactgacaaaaaccact

Visual inspection of the sequence structures revealed some of the genes have splice variants
represented in the GenBank files. Because this analysis is only on PB2, PB1, PA, HA, NP, NA, M1, and
NS1 genes, you need to remove any splice variants.

Remove splice variants from 1997 H5N1

ntSeq97{7}(1) = [];% M2
ntSeq97{8}(1) = [];% NS2

Remove splice variants from 1977 H2N3

ntSeq77{2}(2) = [];% PB1-F2

ntSeq77{7}(1) = [];% M2
ntSeq77{8}(1) = [];% NS2

Remove splice variants from 1985 H2N3

3-66
 Investigating the Bird Flu Virus

ntSeq85{2}(2) = [];% PB1-F2

ntSeq85{7}(1) = [];% M2
ntSeq85{8}(1) = [];% NS2

You need to align the nucleotide sequences to calculate the Ka/Ks ratio. Align protein sequences for
each gene (available in the 'translation' field) using nwalign function, then insert gaps into
nucleotide sequence using seqinsertgaps. Use the function dnds to calculate non-synonymous and
synonymous substitution rates for each of the eight genes in the virus genomes. If you are interested
in seeing the sequence alignments, set the 'verbose' option to true when using dnds.

Influenza gene names

proteins = {'PB2','PB1','PA','HA','NP','NA','M1','NS1'};

H5N1 Virus

for ii = 1:numel(ntSeq97)
[sc,align] = nwalign(ntSeq97{ii}.translation,ntSeq01{ii}.translation,'alpha','aa');
ch97seq = seqinsertgaps(ntSeq97{ii}.Sequence,align(1,:));
ch01seq = seqinsertgaps(ntSeq01{ii}.Sequence,align(3,:));
[dn,ds] = dnds(ch97seq,ch01seq);
H5N1.(proteins{ii}) = dn/ds;
end

H2N3 Virus

for ii = 1:numel(ntSeq77)
[sc,align] = nwalign(ntSeq77{ii}.translation,ntSeq85{ii}.translation,'alpha','aa');
ch77seq = seqinsertgaps(ntSeq77{ii}.Sequence,align(1,:));
ch85seq = seqinsertgaps(ntSeq85{ii}.Sequence,align(3,:));
[dn,ds] = dnds(ch77seq,ch85seq);
H2N3.(proteins{ii}) = dn/ds;
end
H5N1
H2N3

H5N1 =

struct with fields:

PB2: 0.0226
PB1: 0.0240
PA: 0.0307
HA: 0.0943
NP: 0.0517
NA: 0.1015
M1: 0.0460
NS1: 0.3010

H2N3 =

struct with fields:

PB2: 0.0048
PB1: 0.0021
PA: 0.0089

3-67
3 Sequence Analysis

HA: 0.0395
NP: 0.0071
NA: 0.0559
M1: 0
NS1: 0.1954

Note: Ka/Ks ratio results may vary from those shown on [1] due to sequence splice variants.

Visualize Ka/Ks ratios in 3-D bar graph.

H5N1rates = cellfun(@(x)(H5N1.(x)),proteins);
H2N3rates = cellfun(@(x)(H2N3.(x)),proteins);
bar3([H2N3rates' H5N1rates']);
ax = gca;
ax.XTickLabel = {'H2N3','H5N1'};
ax.YTickLabel = proteins;
zlabel('Ka/Ks');
view(-115,16);
title('Ka/Ks Ratios for H5N1 (Chicken) and H2N3 (Mallard) Viruses');

NS1, HA and NA have larger non-synonymous to synonymous ratios compared to the other genes in
both H5N1 and H2N3. Protein sequence changes to these genes have been attributed to an increase
in H5N1 pathogenicity. In particular, changes to the HA gene may provide the virus the ability to
transfer into others species beside birds [2,3].

3-68
 Investigating the Bird Flu Virus

Performing a Phylogenetic Analysis of the HA Protein

The H5N1 virus attaches to cells in the gastrointestinal tract of birds and the respiratory tract of
humans. Changes to the HA protein, which helps bind the virus to a healthy cell and facilitates its
incorporation into the cell, are what allow the virus to affect different organs in the same and
different species. This may provide it the ability to jump from birds to humans [2,3]. You can perform
a phylogenetic analysis of the HA protein from H5N1 virus isolated from chickens at different times
(years) in different regions of Asia and Africa to investigate their relationship to each other.

Load HA amino acid sequence data from 16 regions/times from the MAT-file provided birdflu.mat
or retrieve the up-to-date sequence data from the NCBI repository using the getgenpept function.

load('birdflu.mat','HA')

HA = arrayfun(@(x)getgenpept(x{:}),{HA.Accession});

Create a new structure array containing fields corresponding to amino acid sequence (Sequence) and
source information (Header). You can extract source information from the HA using featureparse
then parse with regexp.

for ii = 1:numel(HA)
source = featureparse(HA(ii),'feature','source');
strain = regexp(source.strain,'A/[Cc]hicken/(\w+\s*\w*).*/(\d+)','tokens');
proteinHA(ii).Header = sprintf('%s_%s',char(strain{1}(1)),char(strain{1}(2)));
proteinHA(ii).Sequence = HA(ii).Sequence;
end

proteinHA(1)

ans =

struct with fields:

Header: 'Nigeria_2006'
Sequence: 'mekivllfaivslvksdqicigyhannsteqvdtimeknvtvthaqdilekthngklcdldgvkplilrdcsvagwllgnpm

Align the HA amino acid sequences using multialign and visualize the alignment with
seqalignviewer.

alignHA = multialign(proteinHA);
seqalignviewer(alignHA);

3-69
3 Sequence Analysis

Calculate the distances between sequences using seqpdist with the Jukes-Cantor method. Use
seqneighjoin to reconstruct a phylogenetic tree using the neighbor-joining method.
Seqneighjoin returns a phytree object.

distHA = seqpdist(alignHA,'method','Jukes-Cantor','alpha','aa');
HA_NJtree = seqneighjoin(distHA,'equivar',alignHA);

Use the view method associated with phytree objects to open the tree in the Phylogenetic Tree
Tool.

view(HA_NJtree);

3-70
 Investigating the Bird Flu Virus

Visualizing Sequence Distances with Multidimensional Scaling (MDS)

Another way to visualize the relationship between sequences is to use multidimensional scaling
(MDS) with the distances calculated for the phylogenetic tree. This functionality is provided by the
cmdscale function in Statistics and Machine Learning Toolbox™.

[Y,eigvals] = cmdscale(distHA);

You can use the eigenvalues returned by cmdscale to help guide your decision of whether to use the
first two or three dimensions in your plot.

sigVecs = [1:3;eigvals(1:3)';eigvals(1:3)'/max(abs(eigvals))];
report = ['Dimension Eigenvalues Normalized' ...
sprintf('\n %d\t %1.4f %1.4f',sigVecs)];
display(report);

report =

'Dimension Eigenvalues Normalized

1 0.0062 1.0000

3-71
3 Sequence Analysis

2 0.0028 0.4462
3 0.0014 0.2209'

The first two dimensions represent a large portion of the data, but the third still contains information
that might help resolve clusters in the sequence data. You can create a three dimensional scatter plot
using plot3 function.

locations = {proteinHA(:).Header};
figure
plot3(Y(:,1),Y(:,2),Y(:,3),'ok');
text(Y(:,1)+0.002,Y(:,2),Y(:,3)+0.001,locations,'interpreter','no');
title('MDS Plot of HA Sequences');
view(-21,12);

Clusters appear to correspond to groupings in the phylogenetic tree. Find the sequences belonging to
each cluster using the subtree method of phytree. One of subtree's required inputs is the node
number (number of leaves + number of branches), which will be the new subtree's root node. For
your example, the cluster containing Hebei and Hong Kong in the MDS plot is equivalent to the
subtree whose root node is Branch 14, which is Node 30 (16 leaves + 14 branches).

cluster1 = get(subtree(HA_NJtree,30),'LeafNames');
cluster2 = get(subtree(HA_NJtree,21),'LeafNames');
cluster3 = get(subtree(HA_NJtree,19),'LeafNames');

Get an index for the sequences belonging to each cluster.

3-72
 Investigating the Bird Flu Virus

[cl1,cl1_ind] = intersect(locations,cluster1);
[cl2,cl2_ind] = intersect(locations,cluster2);
[cl3,cl3_ind] = intersect(locations,cluster3);
[cl4,cl4_ind] = setdiff(locations,{cl1{:} cl2{:} cl3{:}});

Change the color and marker symbols on the MDS plot to correspond to each cluster.

h = plot3(Y(cl1_ind,1),Y(cl1_ind,2),Y(cl1_ind,3),'^',...
Y(cl2_ind,1),Y(cl2_ind,2),Y(cl2_ind,3),'o',...
Y(cl3_ind,1),Y(cl3_ind,2),Y(cl3_ind,3),'d',...
Y(cl4_ind,1),Y(cl4_ind,2),Y(cl4_ind,3),'v');
numClusters = 4;
col = autumn(numClusters);
for i = 1:numClusters
h(i).MarkerFaceColor = col(i,:);
end
set(h(:),'MarkerEdgeColor','k');
text(Y(:,1)+0.002,Y(:,2),Y(:,3),locations,'interpreter','no');
title('MDS Plot of HA Sequences');
view(-21,12);

For more detailed information on using Ka/Ks ratios, phylogenetics and MDS for sequence analysis
see Cristianini and Hahn [5].

Displaying Geographic Regions of the H5N1 Virus on a Map of Africa and Asia

NOTE: You need Mapping Toolbox to produce the following figure.

3-73
3 Sequence Analysis

Using tools from Mapping Toolbox, you can plot the location where each virus was isolated on a map
of Africa and Asia. To do this, you need the latitude and longitude for each location. For information
on finding geospatial data on the internet, see “Find Geospatial Data Online” (Mapping Toolbox).
Latitude and longitude for the capital city of each geographic region where the viruses were isolated
are provided for this example.

Create a geostruct structure, regionHA, that contains the geographic information for each feature,
or sequence, to be displayed. A geostruct is required to have Geometry, Lat, and Lon fields that
specify the feature type, latitude and longitude. This information is used by mapping functions in
Mapping Toolbox to display geospatial data.
[regionHA(1:16).Geometry] = deal('Point');
[regionHA(:).Lat] = deal(9.10, 34.31, 15.31, 39.00, 39.00, 39.00, 55.26,...
15.56, 34.00, 33.14, 34.20, 23.00, 37.35, 44.00,...
22.11, 22.11);
[regionHA(:).Lon] = deal(7.10, 69.08, 32.35, 116.00, 116.00, 116.00,...
65.18, 105.48, 114.00, 131.36, 131.40, 113.00,...
127.00, 127.00, 114.14, 114.14);

A geostruct can also have attribute fields that contain additional information about each feature. Add
attribute fields Name and Cluster to the regionHA structure. The Cluster field contains the
sequence's cluster number, which you will use to identify the sequences' cluster membership.
[regionHA(:).Name] = deal(proteinHA.Header);

[regionHA(cl1_ind).Cluster] = deal(1);
[regionHA(cl2_ind).Cluster] = deal(2);
[regionHA(cl3_ind).Cluster] = deal(3);
[regionHA(cl4_ind).Cluster] = deal(4);

regionHA(1)

ans =

struct with fields:

Geometry: 'Point'
Lat: 9.1000
Lon: 7.1000
Name: 'Nigeria_2006'
Cluster: 3

Create a structure using the makesymbolspec function, which will contain marker and color
specifications for each marker to be displayed on the map. You will pass this structure to the
geoshow function. Symbol markers and colors are set to correspond with the clusters in MDS plot.
clusterSymbols = makesymbolspec('Point',...
{'Cluster',1,'Marker', '^'},...
{'Cluster',2,'Marker', 'o'},...
{'Cluster',3,'Marker', 'd'},...
{'Cluster',4,'Marker', 'v'},...
{'Cluster',[1 4],'MarkerFaceColor',autumn(4)},...
{'Default','MarkerSize', 6},...
{'Default','MarkerEdgeColor','k'});

Load the mapping information and use the geoshow function to plot virus locations on a map.

3-74
 Investigating the Bird Flu Virus

load coast
load topo
figure
fig = gcf;
fig.Renderer = 'zbuffer';
worldmap([-45 85],[0 160])
setm(gca,'mapprojection','robinson',...
'plabellocation',30,'mlabelparallel',-45,'mlabellocation',30)
plotm(lat, long)
geoshow(topo, topolegend, 'DisplayType', 'texturemap')
demcmap(topo)
brighten(.60)

geoshow(regionHA,'SymbolSpec',clusterSymbols);
title('Geographic Locations of HA Sequence in Africa and Asia')

Viewing Geographic Regions of Interest in Google™ Earth

NOTE: You need Mapping Toolbox to export data to a KML-formatted file.

Using the kmlwrite function from Mapping Toolbox, you can write the location and annotation
information for each sequence to a KML-formatted file. Google Earth displays geographic data from
KML files within its Earth browser. Mapping Toolbox's kmlwrite function translates a geostruct,
such as regionHA, into a KML-formatted file to be used by Google Earth. For more information on
kmlwrite, see “Exporting Vector Data to KML” (Mapping Toolbox).

3-75
3 Sequence Analysis

You can further annotate each sequence with information from the Features section of the GenBank
file using the featureparse function. You can then use this information to populate the geostruct,
regionHA, and display it in table form as a description tag for each placemark in the Google Earth
browser. In a geostruct, mandatory fields are Geometry, Lat and Lon field. All other fields are
considered to be attributes of the placemark.

for i = 1:numel(HA)
feats = featureparse(HA(i),'Feature','source');
regionHA(i).Strain = feats.strain;
if isfield(feats,'country')
regionHA(i).Country = feats.country;
else
regionHA(i).Country = 'N/A';
end
year = regexp(regionHA(i).Name,'\d+','match');
regionHA(i).Year = year{1};
% Create a link to GenPept record through the accession number
regionHA(i).AccessionNumber = ...
['<a href="http://www.ncbi.nlm.nih.gov/sites/entrez?db=Protein&cmd=search&term=',...
HA(i).Accession,'">',HA(i).Accession,'</a>'];
end

[regionHA.SequenceLength] = deal(HA.LocusSequenceLength);

Create an attribute structure using the makeattribspec function, which you will use to format the
description table for each marker. The attribute structure dictates the order and formatting of each
attribute. You can also use it to not display one of the attributes in the geostruct, regionHA.

attribStruct = makeattribspec(regionHA);

Remove the Name field and reorder the fields in the attribute structure.

attribStruct = rmfield(attribStruct,'Name');

attribStruct = orderfields(attribStruct,{'AccessionNumber','Strain',...
'SequenceLength','Country','Year','Cluster'});

regionHA = orderfields(regionHA,{'AccessionNumber','Strain',...
'SequenceLength','Country','Year','Cluster','Geometry','Lon','Lat',...
'Name'});

Reformat attribute labels for display in the table.

attribStruct.AccessionNumber.AttributeLabel = '<b>Accession Number</b>';

attribStruct.Strain.AttributeLabel = '<b>Viral Strain</b>';
attribStruct.SequenceLength.AttributeLabel = '<b>Sequence Length</b>';
attribStruct.Country.AttributeLabel = '<b>Country of Origin</b>';
attribStruct.Year.AttributeLabel = '<b>Year Isolated</b>';
attribStruct.Cluster.AttributeLabel = '<b>Cluster Membership</b>';

Viewing the File in Google Earth.

Write the regionHA geostruct to a KML-formatted file in a temporary directory.

kmlDirectory = tempdir;
filename = fullfile(kmlDirectory,'HA_geographic_locations.kml');
kmlwrite(filename,regionHA,'Description',attribStruct,'Name',{regionHA.Strain},...
'Icon','http://maps.google.com/mapfiles/kml/shapes/arrow.png','iconscale',1.5);

3-76
 Investigating the Bird Flu Virus

You can display a KML file in a Google Earth browser [6]. Google Earth must be installed on the
system. On Windows® platforms, display the KML file with:

winopen(filename)

For Unix and MAC users, display the KML file with:

cmd = 'googleearth ';

fullfilename = fullfile(pwd, filename);
system([cmd fullfilename])

For this example, the KML file was previously displayed using Google Earth Pro. The Google Earth
image was then saved using the Google Earth "File->Save Image" menu. This is how the data in your
KML file looks when loaded into Google Earth. To get this view move around and zoom in on the
region over Asia.

3-77
3 Sequence Analysis

Click a placemark to view information about the sequence. The accession number in each data table
is a hyperlink to the GenPept sequence file in the NCBI Protein Database.

3-78
 Investigating the Bird Flu Virus

Optionally, remove the new KML file from your KML output directory.
delete(filename)

close all

References

[1] https://computationalgenomics.blogs.bristol.ac.uk/case_studies/birdflu_demo

[2] Laver, W.G., Bischofberger, N. and Webster, R.G., "Disarming Flu Viruses", Scientific American,
280(1):78-87, 1999.

[3] Suzuki, Y. and Masatoshi, N., "Origin and Evolution of Influenza Virus Hemagglutinin Genes",
Molecular Biology and Evolution, 19(4):501-9, 2002.

[4] Gambaryan, A., et al., "Evolution of the receptor binding phenotype of influenza A(H5) viruses",
Virology, 344(2):432-8, 2006.

3-79
3 Sequence Analysis

[5] Cristianini, N. and Hahn, M.W., "Introduction to Computational Genomics: A Case Studies
Approach", Cambridge University Press, 2007.

[6] Google Earth images were acquired using Google Earth Pro. For more information about Google
Earth and Google Earth Pro, visit http://earth.google.com/

3-80
 Exploring Primer Design

Exploring Primer Design

This example shows how to use the Bioinformatics Toolbox™ to find potential primers that can be
used for automated DNA sequencing.

Introduction

Primer design for PCR can be a daunting task. A good primer should usually meet the following
criteria:

• Length is 18-30 bases.

• Melting temperature is 50-60 degrees Celsius.
• GC content is between 45% and 55%.
• Does not form stable hairpins.
• Does not self dimerize.
• Does not cross dimerize with other primers in the reaction.
• Has a GC clamp at the 3' end of the primer.

This example uses MATLAB® and Bioinformatics Toolbox to find PCR primers for the human
hexosaminidase gene.

First load the hexosaminidase nucleotide sequence from the provided MAT-file
hexosaminidase.mat. The DNA sequence that you want to find primers for is in the Sequence field
of the loaded structure.
load('hexosaminidase.mat','humanHEXA')
sequence = humanHEXA.Sequence;

You can also use the getgenbank function to retrieve the sequence information from the NCBI data
repository and load it into MATLAB. The NCBI reference sequence for HEXA has accession number
NM_000520.
humanHEXA = getgenbank('NM_000520');

Calculating Properties of an Oligonucleotide

The oligoprop function is a useful tool to get properties of oligonucleotide DNA sequences. This
function calculates the GC content, molecular weight, melting temperature, and secondary structure
information. oligoprop returns a structure that has fields with the associated information. Use the
help command to see what other properties oligoprop returns.
nt = oligoprop('AAGCTCAAAAACGCGCGGTATTCGACTGGCGTGATCTATTTTATGCT')

nt =

struct with fields:

GC: 44.6809
GCdelta: 0
Hairpins: [3x47 char]
Dimers: [9x47 char]
MolWeight: 1.4468e+04

3-81
3 Sequence Analysis

MolWeightdelta: 0
Tm: [68.9128 79.7752 85.3393 69.6497 68.2474 75.8931]
Tmdelta: [0 0 0 0 0 0]
Thermo: [4x3 double]
Thermodelta: [4x3 double]

Finding All Potential Forward Primers

The goal is to create a list of all potential forward primers of length 20. You can do this either by
iterating over the entire sequence and taking subsequences at every possible position or by using a
matrix of indices. The following example shows how you can set a matrix of indices and then use it to
create all possible forward subsequences of length 20, in this case N-19 subsequences where N is the
length of the target hexosaminidase sequence. Then you can use the oligoprop function to get
properties for each of the potential primers in the list.
N = length(sequence) % length of the target sequence
M = 20 % desired primer length
index = repmat((0:N-M)',1,M)+repmat(1:M,N-M+1,1);
fwdprimerlist = sequence(index);

for i = N-19:-1:1 % reverse order to pre-allocate structure

fwdprimerprops(i) = oligoprop(fwdprimerlist(i,:));
end

N =

2437

M =

20

Finding All Potential Reverse Primers

After you have all potential forward primers, you can search for reverse primers in a similar fashion.
Reverse primers are found on the complementary strand. To obtain the complementary strand use the
seqcomplement function.
comp_sequence = seqcomplement(sequence);
revprimerlist = seqreverse(comp_sequence(index));

for i = N-19:-1:1 % reverse order to preallocate structure

revprimerprops(i) = oligoprop(revprimerlist(i,:));
end

Filtering Primers Based on GC Content

The GC content information for the primers is in a structure with the field GC. To eliminate all
potential primers that do not meet the criteria stated above (a GC content of 45% to 55%), you can
make a logical indexing vector that indicates which primers have GC content outside the acceptable
range. Extract the GC field from the structure and convert it to a numeric vector.
fwdgc = [fwdprimerprops.GC]';
revgc = [revprimerprops.GC]';

3-82
 Exploring Primer Design

bad_fwdprimers_gc = fwdgc < 45 | fwdgc > 55;

bad_revprimers_gc = revgc < 45 | revgc > 55;

Filtering Primers Based on Their Melting Temperature

The melting temperature is significant when you are designing PCR protocols. Create another logical
indexing vector to keep track of primers with bad melting temperatures. The melting temperatures
from oligoprop are estimated in a variety of ways (basic, salt-adjusted, nearest-neighbor). The
following example uses the nearest-neighbor estimates for melting temperatures with parameters
established by SantaLucia, Jr. [1]. These are stored in the fifth element of the field Tm returned by
oligoprop. The other elements of this field represent other methods to estimate the melting
temperature. You can also use the mean function to compute an average over all the estimates.

fwdtm = cell2mat({fwdprimerprops.Tm}');
revtm = cell2mat({revprimerprops.Tm}');
bad_fwdprimers_tm = fwdtm(:,5) < 50 | fwdtm(:,5) > 60;
bad_revprimers_tm = revtm(:,5) < 50 | revtm(:,5) > 60;

Finding Primers With Self-Dimerization and Hairpin Formation

Self-dimerization and hairpin formation can prevent the primer from binding to the target sequence.
As above, you can create logical indexing vectors to indicate whether the potential primers do or do
not form self-dimers or hairpins.

bad_fwdprimers_dimers = ~cellfun('isempty',{fwdprimerprops.Dimers}');
bad_fwdprimers_hairpin = ~cellfun('isempty',{fwdprimerprops.Hairpins}');
bad_revprimers_dimers = ~cellfun('isempty',{revprimerprops.Dimers}');
bad_revprimers_hairpin = ~cellfun('isempty',{revprimerprops.Hairpins}');

Finding Primers Without a GC Clamp

A strong base pairing at the 3' end of the primer is helpful. Find all the primers that do not end in a G
or C. Remember that all the sequences in the lists are 5'->3'.

bad_fwdprimers_clamp = lower(fwdprimerlist(:,end)) == 'a' | lower(fwdprimerlist(:,end)) == 't';

bad_revprimers_clamp = lower(revprimerlist(:,end)) == 'a' | lower(revprimerlist(:,end)) == 't';

Finding Primers With Nucleotide Repeats

Primers that have stretches of repeated nucleotides can give poor PCR results. These are sequences
with low complexity. To eliminate primers with stretches of four or more repeated bases, use the
function regexp.

fwdrepeats = regexpi(cellstr(fwdprimerlist),'a{4,}|c{4,}|g{4,}|t{4,}','ONCE');
revrepeats = regexpi(cellstr(revprimerlist),'a{4,}|c{4,}|g{4,}|t{4,}','ONCE');
bad_fwdprimers_repeats = ~cellfun('isempty',fwdrepeats);
bad_revprimers_repeats = ~cellfun('isempty',revrepeats);

Find the Primers That Satisfy All the Criteria

The rows of the original list of subsequences correspond to the base number where each
subsequence starts. You can use the logical indexing vectors collected so far and create a new list of
primers that satisfy all the criteria discussed above. The figure shows how the forward primers have
been filtered, where values equal to 1 indicates bad primers and values equal to 0 indicates good
primers.

3-83
3 Sequence Analysis

bad_fwdprimers = [bad_fwdprimers_gc, bad_fwdprimers_tm,...

bad_fwdprimers_dimers, bad_fwdprimers_hairpin,...
bad_fwdprimers_clamp, bad_fwdprimers_repeats];
bad_revprimers = [bad_revprimers_gc, bad_revprimers_tm,...
bad_revprimers_dimers, bad_revprimers_hairpin,...
bad_revprimers_clamp, bad_revprimers_repeats];

good_fwdpos = find(all(~bad_fwdprimers,2));
good_fwdprimers = fwdprimerlist(good_fwdpos,:);
good_fwdprop = fwdprimerprops(good_fwdpos);
N_good_fwdprimers = numel(good_fwdprop)

good_revpos = find(all(~bad_revprimers,2));
good_revprimers = revprimerlist(good_revpos,:);
good_revprop = revprimerprops(good_revpos);
N_good_revprimers = numel(good_revprop)

figure
imagesc([bad_fwdprimers any(bad_fwdprimers,2)]);
title('Filtering candidate forward primers');
ylabel('Primer location');
xlabel('Criteria');
ax = gca;
ax.XTickLabel = char({'%GC','Tm','Dimers','Hairpin','GC clamp','Repeats','All'});
ax.XTickLabelRotation = 45;
colorbar

N_good_fwdprimers =

140

N_good_revprimers =

147

3-84
 Exploring Primer Design

Checking For Cross Dimerization

Cross dimerization can occur between the forward and reverse primer if they have a significant
amount of complementarity. The primers will not function properly if they dimerize with each other.
To check for dimerization, align every forward primer against every reverse primer, using the
swalign function, and keep the low-scoring pairs of primers. This information can be stored in a
matrix with rows representing forward primers and columns representing reverse primers. This
exhaustive calculation can be quite time-consuming. However, there is no point in performing this
calculation on primer pairs where the reverse primer is upstream of the forward primer. Therefore,
these primer pairs can be ignored. The image in the figure shows the pairwise scores before being
thresholded, low scores (dark blue) represent primer pairs that do not dimerize.
scr_mat = [-1,-1,-1,1;-1,-1,1,-1;-1,1,-1,-1;1,-1,-1,-1;];
scr = zeros(N_good_fwdprimers,N_good_revprimers);
for i = 1:N_good_fwdprimers
for j = 1:N_good_revprimers
if good_fwdpos(i) < good_revpos(j)
scr(i,j) = swalign(good_fwdprimers(i,:), good_revprimers(j,:), ...
'SCORINGMATRIX',scr_mat,'GAPOPEN',5,'ALPHA','NT');
else
scr(i,j) = 13; % give a high score to ignore forward primers
% that are after reverse primers
end
end
end

figure

3-85
3 Sequence Analysis

imagesc(scr)
title('Cross dimerization scores')
xlabel('Candidate reverse primers')
ylabel('Candidate forward primers')
colorbar

Low scoring primer pairs are identified as logical one in an indicator matrix.

pairedprimers = scr<=3;

Visualizing Potential Pairs of Primers in the Sequence Domain

An alternative way to present this information is to look at all potential combinations of primers in the
sequence domain. Each dot in the plot represents a possible combination between the forward and
reverse primers after filtering out all those cases with potential cross dimerization.

[f,r] = find(pairedprimers);
figure
plot(good_revpos(r),good_fwdpos(f),'r.','markersize',10)
axis([1 N 1 N])
title('Primer selection graph')
xlabel('Reverse primer positions')
ylabel('Forward primer positions')

3-86
 Exploring Primer Design

Selecting a Primer Pair to Amplify a Specific Region

You can use the information calculated so far to find the best primer pairs that allow amplification of
the 220bp region from position 880 to 1100. First, you find all pairs that can cover the required
region, taking into account the length of the primer. Then, you calculate the Euclidean distance of the
actual positions to the desired ones, and re-order the list starting with the closest distance.

pairs = find(good_fwdpos(f)<(880-M) & good_revpos(r)>1100);

dist = (good_fwdpos(f(pairs))-(880-M)).^2 + (good_revpos(r(pairs))-(1100)).^2;
[dist,h] = sort(dist);
pairs = pairs(h);

hold on
plot(good_revpos(r(pairs)),good_fwdpos(f(pairs)),'b.','markersize',10)
plot([1100 1100],[1 880-M],'g')
plot([1100 N],[880-M 880-M],'g')

3-87
3 Sequence Analysis

Retrieve Primer Pairs

Use the sprintf function to generate a report with the ten best pairs and associated information.
These primer pairs can then be verified experimentally. These primers can also be 'BLASTed' using
the blastncbi function to check specificity.
Primers = sprintf('Fwd/Rev Primers Start End %%GC mT Length\n\n');
for i = 1:10
fwd = f(pairs(i));
rev = r(pairs(i));
Primers = sprintf('%s%-21s%-6d%-6d%-4.4g%-4.4g\n%-21s%-6d%-6d%-4.4g%-7.4g%-6d\n\n', ...
Primers, good_fwdprimers(fwd,:),good_fwdpos(fwd),good_fwdpos(fwd)+M-1,good_fwdprop(fwd).GC,go
good_revprimers(rev,:),good_revpos(rev)+M-1,good_revpos(rev),good_revprop(rev).GC,go
good_revpos(rev) - good_fwdpos(fwd) );
end
disp(Primers)

Fwd/Rev Primers Start End %GC mT Length

tacatctcgccattacctgc 732 751 50 55.61

tcaacctcatctcctccaag 1181 1162 50 54.8 430

atacatctcgccattacctg 731 750 45 52.87

tcaacctcatctcctccaag 1181 1162 50 54.8 431

tacatctcgccattacctgc 732 751 50 55.61

aaatcaacctcatctcctcc 1184 1165 45 52.9 433

3-88
 Exploring Primer Design

tacatctcgccattacctgc 732 751 50 55.61

gaaatcaacctcatctcctc 1185 1166 45 51.08 434

atacatctcgccattacctg 731 750 45 52.87

aaatcaacctcatctcctcc 1184 1165 45 52.9 434

atacatctcgccattacctg 731 750 45 52.87

gaaatcaacctcatctcctc 1185 1166 45 51.08 435

ggatacatctcgccattacc 729 748 50 53.45

tcaacctcatctcctccaag 1181 1162 50 54.8 433

tacatctcgccattacctgc 732 751 50 55.61

gtgaaatcaacctcatctcc 1187 1168 45 51.63 436

tacatctcgccattacctgc 732 751 50 55.61

ggtgaaatcaacctcatctc 1188 1169 45 51.63 437

atacatctcgccattacctg 731 750 45 52.87

gtgaaatcaacctcatctcc 1187 1168 45 51.63 437

Find Restriction Enzymes That Cut Inside the Primer

Use the rebasecuts function to list all the restriction enzymes from the REBASE® database [2] that
will cut a primer. These restriction enzymes can be used in the design of cloning experiments. For
example, you can use this on the first pair of primers from the list of possible primers that you just
calculated.

fwdprimer = good_fwdprimers(f(pairs(1)),:)
fwdcutter = unique(rebasecuts(fwdprimer))

revprimer = good_revprimers(r(pairs(1)),:)
revcutter = unique(rebasecuts(revprimer))

fwdprimer =

'tacatctcgccattacctgc'

fwdcutter =

14x1 cell array

{'AbaSI' }
{'Acc36I'}
{'BfuAI' }
{'BmeDI' }
{'BspMI' }
{'BveI' }
{'FspEI' }
{'LpnPI' }
{'MspJI' }
{'RlaI' }
{'SetI' }
{'SgeI' }

3-89
3 Sequence Analysis

{'SgrTI' }
{'YkrI' }

revprimer =

'tcaacctcatctcctccaag'

revcutter =

12x1 cell array

{'AbaSI' }
{'AspBHI'}
{'BmeDI' }
{'BsaXI' }
{'FspEI' }
{'MnlI' }
{'MspJI' }
{'RlaI' }
{'SetI' }
{'SgeI' }
{'SgrTI' }
{'YkrI' }

References

[1] SantaLucia, J. Jr., "A unified view of polymer, dumbbell, and oligonucleotide DNA nearest-neighbor
thermodynamics", PNAS, 95(4):1460-5, 1998.

[2] Roberts, R.J., et al., "REBASE--restriction enzymes and DNA methyltransferases", Nucleic Acids
Research, 33:D230-2, 2005.

3-90
 Identifying Over-Represented Regulatory Motifs

Identifying Over-Represented Regulatory Motifs

This example illustrates a simple approach to searching for potential regulatory motifs in a set of co-
expressed genomic sequences by identifying significantly over-represented ungapped words of fixed
length. The discussion is based on the case study presented in Chapter 10 of "Introduction to
Computational Genomics. A Case Studies Approach" [1].

Introduction

The circadian clock is the 24 hour cycle of the physiological processes that synchronize with the
external day-night cycle. Most of the work on the circadian oscillator in plants has been carried out
using the model plant Arabidopsis thaliana. In this organism, the regulation of a series of genes that
need to be turned on or off at specific time of the day and night, is accomplished by small regulatory
sequences found upstream the genes in question. One such regulatory motif, AAAATATCT, also known
as the Evening Element (EE), has been identified in the promoter regions of circadian clock-regulated
genes that show peak expression in the evening [2].

Loading Upstream Regions of Clock-Regulated Genes

We consider three sets of clock-regulated genes, clustered according to the time of the day when they
are maximally expressed: set 1 corresponds to 1 KB-long upstream regions of genes whose
expression peak in the morning (8am-4pm); set 2 corresponds to 1 KB-long upstream regions of
genes whose expression peak in the evening (4pm-12pm); set 3 corresponds to 1 KB-long upstream
regions of genes whose expression peak in the night (12pm-8am). Because we are interested in a
regulatory motif in evening genes, set 2 represents our target set, while set 1 and set 3 are used as
background. In each set, the sequences and their respective reverse complements are concatenated
to each other, with individual sequences separated by a gap symbol (-).
load evemotifdemodata.mat;

% === concatenate both strands

s1 = [[set1.Sequence] seqrcomplement([set1.Sequence])];
s2 = [[set2.Sequence] seqrcomplement([set2.Sequence])];
s3 = [[set3.Sequence] seqrcomplement([set3.Sequence])];

% === compute length and number of sequences in each set

L1 = length(set1(1).Sequence);
L2 = length(set2(1).Sequence);
L3 = length(set3(1).Sequence);

N1 = numel(set1) * 2;
N2 = numel(set2) * 2;
N3 = numel(set3) * 2;

% === add separator between sequences

seq1 = seqinsertgaps(s1, 1:L1:(L1*N1)+N1, 1);
seq2 = seqinsertgaps(s2, 1:L2:(L2*N2)+N2, 1);
seq3 = seqinsertgaps(s3, 1:L3:(L3*N3)+N3, 1);

Identifying Over-Represented Words

To determine which candidate motif is over-represented in a given target set with respect to the
background set, we identify all possible W-mers (words of length W) in both sets and compute their
frequency. A word is considered over-represented if its frequency in the target set is significantly
higher than the frequency in the background set. This difference is also called "margin".

3-91
3 Sequence Analysis

type findOverrepresentedWords

function [nmersSorted, freqDiffSorted] = findOverrepresentedWords(seq, seq0, W)

% FINDOVERREPRESENTEDWORDS helper for evemotifdemo

% Copyright 2007 The MathWorks, Inc.

%=== find and count words of length W

nmers0 = nmercount(seq0, W);
nmers = nmercount(seq, W);

%=== compute frequency of words

f = [nmers{:,2}]/(length(seq) - W + 1);
f0 = [nmers0{:,2}]/(length(seq0) - W + 1);

%=== determine words common to both set

[nmersInt, i1, i2] = intersect(nmers(:,1),nmers0(:,1));
freqDiffInt = (f(i1) - f0(i2))';

%=== determine words specific to one set only

[nmersXOr, i3, i4] = setxor(nmers(:,1),nmers0(:,1));
c0 = nmers(i3,1);
d0 = nmers0(i4,1);
nmersXOr = [c0; d0];
freqDiffXOr = [f(i3) -f0(i4)]';

%=== define all words and their difference in frequency (margin)

nmersAll = [nmersInt; nmersXOr];
freqDiff = [freqDiffInt; freqDiffXOr];

%=== sort according to descending difference in frequency

[freqDiffSorted, freqDiffSortedIndex] = sort(freqDiff, 'descend');
nmersSorted = nmersAll(freqDiffSortedIndex);

The Evening Element Motif

If we consider all words of length W = 9 that appear more frequently in the target set (upstream
region of genes highly expressed in the evening) with respect to the background set (upstream region
of genes highly expressed in the morning and night), we notice that the most over-represented word
is 'AAAATATCT', also known as the Evening Element (EE) motif.

W = 9;

[words, freqDiff] = findOverrepresentedWords(seq2, [seq1 seq3],W);

words(1:10)
freqDiff(1:10)

ans =

10x1 cell array

{'AAAATATCT'}
{'AGATATTTT'}
{'CTCTCTCTC'}
{'GAGAGAGAG'}

3-92
 Identifying Over-Represented Regulatory Motifs

{'AGAGAGAGA'}
{'TCTCTCTCT'}
{'AAATATCTT'}
{'AAGATATTT'}
{'AAAAATATC'}
{'GATATTTTT'}

ans =

1.0e-03 *

0.1439
0.1439
0.1140
0.1140
0.1074
0.1074
0.0713
0.0713
0.0695
0.0695

Filtering out Repeats

Besides the EE motif, other words of length W = 9 appear to be over-represented in the target set. In
particular, we notice the presence of repeats, i.e., words consisting of a single nucleotide or dimer
repeated for the entire word length, such as 'CTCTCTCTC'. This phenomenon is quite common in
genomic sequences and generally is associated with non-functional components. Because in this
context the repeats are unlikely to be biologically significant, we filter them out.
% === determine repeats
wordsN = numel(words);
r = zeros(wordsN,1);

for i = 1:wordsN
if (all(words{i}(1:2:end) == words{i}(1)) && ... % odd positions are the same
all(words{i}(2:2:end) == words{i}(2))) % even positions are the same
r(i) = 1;
end
end
r = logical(r);

% === filter out repeats

words = words(~r);
freqDiff = freqDiff(~r);

% === consider the top 10 motifs

motif = words(1:10)
margin = freqDiff(1:10)

EEMotif = motif{1}
EEMargin = margin(1)

motif =

3-93
3 Sequence Analysis

10x1 cell array

{'AAAATATCT'}
{'AGATATTTT'}
{'AAATATCTT'}
{'AAGATATTT'}
{'AAAAATATC'}
{'GATATTTTT'}
{'AAATAAAAT'}
{'ATTTTATTT'}
{'TAAATAAAA'}
{'TTTTATTTA'}

margin =

1.0e-03 *

0.1439
0.1439
0.0713
0.0713
0.0695
0.0695
0.0656
0.0656
0.0600
0.0600

EEMotif =

'AAAATATCT'

EEMargin =

1.4393e-04

After removing the repeats, we observe that the EE motif ('AAAATATCT') and its reverse complement
('AGATATTTT') are at the top of the list. The other over-represented words are either simple variants
of the EE motif, such as 'AAATATCTT', 'AAAAATATC', 'AAATATCTC', or their reverse complements,
such as 'AAGATATTT', 'GATATTTTT', 'GAGATATTT'.

Assessing the Statistical Significance of Margins

Various techniques can be used to assess the statistical significance of the margin computed for the
EE motif. For example, we can repeat the analysis using some control sequences and evaluate the
resulting margins with respect to the EE margin. Genomic regions of Arabidopsis thaliana that are
further away from the transcription start site are good candidates for this purpose. Alternatively, we
could randomly split and shuffle the sequences under consideration and use these as controls.
Another simple solution is to generate random sequences according to the nucleotide composition of
the three original sets of sequences, as shown below.

% === find base composition of each set

bases1 = basecount(s1);

3-94
 Identifying Over-Represented Regulatory Motifs

bases2 = basecount(s2);
bases3 = basecount(s3);

% === generate random sequences according to base composition

rs1 = randseq(length(s1),'fromstructure', bases1);
rs2 = randseq(length(s2),'fromstructure', bases2);
rs3 = randseq(length(s3),'fromstructure', bases3);

% === add separator between sequences

rseq1 = seqinsertgaps(rs1, 1:L1:(L1*N1)+N1, 1);
rseq2 = seqinsertgaps(rs2, 1:L2:(L2*N2)+N2, 1);
rseq3 = seqinsertgaps(rs3, 1:L3:(L3*N3)+N3, 1);

% === compute margins for control set

[words, freqDiff] = findOverrepresentedWords(rseq2, [rseq1 rseq3],W);

The variable ctrlMargin holds the estimated margins of the top motifs for each of the 100 control
sequences generated as described above. The distribution of these margins can be approximated by
the extreme value distribution. We use the function gevfit from the Statistics and Machine Learning
Toolbox™ to estimate the parameters (shape, scale, and location) of the extreme value distribution
and we overlay a scaled version of its probability density function, computed using gevpdf, with the
histogram of the margins of the control sequences.

% === estimate parameters of distribution

nCtrl = length(ctrlMargin);
buckets = ceil(nCtrl/10);
parmhat = gevfit(ctrlMargin);
k = parmhat(1); % shape parameter
sigma = parmhat(2); % scale parameter
mu = parmhat(3); % location parameter

% === compute probability density function

x = linspace(min(ctrlMargin), max([ctrlMargin EEMargin]));
y = gevpdf(x, k, sigma, mu);

% === scale probability density function

[v, c] = hist(ctrlMargin,buckets);
binWidth = c(2) - c(1);
scaleFactor = nCtrl * binWidth;

% === overlay
figure()
hold on;
hist(ctrlMargin, buckets);
h = findobj(gca,'Type','patch');
h.FaceColor = [.9 .9 .9];
plot(x, scaleFactor * y, 'r');
stem(EEMargin, 1, 'b');
xlabel('Margin');
ylabel('Number of sequences');
legend('Ctrl Margins', 'EVD pdf', 'EE Margin');

3-95
3 Sequence Analysis

The control margins are the differences in frequency that we would expect to find when a word is
over-represented by chance alone. The margin relative to the EE motif is clearly significantly larger
than the control margins, and does not fit within the probability density curve of the random controls.
Because the EE margin is larger than all 100 control margins, we can conclude that the over-
representation of the EE motif in the target set is statistically significant and the p-value estimate is
less than 0.01.

Selecting Motif Length

If we repeat the search for over-represented words of length W = 6...11, we observe that all the top
motifs are either substrings (if W < 9) or superstrings (if W > 9) of the EE motif. Thus, how do we
decide what is the correct length of this motif? We can expect that the optimal length maximizes the
difference in frequency between the motif in the target set and the same motif in the background set.
However, in order to compare the margin across different lengths, the margin must be normalized to
account for the natural tendency of shorter words to occur more frequently. We perform this
normalization by dividing each margin by the margin corresponding to the most over-represented
word of identical length in a random set of sequences with a nucleotide composition similar to the
target set. For convenience, the top over-represented words for length W = 6...11 and their margins
are stored in the variables topMotif and topMargin. Similarly, the top over-represented words for
length W = 6...11 and their margins in a random set are stored in the variables rTopMotif and
rTopMargin.
% === top over-represented words, W = 6...11 in set 2 (evening)
topMotif
topMargin

3-96
 Identifying Over-Represented Regulatory Motifs

% === top over-represented words, W = 6...11 in random set

rTopMotif
rTopMargin

% === compute score

score = topMargin ./ rTopMargin;
[bestScore, bestLength] = max(score);

% === plot
figure()
plot(6:11, score(6:11));
xlabel('Motif length');
ylabel('Normalized margin');
title('Optimal motif length');
hold on
line([bestLength bestLength], [0 bestScore], 'LineStyle', '-.')

topMotif =

11x1 cell array

{0x0 double }
{0x0 double }
{0x0 double }
{0x0 double }
{0x0 double }
{'AATATC' }
{'AATATCT' }
{'AAATATCT' }
{'AAAATATCT' }
{'AAAAATATCT' }
{'AAAAAATATCT'}

topMargin =

1.0e-03 *

NaN
NaN
NaN
NaN
NaN
0.3007
0.2607
0.2074
0.1439
0.0648
0.0424

rTopMotif =

11x1 cell array

{0x0 double }
{0x0 double }

3-97
3 Sequence Analysis

{0x0 double }
{0x0 double }
{0x0 double }
{'ATTATA' }
{'TATAATA' }
{'TTATTAAA' }
{'GTTATTAAA' }
{'ATTATATATC' }
{'ATGTTATTATT'}

rTopMargin =

1.0e-03 *

NaN
NaN
NaN
NaN
NaN
0.5650
0.2374
0.0972
0.0495
0.0279
0.0183

3-98
 Identifying Over-Represented Regulatory Motifs

By plotting the normalized margin versus the motif length, we find that length W = 9 is the most
informative in discriminating over-represented motifs in the target sequence (evening set) against the
background set (morning and night sets).

Determining the Evening Element Motif Presence Among Clock-Regulated Genes

Although the EE Motif has been identified and experimentally validated as a regulatory motif for
genes whose expression peaks in the evening hours, it is not shared by all evening genes, nor is it
exclusive of these genes. We count the occurrences of the EE motif in the three sequence sets and
determine what proportion of genes in each set contain the motif.

EECount = zeros(3,1);

% === determine positions where EE motif occurs

loc1 = strfind(seq1, EEMotif);
loc2 = strfind(seq2, EEMotif);
loc3 = strfind(seq3, EEMotif);

% === count occurrences

EECount(1) = length(loc1);
EECount(2) = length(loc2);
EECount(3) = length(loc3);

% === find proportions of genes with EE Motif

NumGenes = [N1; N2; N3] / 2;
EEProp = EECount ./ NumGenes;

% === plot
figure()
bar(EEProp, 0.5);
ylabel('Proportion of genes containing EE Motif');
xlabel('Gene set');
title('Presence of EE Motif');
ylim([0 1])
ax = gca;
ax.XTickLabel = {'morning', 'evening', 'night'};

3-99
3 Sequence Analysis

It appears as though about 9% of genes in set 1, 40% of genes in set 2, and 13% of genes in set 3
have the EE motif. Thus, not all genes in set 2 have the motif, but it is clearly enriched in this group.

Analyzing the Evening Element Motif Location

Unlike many other functional motifs, the EE motif does not appear to accumulate at specific gene
locations in the set of sequences analyzed. After determining the location of each occurrence with
respect to the transcription start site (TSS), we observe a relatively uniform distribution of
occurrences across the upstream region of the genes considered, with the possible exception of the
middle region (between 400 and 500 bases upstream of the TSS).

offset = rem(loc2, 1001);

figure();
hist(offset, 100);
xlabel('Offset in upstream region (TSS = 0)');
ylabel('Number of sequences');

3-100
 Identifying Over-Represented Regulatory Motifs

References

[1] Cristianini, N. and Hahn, M.W., "Introduction to Computational Genomics: A Case Studies
Approach", Cambridge University Press, 2007.

[2] Harmer, S.L., et al., "Orchestrated Transcription of Key Pathways in Arabidopsis by the Circadian
Clock", Science, 290(5499):2110-3, 2000.

3-101
3 Sequence Analysis

Predicting and Visualizing the Secondary Structure of RNA

Sequences

This example illustrates how to use the rnafold and rnaplot functions to predict and plot the
secondary structure of an RNA sequence.

Introduction

RNA plays an important role in the cell, both as genetic information carrier (mRNA) and as functional
element (tRNA, rRNA). Because the function of an RNA sequence is largely associated with its
structure, predicting the RNA structure from its sequence has become increasingly important.
Because base pairing and base stacking represent the majority of the free energy contribution to
folding, a good estimation of secondary structure can be very helpful not only in the interpretation of
the function and reactivity, but also in the analysis of the tertiary structure of the RNA molecule.

RNA Secondary Structure Prediction Using Nearest-Neighbor Thermodynamic Model

The secondary structure of an RNA sequence is determined by the interaction between its bases,
including hydrogen bonding and base stacking. One of the many methods for RNA secondary
structure prediction uses the nearest-neighbor model and minimizes the total free energy associated
with an RNA structure. The minimum free energy is estimated by summing individual energy
contributions from base pair stacking, hairpins, bulges, internal loops and multi-branch loops. The
energy contributions of these elements are sequence- and length-dependent and have been
experimentally determined [1]. The rnafold function uses the nearest-neighbor thermodynamic
model to predict the minimum free-energy secondary structure of an RNA sequence. More
specifically, the algorithm implemented in rnafold uses dynamic programming to compute the
energy contributions of all possible elementary substructures and then predicts the secondary
structure by considering the combination of elementary substructures whose total free energy is
minimum. In this computation, the contribution of coaxially stacked helices is not accounted for, and
the formation of pseudoknots (non-nested structural elements) is forbidden.

Secondary Structure of Transfer RNA Phenylalanine

tRNAs are small molecules (73-93 nucleotides) that during translation transfer specific amino acids to
the growing polypeptide chain at the ribosomal site. Although at least one tRNA molecule exists for
each amino acid type, both secondary and tertiary structures are well conserved among the various
tRNA types, most likely because of the necessity of maintaining reliable interaction with the
ribosome. We consider the following tRNA-Phe sequence from Saccharomyces cerevisiae and predict
the minimum free-energy secondary structure using the function rnafold.
% === Predict secondary structure in bracket notation
phe_seq = 'GCGGAUUUAGCUCAGUUGGGAGAGCGCCAGACUGAAGAUCUGGAGGUCCUGUGUUCGAUCCACAGAAUUCGCACCA';
phe_str = rnafold(phe_seq)

phe_str =

'(((((((..((((........)))).(((((.......))))).....(((((.......))))))))))))....'

In the bracket notation, each dot represents an unpaired base, while a pair of equally nested, opening
and closing brackets represents a base pair. Alternative representations of RNA secondary structures
can be drawn using the function rnaplot. For example, the structure predicted above can be

3-102
 Predicting and Visualizing the Secondary Structure of RNA Sequences

displayed as a rooted tree, where leaf nodes correspond to unpaired residues and internal nodes
(except the root) correspond to base pairs. You can display the position and type of each residue by
clicking on the corresponding node.

% === Plot RNA secondary structure as tree

rnaplot(phe_str, 'seq', phe_seq, 'format', 'tree');

The tRNA secondary structure is commonly represented in a diagram plot and resembles a clover
leaf. It displays four base-paired stems (or "arms") and three loops. Each of the four stems has been
extensively studied and characterized: acceptor stem (positions 1-7 and 66-72), D-stem (positions
10-13 and 22-25), anticodon stem (positions 27-31 and 39-43) and T-stem (positions 49-53 and 61-65).
We can draw the tRNA secondary structure as a two-dimensional plot where each residue is identified
by a dot and the backbone and the hydrogen bonds are represented as lines between the dots. The
stems consist of consecutive stretches of base paired residues (blue dots), while the loops are formed
by unpaired residues (red dots).

% === Plot the secondary structure using the dot diagram representation
rnaplot(phe_str, 'seq', phe_seq, 'format', 'dot');

text(500, 200, 'T-stem');

text(100, 600, 'Anticodon stem');
text(550, 650, 'D-stem stem');
text(700, 400, 'Acceptor stem');

3-103
3 Sequence Analysis

While all the stems are important for a proper three-dimensional folding of the molecule and
successful interplay with ribosome and tRNA synthetases, the acceptor stem and the anticodon stem
are particularly interesting because they include the attachment site and the anticodon triplet. The
attachment site (positions 74-76) occurs at the 3' end of the RNA chains and consists of the sequence
C-C-A in all amino acid acceptor stems. The anticodon triplet consists of 3 bases that pair with a
complementary codon in the messenger RNA. In the case of Phe-tRNA, the anticodon sequence A-A-G
(positions 34-36) pairs with the mRNA codon U-U-C, encoding the amino acid phenylalanine. We can
redraw the structure and highlight these regions in the acceptor stem and anticodon stem by using
the selection property:

aag_pos = 34:36;
cca_pos = 74:76;

rnaplot(phe_str, 'sequence', phe_seq, 'format', 'diagram', ...

'selection', [aag_pos, cca_pos]);

3-104
 Predicting and Visualizing the Secondary Structure of RNA Sequences

The segregation of the sequence into four separate stems is better appreciated by displaying the
structure as graph plot. Each residue is represented on the abscissa and semi-elliptical lines connect
bases that pair with each other. The lack of pseudoknots in the secondary structure is reflected by the
absence of intersecting lines. This is expected in tRNA secondary structures and anticipated because
the dynamic programming method used does not allow pseudoknots.

rnaplot(phe_str, 'sequence', phe_seq, 'format', 'graph');

3-105
3 Sequence Analysis

Similar observations can be drawn by displaying the secondary structure as a circle, where each base
is represented by a dot on the circumference of a circle of arbitrary size, and bases that pair with
each other are connected by lines. The lines are visually clustered into four distinct groups, separated
by stretched of unpaired residues. We can hide the unpaired residues by using H.Unpaired, the
handle returned with the colorby property set to state.

[ha, H] = rnaplot(phe_str, 'sequence', phe_seq, 'format', 'circle', ...

'colorby', 'state');
H.Unpaired.Visible = 'off';
legend off;

3-106
 Predicting and Visualizing the Secondary Structure of RNA Sequences

As you can see, the outputs of the rnaplot function include a MATLAB® structure H consisting of
handles that can be used to change the aspect properties of various residue subsets. For example, if
you set the color scheme using the colorby property set to residue, the dots are colored according
to the residue type, and you can change their property using the appropriate handle.

[ha, H] = rnaplot(phe_str, 'sequence', phe_seq, 'format', 'circle', 'colorby', 'residue')

ha =

Axes (Bioinfo:rnaplot:circle) with properties:

XLim: [-1 1]
YLim: [-1 1.1000]
XScale: 'linear'
YScale: 'linear'
GridLineStyle: '-'
Position: [0.1124 0.1100 0.6703 0.8150]
Units: 'normalized'

Use GET to show all properties

H =

struct with fields:

3-107
3 Sequence Analysis

A: [1x1 Line]
C: [1x1 Line]
G: [1x1 Line]
U: [1x1 Line]
Selected: [0x1 Line]

H.G.Color = [0.5 0.5 0.5];

H.G.Marker = '*';
H.C.Color = [0.5 0.5 0.5];
H.C.Marker = '+';

3-108
 Predicting and Visualizing the Secondary Structure of RNA Sequences

Conservation of Transfer RNA Phenylalanine

Despite some differences in their primary sequences, tRNAs molecules present a secondary structure
pattern that is well conserved across the three phylogenetic domains. Consider the structure of the
tRNA-Phe of one representative organism for each phylogenetic domain: Saccharomyces cerevisiae
for the Eukaryotes, Haloarcula marismortui for the Archaea, and Thermus thermophilus for the
Bacteria. Then predict and plot their secondary structures using the mountain plot representation.

yeast = 'GCGGACUUAGCUCAGUUGGGAGAGCGCCAGACUGAAGAUCUGGAGGUCCUGUGUUCGAUCCACAGAGUUCGCACCA';
halma = 'GCCGCCUUAGCUCAGACUGGGAGAGCACUCGACUGAAGAUCGAGCUGUCCCCGGUUCAAAUCCGGGAGGCGGCACCA';
theth = 'GCCGAGGUAGCUCAGUUGGUAGAGCAUGCGACUGAAAAUCGCAGUGUCGGCGGUUCGAUUCCGCCCCUCGGCACCA';

yeast_str = rnafold(yeast);
theth_str = rnafold(theth);
halma_str = rnafold(halma);

h1 = rnaplot(yeast_str, 'sequence', yeast, 'format', 'mountain');

title(h1, 'tRNA-Phe Saccharomyces cerevisiae');
legend hide;

h2 = rnaplot(halma_str, 'sequence', halma, 'format', 'mountain');

title(h2, 'tRNA-Phe Haloarcula marismortui');
legend hide;

h3 = rnaplot(theth_str, 'sequence', theth, 'format', 'mountain');

title(h3, 'tRNA-Phe Thermus thermophilus');
legend hide;

3-109
3 Sequence Analysis

3-110
Predicting and Visualizing the Secondary Structure of RNA Sequences

3-111
3 Sequence Analysis

The similarity among the resulting structures is striking, the only difference being one extra residue
in the D-loop of Haloarcula marismortui, displayed in the first flat slope in the mountain plot.

The G-U Wobble Base Pair

Besides the Watson-Crick base pairs (A-U, G-C), virtually every class of functional RNA presents G-U
wobble base pairs. G-U pairs have an array of distinctive chemical, structural and conformational
properties: they have high affinity for metal ions, they are almost thermodynamically as stable as
Watson-Crick base pairs, and they present conformational flexibility to different environments. The
wobble pair at the third position of the acceptor helix of tRNA is very highly conserved in almost all
organisms. This conservation suggests that the G-U pair possesses unique features that can hardly be
duplicated by other pairs. You can observe the base pair type distribution on the secondary structure
diagram by coloring the base pairs according to their type.

rnaplot(yeast_str, 'sequence', yeast, 'format', 'diagram', 'colorby', 'pair');

3-112
 Predicting and Visualizing the Secondary Structure of RNA Sequences

References

[1] Matthews, D., Sabina, J., Zuker, M., and Turner, D. "Expanded sequence dependence of
thermodynamic parameters improves prediction of RNA secondary structure", Journal of Molecular
Biology, 288(5):911-40, 1999.

3-113
3 Sequence Analysis

Using HMMs for Profile Analysis of a Protein Family

This example shows how HMM profiles are used to characterize protein families. Profile analysis is a
key tool in bioinformatics. The common pairwise comparison methods are usually not sensitive and
specific enough for analyzing distantly related sequences. In contrast, Hidden Markov Model (HMM)
profiles provide a better alternative to relate a query sequence to a statistical description of a family
of sequences. HMM profiles use a position-specific scoring system to capture information about the
degree of conservation at various positions in the multiple alignment of these sequences. HMM
profile analysis can be used for multiple sequence alignment, for database searching, to analyze
sequence composition and pattern segmentation, and to predict protein structure and locate genes by
predicting open reading frames.

Accessing PFAM Databases

Start this example with an already built HMM of a protein family. Retrieve the model for the well-
known 7-fold transmembrane receptor from the Sanger Institute database. The PFAM key number is
PF00002. Also retrieve the pre-aligned sequences used to train this model. More information about
the PFAM database can be found at http://pfam.xfam.org/.

hmm_7tm = gethmmprof(2);
seed_seqs = gethmmalignment(2,'type','seed');

For your convenience, previously downloaded sequences are included in a MAT-file. Note that data in
public repositories is frequently curated and updated; therefore the results of this example might be
slightly different when you use up-to-date datasets.

load('gpcrfam.mat','hmm_7tm','seed_seqs')

Models and alignments can also be stored and parsed in later directly from the files using the
pfamhmmread, fastaread and multialignread functions.

Display the names and contents of the first three loaded sequences using the seqdisp command.

seqdisp(seed_seqs([1 2 3]),'row',70)

ans =

23x81 char array

'>VIPR2_HUMAN/123-371 '
' 1 YILVKAIYTL GYSVS.LMSL ATGSIILCLF .RKLHCTR.N YIHLNLFLSF ILRAISVLVK .DDVLYSSS.'
' 71 GTLHCPD... .......... .......... ....QPSSW. ..V.GCKLSL VFLQYCIMAN FFWLLVEGLY'
'141 LHTLLVA... ...MLPP.RR CFLAYLLIGW GLPTVCIGAW TAAR...... .........L YLED......'
'211 ......TGC. WDTN.DHSVP W....WVIRI PILISIIVNF VLFISIIRIL LQKLT..... .SPDVGGNDQ'
'281 SQY....... .......... .......... ....KRLAKS TLLLIPLFGV HYMV..FAVF PISI...S.S'
'351 KYQILFELCL GSF....QGL VV '
' '
'>VIPR_CARAU/100-348 '
' 1 FRSVKIGYTI GHSVS.LISL TTAIVILCMS .RKLHCTR.N YIHMHLFVSF ILKAIAVFVK .DAVLYDVIQ'
' 71 ESDNCS.... .......... .......... .....TASV. ....GCKAVI VFFQYCIMAS FFWLLVEGLY'
'141 LHALLAVS.. ...FFSE.RK YFWWYILIGW GGPTIFIMAW SFAK...... .........A YFND......'
'211 ......VGC. WDIIENSDLF W....WIIKT PILASILMNF ILFICIIRIL RQKIN..... .CPDIGRNES'
'281 NQY....... .......... .......... ....SRLAKS TLLLIPLFGI NFII..FAFI PENI...K.T'
'351 ELRLVFDLIL GSF....QGF VV '

3-114
 Using HMMs for Profile Analysis of a Protein Family

' '
'>VIPR1_RAT/140-386 '
' 1 YNTVKTGYTI GYSLS.LASL LVAMAILSLF .RKLHCTR.N YIHMHLFMSF ILRATAVFIK .DMALFNSG.'
' 71 EIDHCS.... .......... .......... .....EASV. ....GCKAAV VFFQYCVMAN FFWLLVEGLY'
'141 LYTLLAVS.. ...FFSE.RK YFWGYILIGW GVPSVFITIW TVVR...... .........I YFED......'
'211 ......FGC. WDTI.INSSL W....WIIKA PILLSILVNF VLFICIIRIL VQKLR..... .PPDIGKNDS'
'281 SPY....... .......... .......... ....SRLAKS TLLLIPLFGI HYVM..FAFF PDNF...K.A'
'351 QVKMVFELVV GSF....QGF VV '

More information regarding how to store the profile HMM information in a MATLAB® structure is
found in the help for hmmprofstruct.

Profile HMM Alignment

To test the profile HMM alignment tool you can re-align the sequences from the multiple alignment to
the HMM model. First erase the periods in sequences used to format the downloaded aligned
sequences. Doing this removes the alignment information from the sequences.

seqs = strrep({seed_seqs.Sequence},'.','');
names = {seed_seqs.Header};

Now align all the proteins to the HMM profile.

fprintf('Aligning sequences ')

scores = zeros(numel(seqs),1);
aligned_seqs = cell(numel(seqs),1);
for sn=1:numel(seqs)
fprintf('.')
[scores(sn),aligned_seqs{sn}]=hmmprofalign(hmm_7tm,seqs{sn});
end
fprintf('\n')

Aligning sequences ................................

Next, send the results to the Web Browser to better explore the new multiple alignment. Columns
marked with * at the bottom indicate when the model was in a "match" or "delete" state.

hmmprofmerge(aligned_seqs,names,scores)

You can also explore the alignment from the command window; the hmmprofmerge function with one
output argument places the aligned sequences into a char array.

str = hmmprofmerge(aligned_seqs);
str(1:10,1:80)

ans =

10x80 char array

'YILVKAIYTLGYSVS.LMSLATGSIILCLF.RKLHCTR.NYIHLNLFLSFILRAISVLVK.DDVLYSSSG-TLH......'
'FRSVKIGYTIGHSVS.LISLTTAIVILCMS.RKLHCTR.NYIHMHLFVSFILKAIAVFVK.DAVLYDVIQESDN......'
'YNTVKTGYTIGYSLS.LASLLVAMAILSLF.RKLHCTR.NYIHMHLFMSFILRATAVFIK.DMALFNSG-EIDH......'
'FGAIKTGYTIGHSLS.LISLTAAMIILCIF.RKLHCTR.NYIHMHLFMSFIMRAIAVFIK.DIVLFESG-ESDH......'
'YLSVKALYTVGYSTS.LVTLTTAMVILCRF.RKLHCTR.NFIHMNLFVSFMLRAISVFIK.DWILYAEQD-SSH......'
'FSTVKIIYTTGHSIS.IVALCVAIAILVAL.RRLHCPR.NYIHTQLFATFILKASAVFLK.DAAIFQGDS-TDH......'
'LSTLKQLYTAGYATS.LISLITAVIIFTCF.RKFHCTR.NYIHINLFVSFILRATAVFIK.DAVLFSDET-QNH......'

3-115
3 Sequence Analysis

'FDRLGMIYTVGYSVS.LASLTVAVLILAYF.RRLHCTR.NYIHMHLFLSFMLRAVSIFVK.DAVLYSGATLDEA......'
'FERLYVMYTVGYSIS.FGSLAVAILIIGYF.RRLHCTR.NYIHMHLFVSFMLRATSIFVK.DRVVHAHIGVKEL......'
'ALNLFYLTIIGHGLS.IASLLISLGIFFYF.KSLSCQR.ITLHKNLFFSFVCNSVVTIIH.LTAVANNQALVAT......'

Looking for Similarity with Sequence Comparison

Having a profile HHM which describes this family has several advantages over plain sequence
comparison. Suppose that you have a new oligonucleotide that you want to relate to the 7-
transmembrane receptor family. For this example, get a protein sequence from NCBI and extract the
aminoacid sequence.
mousegpcr = getgenpept('NP_783573');
Bai3 = mousegpcr.Sequence;

This sequence is also provided in the MAT-file gpcrfam.mat.

load('gpcrfam.mat','mousegpcr')
Bai3 = mousegpcr.Sequence;

seqdisp(Bai3,'row',70)

ans =

22x82 char array

' 1 MKAVRNLLIY IFSTYLLVMF GFNAAQDFWC STLVKGVIYG SYSVSEMFPK NFTNCTWTLE NPDPTKYSIY'

' 71 LKFSKKDLSC SNFSLLAYQF DHFSHEKIKD LLRKNHSIMQ LCSSKNAFVF LQYDKNFIQI RRVFPTDFPG'
' 141 LQKKVEEDQK SFFEFLVLNK VSPSQFGCHV LCTWLESCLK SENGRTESCG IMYTKCTCPQ HLGEWGIDDQ'
' 211 SLVLLNNVVL PLNEQTEGCL TQELQTTQVC NLTREAKRPP KEEFGMMGDH TIKSQRPRSV HEKRVPQEQA'
' 281 DAAKFMAQTG ESGVEEWSQW SACSVTCGQG SQVRTRTCVS PYGTHCSGPL RESRVCNNTA LCPVHGVWEE'
' 351 WSPWSLCSFT CGRGQRTRTR SCTPPQYGGR PCEGPETHHK PCNIALCPVD GQWQEWSSWS HCSVTCSNGT'
' 421 QQRSRQCTAA AHGGSECRGP WAESRECYNP ECTANGQWNQ WGHWSGCSKS CDGGWERRMR TCQGAAVTGQ'
' 491 QCEGTGEEVR RCSEQRCPAP YEICPEDYLI SMVWKRTPAG DLAFNQCPLN ATGTTSRRCS LSLHGVASWE'
' 561 QPSFARCISN EYRHLQHSIK EHLAKGQRML AGDGMSQVTK TLLDLTQRKN FYAGDLLVSV EILRNVTDTF'
' 631 KRASYIPASD GVQNFFQIVS NLLDEENKEK WEDAQQIYPG SIELMQVIED FIHIVGMGMM DFQNSYLMTG'
' 701 NVVASIQKLP AASVLTDINF PMKGRKGMVD WARNSEDRVV IPKSIFTPVS SKELDESSVF VLGAVLYKNL'
' 771 DLILPTLRNY TVVNSKVIVV TIRPEPKTTD SFLEIELAHL ANGTLNPYCV LWDDSKSNES LGTWSTQGCK'
' 841 TVLTDASHTK CLCDRLSTFA ILAQQPREIV MESSGTPSVT LIVGSGLSCL ALITLAVVYA ALWRYIRSER'
' 911 SIILINFCLS IISSNILILV GQTQTHNKSI CTTTTAFLHF FFLASFCWVL TEAWQSYMAV TGKIRTRLIR'
' 981 KRFLCLGWGL PALVVATSVG FTRTKGYGTD HYCWLSLEGG LLYAFVGPAA AVVLVNMVIG ILVFNKLVSR'
'1051 DGILDKKLKH RAGQMSEPHS GLTLKCAKCG VVSTTALSAT TASNAMASLW SSCVVLPLLA LTWMSAVLAM'
'1121 TDKRSILFQI LFAVFDSLQG FVIVMVHCIL RREVQDAFRC RLRNCQDPIN ADSSSSFPNG HAQIMTDFEK'
'1191 DVDIACRSVL HKDIGPCRAA TITGTLSRIS LNDDEEEKGT NPEGLSYSTL PGNVISKVII QQPTGLHMPM'
'1261 SMNELSNPCL KKENTELRRT VYLCTDDNLR GADMDIVHPQ ERMMESDYIV MPRSSVSTQP SMKEESKMNI'
'1331 GMETLPHERL LHYKVNPEFN MNPPVMDQFN MNLDQHLAPQ EHMQNLPFEP RTAVKNFMAS ELDDNVGLSR'
'1401 SETGSTISMS SLERRKSRYS DLDFEKVMHT RKRHMELFQE LNQKFQTLDR FRDIPNTSSM ENPAPNKNPW'
'1471 DTFKPPSEYQ HYTTINVLDT EAKDTLELRP AEWEKCLNLP LDVQEGDFQT EV '

First, using local alignment compare the new sequence to one of the sequences in the multiple
alignment. For instance use the first sequence, in this case the human protein 'VIPR2'. The Smith-
Waterman algorithm (swalign) can make use of scoring matrices. Scoring matrices can capture the
probability of substitution of symbols. The sequences in this example are known to be only distantly
related, so BLOSUM30 is a good choice for the scoring matrix.
VIPR2 = seqs{1};
[sc_aa_affine, alignment] = swalign(Bai3,VIPR2,'ScoringMatrix',...

3-116
 Using HMMs for Profile Analysis of a Protein Family

'blosum30','gapopen',5,'extendgap',3,'showscore',true);

sc_aa_affine

sc_aa_affine =

69.6000

By looking at the scoring space, apparently, both sequences are related. However, this relationship
could not be inferred from a dot plot.

Bai3_aligned_region = strrep(alignment(1,:),'-','');
seqdotplot(VIPR2,Bai3_aligned_region,7,2)
ylabel('VIPR2'); xlabel('Bai3');

3-117
3 Sequence Analysis

Is either of these two examples enough evidence to affirm that these sequences are related? One way
to test this is to randomly create a fake sequence with the same distribution of amino acids and see
how it aligns to the family. Notice that the score of the local alignment between the fake sequence
and the VIPR2 protein is not significantly lower than the score of the alignment between the Bia3 and
VIPR2 proteins. To ensure reproducibility of the results of this example, we reset the global random
generator.

rng(0,'twister');
fakeSeq = randseq(1000,'FROMSTRUCTURE',aacount(VIPR2));
sc_fk_affine = swalign(fakeSeq,VIPR2,'ScoringMatrix','blosum30',...
'gapopen',5,'extendgap',3,'showscore',true)

sc_fk_affine =

60.4000

3-118
 Using HMMs for Profile Analysis of a Protein Family

In contrast, when you align both sequences to the family using the trained profile HMM, the score of
aligning the target sequence to the family profile is significantly larger than the score of aligning the
fake sequence.

sc_aa_hmm = hmmprofalign(hmm_7tm,Bai3)
sc_fk_hmm = hmmprofalign(hmm_7tm,fakeSeq)

sc_aa_hmm =

214.5286

sc_fk_hmm =

-49.1624

Exploring Profile HMM Alignment Options

Similarly to the swalign alignment function, when you use profile alignments you can visualize the
scoring space using the showscore option to the hmmprofalign function.

Display Bai3 aligned to the 7tm_2 family.

hmmprofalign(hmm_7tm,Bai3,'showscore',true);
title('log-odds score for best path: Bai3');

3-119
3 Sequence Analysis

Display the "fake" sequence aligned to the 7tm_2 family.

hmmprofalign(hmm_7tm,fakeSeq,'showscore',true);
title('log-odds score for best path: fake sequence');

3-120
 Using HMMs for Profile Analysis of a Protein Family

Display Bai3 globally aligned to the 7tm_2 family.

[sc_aa_hmm,align,ptrs] = hmmprofalign(hmm_7tm,Bai3);
Bai3_hmmaligned_region = Bai3(min(ptrs):max(ptrs));
hmmprofalign(hmm_7tm,Bai3_hmmaligned_region,'showscore',true);
title('log-odds score for best path: Bai3 aligned globally');

3-121
3 Sequence Analysis

Align tandemly repeated domains.

naa = numel(Bai3_hmmaligned_region);
repeats = randseq(1000,'FROMSTRUCTURE',aacount(Bai3)); %artificial example
repeats(200+(1:naa)) = Bai3_hmmaligned_region;
repeats(500+(1:naa)) = Bai3_hmmaligned_region;
repeats(700+(1:naa)) = Bai3_hmmaligned_region;
hmmprofalign(hmm_7tm,repeats,'showscore',true);
title('log-odds score for best path: Bai3 tandem repeats');

3-122
 Using HMMs for Profile Analysis of a Protein Family

Searching for Fragment Domains

In MATLAB®, you can search for fragment domains by manually activating the B->M and M->E
transition probabilities of the HMM model.

hmm_7tm_f = hmm_7tm;
hmm_7tm_f.BeginX(3:end)=.002;
hmm_7tm_f.MatchX(1:end-1,4)=.002;

Create a random sequence, or fragment model, with a small insertion of the Bai3 protein:

fragment = randseq(1000,'FROMSTRUCTURE',aacount(Bai3));
fragment(501:550) = Bai3_hmmaligned_region(101:150);

Try aligning the random sequence with the inserted peptide to both models, the global and fragment
model:

hmmprofalign(hmm_7tm,fragment,'showscore',true);
title('log-odds score for best path: PF00002 global ');
hmmprofalign(hmm_7tm_f,fragment,'showscore',true);
title('log-odds score for best path: PF00002 fragment domains');

3-123
3 Sequence Analysis

3-124
 Using HMMs for Profile Analysis of a Protein Family

Exploring the Profile HMMs

The function showhmmprof is an interactive tool to explore the profile HMM. Try right and left mouse
clicks over the model figures. There are three plots for each model: (1) the symbol emission
probabilities in the Match states, (2) the symbol emission probabilities in the Insert states, and (3)
the Transition probabilities.

showhmmprof(hmm_7tm,'scale','logodds')

3-125
3 Sequence Analysis

3-126
 Using HMMs for Profile Analysis of a Protein Family

An alternative method to explore a profile HMM is by creating a sequence logo from the multiple
alignment. A sequence logo displays the frequency of bases found at each position within a given
region, usually for a binding site. Using the hmm_7tm sequences, consider the portion of the
Parathyroid hormone-related peptide receptor (precursor) found at the n-terminus of the
PTRR_Human sequence. The seqlogo allows a quick visual comparison of how well this region is
conserved across the 7tm family.

seqlogo(str,'startat',1,'endat',20,'alphabet','AA')

3-127
3 Sequence Analysis

Profile Estimation

Profile HMMs can also be estimated from a multiple alignment. As new sequences related to the
family are found, it is possible to re-estimate the model parameters.

hmm_7tm_new = hmmprofestimate(hmm_7tm,str)

hmm_7tm_new =

struct with fields:

Name: '7tm_2'
PfamAccessionNumber: 'PF00002.19'
ModelDescription: '7 transmembrane receptor (Secretin family)'
ModelLength: 243
Alphabet: 'AA'
MatchEmission: [243x20 double]
InsertEmission: [243x20 double]
NullEmission: [0.0768 0.0418 0.0396 0.0305 0.0201 0.0378 ... ]
BeginX: [244x1 double]
MatchX: [242x4 double]

3-128
 Using HMMs for Profile Analysis of a Protein Family

InsertX: [242x2 double]

DeleteX: [242x2 double]
FlankingInsertX: [2x2 double]
LoopX: [2x2 double]
NullX: [2x1 double]

In case your sequences are not pre-aligned, you can also utilize the multialign function before
estimating a new HMM profile. It is possible to refine the HMM profile by re-aligning the sequences
to the model and re-estimating the model iteratively until you converge to a locally optimal model.

aligned_seqs = multialign(seqs);
hmm_7tm_ma = hmmprofestimate(hmmprofstruct(270),aligned_seqs)
showhmmprof(hmm_7tm_ma,'scale','logodds')
close; close; % close insertion emission prob. and transition prob.

hmm_7tm_ma =

struct with fields:

ModelLength: 270
Alphabet: 'AA'
MatchEmission: [270x20 double]
InsertEmission: [270x20 double]
NullEmission: [0.0500 0.0500 0.0500 0.0500 0.0500 0.0500 0.0500 ... ]
BeginX: [271x1 double]
MatchX: [269x4 double]
InsertX: [269x2 double]
DeleteX: [269x2 double]
FlankingInsertX: [2x2 double]
LoopX: [2x2 double]
NullX: [2x1 double]

Align all sequences to the new model.

3-129
3 Sequence Analysis

fprintf('Aligning sequences ')

scores = zeros(numel(seqs),1);
aligned_seqs = cell(numel(seqs),1);
for sn=1:numel(seqs)
fprintf('.')
[scores(sn),aligned_seqs{sn}]=hmmprofalign(hmm_7tm_ma,seqs{sn});
end
fprintf('\n')

str = hmmprofmerge(aligned_seqs);
str(1:10,1:80)

Aligning sequences ................................

ans =

10x80 char array

'YILVKAIYTLGYSVSLMSLATGSIILCLF.RKLHCTRNYIHLNLFLSFILRAISVLVKDDVLYSS---SGTLHCP-....'
'FRSVKIGYTIGHSVSLISLTTAIVILCMS.RKLHCTRNYIHMHLFVSFILKAIAVFVKDAVLYDVIQ--ESDNCS-....'
'YNTVKTGYTIGYSLSLASLLVAMAILSLF.RKLHCTRNYIHMHLFMSFILRATAVFIKDMALFNS---GEIDHCS-....'
'FGAIKTGYTIGHSLSLISLTAAMIILCIF.RKLHCTRNYIHMHLFMSFIMRAIAVFIKDIVLFES---GESDHCH-....'
'YLSVKALYTVGYSTSLVTLTTAMVILCRF.RKLHCTRNFIHMNLFVSFMLRAISVFIKDWILYAE---QDSSHCF-....'
'FSTVKIIYTTGHSISIVALCVAIAILVAL.RRLHCPRNYIHTQLFATFILKASAVFLKDAAIFQG---DSTDHCS-....'
'LSTLKQLYTAGYATSLISLITAVIIFTCF.RKFHCTRNYIHINLFVSFILRATAVFIKDAVLFSD---ETQNHCL-....'
'FDRLGMIYTVGYSVSLASLTVAVLILAYF.RRLHCTRNYIHMHLFLSFMLRAVSIFVKDAVLYSGATLDEAERLTE....'
'FERLYVMYTVGYSISFGSLAVAILIIGYF.RRLHCTRNYIHMHLFVSFMLRATSIFVKDRVVHAHIGVKELESLIM....'
'ALNLFYLTIIGHGLSIASLLISLGIFFYF.KSLSCQRITLHKNLFFSFVCNSVVTIIHLTAVANNQALVATNP---....'

Show the aligned sequences in the Help Browser.

hmmprofmerge(aligned_seqs,names,scores)

3-130
 Predicting Protein Secondary Structure Using a Neural Network

Predicting Protein Secondary Structure Using a Neural

Network

This example shows a secondary structure prediction method that uses a feed-forward neural
network and the functionality available with the Deep Learning Toolbox™.

It is a simplified example intended to illustrate the steps for setting up a neural network with the
purpose of predicting secondary structure of proteins. Its configuration and training methods are not
meant to be necessarily the best solution for the problem at hand.

Introduction

Neural network models attempt to simulate the information processing that occurs in the brain and
are widely used in a variety of applications, including automated pattern recognition.

The Rost-Sander data set [1] consists of proteins whose structures span a relatively wide range of
domain types, composition and length. The file RostSanderDataset.mat contains a subset of this
data set, where the structural assignment of every residue is reported for each protein sequence.
load RostSanderDataset.mat

N = numel(allSeq);

id = allSeq(7).Header % annotation of a given protein sequence

seq = int2aa(allSeq(7).Sequence) % protein sequence
str = allSeq(7).Structure % structural assignment

id =

'1CSE-ICOMPLEX(SERINEPROTEINASE-INHIBITOR)03-JU'

seq =

'KSFPEVVGKTVDQAREYFTLHYPQYNVYFLPEGSPVTLDLRYNRVRVFYNPGTNVVNHVPHVG'

str =

'CCCHHHCCCCHHHHHHHHHHHCCCCEEEEEECCCCEECCCCCCEEEEEEECCCCEECCCCEEC'

In this example, you will build a neural network to learn the structural state (helix, sheet or coil) of
each residue in a given protein, based on the structural patterns observed during a training phase.
Due to the random nature of some steps in the following approach, numeric results might be slightly
different every time the network is trained or a prediction is simulated. To ensure reproducibility of
the results, we reset the global random generator to a saved state included in the loaded file, as
shown below:
rng(savedState);

Defining the Network Architecture

For the current problem we define a neural network with one input layer, one hidden layer and one
output layer. The input layer encodes a sliding window in each input amino acid sequence, and a

3-131
3 Sequence Analysis

prediction is made on the structural state of the central residue in the window. We choose a window
of size 17 based on the statistical correlation found between the secondary structure of a given
residue position and the eight residues on either side of the prediction point [2]. Each window
position is encoded using a binary array of size 20, having one element for each amino acid type. In
each group of 20 inputs, the element corresponding to the amino acid type in the given position is set
to 1, while all other inputs are set to 0. Thus, the input layer consists of R = 17x20 input units, i.e. 17
groups of 20 inputs each.

In the following code, we first determine for each protein sequence all the possible subsequences
corresponding to a sliding window of size W by creating a Hankel matrix, where the ith column
represents the subsequence starting at the ith position in the original sequence. Then for each
position in the window, we create an array of size 20, and we set the jth element to 1 if the residue in
the given position has a numeric representation equal to j.
W = 17; % sliding window size

% === binarization of the inputs

for i = 1:N
seq = double(allSeq(i).Sequence); % current sequence
win = hankel(seq(1:W),seq(W:end)); % all possible sliding windows
myP = zeros(20*W,size(win,2)); % input matrix for current sequence
for k = 1:size(win, 2)
index = 20*(0:W-1)' + win(:,k); % input array for each position k
myP(index,k) = 1;
end
allSeq(i).P = myP;
end

The output layer of our neural network consists of three units, one for each of the considered
structural states (or classes), which are encoded using a binary scheme. To create the target matrix
for the neural network, we first obtain, from the data, the structural assignments of all possible
subsequences corresponding to the sliding window. Then we consider the central position in each
window and transform the corresponding structural assignment using the following binary encoding:
1 0 0 for coil, 0 1 0 for sheet, 0 0 1 for helix.
cr = ceil(W/2); % central residue position

% === binarization of the targets

for i = 1:N
str = double(allSeq(i).Structure); % current structural assignment
win = hankel(str(1:W),str(W:end)); % all possible sliding windows
myT = false(3,size(win,2));
myT(1,:) = win(cr,:) == double('C');
myT(2,:) = win(cr,:) == double('E');
myT(3,:) = win(cr,:) == double('H');
allSeq(i).T = myT;
end

You can perform the binarization of the input and target matrix described in the two steps above in a
more concise way by executing the following equivalent code:
% === concise binarization of the inputs and targets
for i = 1:N
seq = double(allSeq(i).Sequence);
win = hankel(seq(1:W),seq(W:end)); % concurrent inputs (sliding windows)

3-132
 Predicting Protein Secondary Structure Using a Neural Network

% === binarization of the input matrix

allSeq(i).P = kron(win,ones(20,1)) == kron(ones(size(win)),(1:20)');

% === binarization of the target matrix

allSeq(i).T = allSeq(i).Structure(repmat((W+1)/2:end-(W-1)/2,3,1)) == ...
repmat(('CEH')',1,length(allSeq(i).Structure)-W+1);
end

Once we define the input and target matrices for each sequence, we create an input matrix, P, and
target matrix, T, representing the encoding for all the sequences fed into the network.
% === construct input and target matrices
P = double([allSeq.P]); % input matrix
T = double([allSeq.T]); % target matrix

Creating the Neural Network

The problem of secondary structure prediction can be thought of as a pattern recognition problem,
where the network is trained to recognize the structural state of the central residue most likely to
occur when specific residues in the given sliding window are observed. We create a pattern
recognition neural network using the input and target matrices defined above and specifying a
hidden layer of size 3.
hsize = 3;
net = patternnet(hsize);
net.layers{1} % hidden layer
net.layers{2} % output layer

ans =

Neural Network Layer

name: 'Hidden'
dimensions: 3
distanceFcn: (none)
distanceParam: (none)
distances: []
initFcn: 'initnw'
netInputFcn: 'netsum'
netInputParam: (none)
positions: []
range: [3x2 double]
size: 3
topologyFcn: (none)
transferFcn: 'tansig'
transferParam: (none)
userdata: (your custom info)

ans =

Neural Network Layer

name: 'Output'
dimensions: 0
distanceFcn: (none)
distanceParam: (none)

3-133
3 Sequence Analysis

distances: []
initFcn: 'initnw'
netInputFcn: 'netsum'
netInputParam: (none)
positions: []
range: []
size: 0
topologyFcn: (none)
transferFcn: 'softmax'
transferParam: (none)
userdata: (your custom info)

Training the Neural Network

The pattern recognition network uses the default Scaled Conjugate Gradient algorithm for training,
but other algorithms are available (see the Deep Learning Toolbox documentation for a list of
available functions). At each training cycle, the training sequences are presented to the network
through the sliding window defined above, one residue at a time. Each hidden unit transforms the
signals received from the input layer by using a transfer function logsig to produce an output signal
that is between and close to either 0 or 1, simulating the firing of a neuron [2]. Weights are adjusted
so that the error between the observed output from each unit and the desired output specified by the
target matrix is minimized.

% === use the log sigmoid as transfer function

net.layers{1}.transferFcn = 'logsig';

% === train the network

[net,tr] = train(net,P,T);

3-134
 Predicting Protein Secondary Structure Using a Neural Network

During training, the training tool window opens and displays the progress. Training details such as
the algorithm, the performance criteria, the type of error considered, etc. are shown.

Use the function view to generate a graphical view of the neural network.

view(net)

3-135
3 Sequence Analysis

One common problem that occurs during neural network training is data overfitting, where the
network tends to memorize the training examples without learning how to generalize to new
situations. The default method for improving generalization is called early stopping and consists in
dividing the available training data set into three subsets: (i) the training set, which is used for
computing the gradient and updating the network weights and biases; (ii) the validation set, whose
error is monitored during the training process because it tends to increase when data is overfitted;
and (iii) the test set, whose error can be used to assess the quality of the division of the data set.

When using the function train, by default, the data is randomly divided so that 60% of the samples
are assigned to the training set, 20% to the validation set, and 20% to the test set, but other types of
partitioning can be applied by specifying the property net.divideFnc (default dividerand). The
structural composition of the residues in the three subsets is comparable, as seen from the following
survey:

[i,j] = find(T(:,tr.trainInd));
Ctrain = sum(i == 1)/length(i);
Etrain = sum(i == 2)/length(i);
Htrain = sum(i == 3)/length(i);

3-136
 Predicting Protein Secondary Structure Using a Neural Network

[i,j] = find(T(:,tr.valInd));
Cval = sum(i == 1)/length(i);
Eval = sum(i == 2)/length(i);
Hval = sum(i == 3)/length(i);

[i,j] = find(T(:,tr.testInd));
Ctest = sum(i == 1)/length(i);
Etest = sum(i == 2)/length(i);
Htest = sum(i == 3)/length(i);

figure()
pie([Ctrain; Etrain; Htrain]);
title('Structural assignments in training data set');
legend('C', 'E', 'H')

figure()
pie([Cval; Eval; Hval]);
title('Structural assignments in validation data set');
legend('C', 'E', 'H')

figure()
pie([Ctest; Etest; Htest]);
title('Structural assignments in testing data set ');
legend('C', 'E', 'H')

3-137
3 Sequence Analysis

3-138
 Predicting Protein Secondary Structure Using a Neural Network

The function plotperform display the trends of the training, validation, and test errors as training
iterations pass.

figure()
plotperform(tr)

3-139
3 Sequence Analysis

The training process stops when one of several conditions (see net.trainParam) is met. For
example, in the training considered, the training process stops when the validation error increases
for a specified number of iterations (6) or the maximum number of allowed iterations is reached
(1000).

% === display training parameters

net.trainParam

% === plot validation checks and gradient

figure()
plottrainstate(tr)

ans =

Function Parameters for 'trainscg'

Show Training Window Feedback showWindow: true

Show Command Line Feedback showCommandLine: false
Command Line Frequency show: 25
Maximum Epochs epochs: 1000
Maximum Training Time time: Inf
Performance Goal goal: 0
Minimum Gradient min_grad: 1e-06
Maximum Validation Checks max_fail: 6
Sigma sigma: 5e-05

3-140
 Predicting Protein Secondary Structure Using a Neural Network

Lambda lambda: 5e-07

Analyzing the Network Response

To analyze the network response, we examine the confusion matrix by considering the outputs of the
trained network and comparing them to the expected results (targets).

O = sim(net,P);
figure()
plotconfusion(T,O);

3-141
3 Sequence Analysis

The diagonal cells show the number of residue positions that were correctly classified for each
structural class. The off-diagonal cells show the number of residue positions that were misclassified
(e.g. helical positions predicted as coiled positions). The diagonal cells correspond to observations
that are correctly classified. Both the number of observations and the persentage of the total number
of observations are shown in each cell. The column on the far right of the plot shows the percentages
of all the examples predicted to belong to each class that are correctly and incorrectly classified.
These metrics are often called the precision (or positive predictive value) and false discovery rate,
respectively. The row at the bottom of the plot shows the percentages of all the examples belonging to
each class that are correctly and incorrectly classified. These metrics are often called the recall (or
true positive rate) and false negative rate, respectively. The cell in the bottom right of the plot shows
the overall accuracy.

We can also consider the Receiver Operating Characteristic (ROC) curve, a plot of the true positive
rate (sensitivity) versus the false positive rate (1 - specificity).

3-142
 Predicting Protein Secondary Structure Using a Neural Network

figure()
plotroc(T,O);

Refining the Neural Network for More Accurate Results

The neural network that we have defined is relative simple. To achieve some improvements in the
prediction accuracy we could try one of the following:

3-143
3 Sequence Analysis

• Increase the number of training vectors. Increasing the number of sequences dedicated to
training requires a larger curated database of protein structures, with an appropriate distribution
of coiled, helical and sheet elements.
• Increase the number of input values. Increasing the window size or adding more relevant
information, such as biochemical properties of the amino acids, are valid options.
• Use a different training algorithm. Various algorithms differ in memory and speed requirements.
For example, the Scaled Conjugate Gradient algorithm is relatively slow but memory efficient,
while the Levenberg-Marquardt is faster but more demanding in terms of memory.
• Increase the number of hidden neurons. By adding more hidden units we generally obtain a more
sophisticated network with the potential for better performances but we must be careful not to
overfit the data.

We can specify more hidden layers or increased hidden layer size when the pattern recognition
network is created, as shown below:

hsize = [3 4 2];
net3 = patternnet(hsize);

hsize = 20;
net20 = patternnet(hsize);

We can also assign the network initial weights to random values in the range -0.1 to 0.1 as suggested
by the study reported in [2] by setting the net20.IW and net20.LW properties as follows:

% === assign random values in the range -.1 and .1 to the weights
net20.IW{1} = -.1 + (.1 + .1) .* rand(size(net20.IW{1}));
net20.LW{2} = -.1 + (.1 + .1) .* rand(size(net20.LW{2}));

In general, larger networks (with 20 or more hidden units) achieve better accuracy on the protein
training set, but worse accuracy in the prediction accuracy. Because a 20-hidden-unit network
involves almost 7,000 weights and biases, the network is generally able to fit the training set closely
but loses the ability of generalization. The compromise between intensive training and prediction
accuracy is one of the fundamental limitations of neural networks.

net20 = train(net20,P,T);

O20 = sim(net20,P);
numWeightsAndBiases = length(getx(net20))

numWeightsAndBiases =

6883

3-144
 Predicting Protein Secondary Structure Using a Neural Network

You can display the confusion matrices for training, validation and test subsets by clicking on the
corresponding button in the training tool window.

Assessing Network Performance

You can evaluate structure predictions in detail by calculating prediction quality indices [3], which
indicate how well a particular state is predicted and whether overprediction or underprediction has
occurred. We define the index pcObs(S) for state S (S = {C, E, H}) as the number of residues

3-145
3 Sequence Analysis

correctly predicted in state S, divided by the number of residues observed in state S. Similarly, we
define the index pcPred(S) for state S as the number of residues correctly predicted in state S,
divided by the number of residues predicted in state S.

[i,j] = find(compet(O));
[u,v] = find(T);

% === compute fraction of correct predictions when a given state is observed

pcObs(1) = sum(i == 1 & u == 1)/sum (u == 1); % state C
pcObs(2) = sum(i == 2 & u == 2)/sum (u == 2); % state E
pcObs(3) = sum(i == 3 & u == 3)/sum (u == 3); % state H

% === compute fraction of correct predictions when a given state is predicted

pcPred(1) = sum(i == 1 & u == 1)/sum (i == 1); % state C
pcPred(2) = sum(i == 2 & u == 2)/sum (i == 2); % state E
pcPred(3) = sum(i == 3 & u == 3)/sum (i == 3); % state H

% === compare quality indices of prediction

figure()
bar([pcObs' pcPred'] * 100);
ylabel('Correctly predicted positions (%)');
ax = gca;
ax.XTickLabel = {'C';'E';'H'};
legend({'Observed','Predicted'});

3-146
 Predicting Protein Secondary Structure Using a Neural Network

These quality indices are useful for the interpretation of the prediction accuracy. In fact, in cases
where the prediction technique tends to overpredict/underpredict a given state, a high/low prediction
accuracy might just be an artifact and does not provide a measure of quality for the technique itself.

Conclusions

The method presented here predicts the structural state of a given protein residue based on the
structural state of its neighbors. However, there are further constraints when predicting the content
of structural elements in a protein, such as the minimum length of each structural element.
Specifically, a helix is assigned to any group of four or more contiguous residues, and a sheet is
assigned to any group of two or more contiguous residues. To incorporate this type of information, an
additional network can be created so that the first network predicts the structural state from the
amino acid sequence, and the second network predicts the structural element from the structural
state.

References

[1] Rost, B., and Sander, C., "Prediction of protein secondary structure at better than 70% accuracy",
Journal of Molecular Biology, 232(2):584-99, 1993.

[2] Holley, L.H. and Karplus, M., "Protein secondary structure prediction with a neural network",
PNAS, 86(1):152-6, 1989.

[3] Kabsch, W., and Sander, C., "How good are predictions of protein secondary structure?", FEBS
Letters, 155(2):179-82, 1983.

3-147
3 Sequence Analysis

Visualizing the Three-Dimensional Structure of a Molecule

This example shows how to display, inspect and annotate the three-dimensional structure of
molecules. This example performs a three-dimensional superposition of the structures of two related
proteins.

Introduction

Ubiquitin is a small protein of approximately 76 amino acids, found in all eukaryotic cells and very
well conserved among species. Through post-translational modification of a variety of proteins,
ubiquitin is involved in many diverse biological processes, including protein degradation, protein
trafficking, DNA repair, gene regulation, etc. Because of its ubiquitous presence in cells and its
involvement in many fundamental processes, ubiquitin has been the focus of extensive research at
the sequence, structural, and functional level.

You can view the three-dimensional structure of ubiquitin by downloading the crystal structure file
from the PDB database and then displaying it using the molviewer function. By default, the protein
structure is rendered such that each atom is represented by a ball and each bond is represented by a
stick. You can change the mode of rendering by selecting display options below the figure. You can
also rotate and manipulate the structure by click-dragging the protein or by entering Rasmol
commands in the Scripting Console.

In this example, we will explore the structural characteristics of ubiquitin through combinations of
Rasmol commands passed to the evalrasmolscript function. However, you can perform the same
analysis by using the Molecule Viewer window. The information for the ubiquitin protein is provided
in the MAT-file ubilikedata.mat.

load('ubilikedata.mat','ubi')

Alternatively, you can use the getpdb function to retrieve the protein information from the PDB
repository and load it into MATLAB®. Note that data in public repositories is frequently curated and
updated; therefore the results of this example might be slightly different when you use up-to-date
datasets.

ubi = getpdb('1ubi');

h1 = molviewer(ubi);

3-148
 Visualizing the Three-Dimensional Structure of a Molecule

evalrasmolscript(h1, 'select all; wireframe 100; background black;');

Rendering the Molecule

We can look at the ubiquitin fold by using the "cartoon" rendering, which clearly displays the
secondary structure elements. We restrict our selection to the protein, since we are not interested in
displaying other heterogeneous particles, such as water molecules.

% Display the molecule as cartoon and color the atoms according to their
% secondary structure assignment. Then remove other atoms and bonds.
evalrasmolscript(h1, ['spacefill off; wireframe off; ' ...
'restrict protein; cartoon on; color structure; ' ...
'center selected;']);

3-149
3 Sequence Analysis

Exploring the Molecule by Spinning and Zooming

The ubiquitin fold consists of five antiparallel beta strands, one alpha helix, a small 3-10 helix, and
several turns and loops. The fold resembles a small barrel, with the beta sheet forming one side and
the alpha helix forming the other side of the barrel. The bottom part is closed by the 3-10 helix. We
can better appreciate the compact, globular fold of ubiquitin by spinning the structure 360 degrees
and by zooming in and out using the "move" command.

% Animate the display by making the structure spin and zoom in

evalrasmolscript(h1, ['move 0 180 0 40 0 0 0 0 5; ' ... %
... % rotate y by 180, zoom in by 40, time = 5 sec
'move 0 180 0 -40 0 0 0 0 5;']);
% rotate y by 180, zoom out by 40, time = 5 sec

3-150
 Visualizing the Three-Dimensional Structure of a Molecule

Evaluating the Amino Acid Charge Distribution in the Structure

The compactness and high stability of the ubiquitin fold is related to the spatial distribution of
hydrophobic and hydrophilic amino acids in the folded state. We can look at the distribution of
charged amino acids by selecting positively and negatively charged residues and then by rendering
these atoms with different colors (red and blue respectively). We can also render water molecules as
white to see their relationship to the charged residues.
evalrasmolscript(h1, ['select protein; color gray; ' ...
'select positive; color red; spacefill 300; ' ...
'select negative; color blue; spacefill 300; ' ...
'select HOH; color white; spacefill 100;']); % water atoms

The charged amino acids are located primarily on the surface exposed to the solvent, where they
interact with the water molecules. In particular, we notice that the charge distribution is not uniform
across the sides of the ubiquitin's barrel. In fact, the side with the alpha helix appears to be more
crowded with charged amino acids than the side containing the beta strands.

3-151
3 Sequence Analysis

Exploring the Hydrophobicity Profile of the Structure

We can perform a similar analysis by looking at the spatial distribution of some hydrophobic amino
acids, such as Alanine, Isoleucine, Valine, Leucine and Methionine. You can also use the Rasmol label
"hydrophobic" to select all hydrophobic residues.

% color hydrophobic amino acids green

evalrasmolscript(h1, ['select all; spacefill off; color gray; ' ...
'select Ala or Ile or Val or Leu or Met; ' ...
'color green; wireframe 100;' ...
'move 90 0 0 0 0 0 0 0 1; move 0 -45 0 0 0 0 0 0 1']);

Unlike the charged amino acids above, the hydrophobic amino acids are located primarily in the
interior of the barrel. This gives high stability to the ubiquitin fold, since hydrophobic amino acids are
shielded from the solvent, making the protein structure compact and tight.

3-152
 Visualizing the Three-Dimensional Structure of a Molecule

Measuring Atomic Distances

Ubiquitin displays a tight fold with one alpha helix traversing one side of the small barrel. The length
of this alpha helix presents some variation among the representatives of the ubiquitin-like protein
family. We can determine the actual size of the helix either by double clicking on the relevant atoms
or by using MATLAB® and Rasmol commands as follows.

% reset the display to cartoons

evalrasmolscript(h1, ['reset; select all; spacefill off; wireframe off; '...
'cartoon on; color structure;']);

% determine the boundaries of the alpha helix

initHelixRes = ubi.Helix(1).initSeqNum % alpha helix starting residue

initHelixRes = 23

endHelixRes = ubi.Helix(1).endSeqNum % alpha helix ending residue

endHelixRes = 34

% highlight the starting and ending residues of helix

evalrasmolscript(h1, ['select ' num2str(initHelixRes) ' or ' ...
num2str(endHelixRes) '; color red; wireframe 100;']);

% determine atom numbers for starting and ending residues

initHelixAtoms = ubi.Model.Atom([ubi.Model.Atom(:).resSeq]==initHelixRes);
endHelixAtoms = ubi.Model.Atom([ubi.Model.Atom(:).resSeq]==endHelixRes);
initHelix = min([initHelixAtoms.AtomSerNo]); % Helix starting atom
endHelix = min([endHelixAtoms.AtomSerNo]); % Helix ending atom
evalrasmolscript(h1, ['measure ' num2str(initHelix) ' ' num2str(endHelix) ';']);

3-153
3 Sequence Analysis

Displaying and Labeling Lysine Residues in Ubiquitin Structure

The process of ubiquitination - the attachment of a ubiquitin molecule to a target protein - is

mediated by the formation of an isopeptide bond between the C-terminal 4-residue tail of ubiquitin
and a Lysine of the target protein. If the target protein is another ubiquitin, the process is called
polyubiquitination. Polyubiquitin chains consisting of at least four ubiquitins are used to tag the
target proteins for degradation by the proteasome. All seven Lysines in ubiquitin can be used in the
polyubiquitination process, resulting in different chains that alter the target protein in different ways.
We can look at the spatial distribution of Lysines on the ubiquitin fold by selecting and labeling the
alpha carbons of each Lysine in the structure.
% highlight the Lysine residues in the structure and the C-terminal tail
% involved in the isopeptide bond formation
evalrasmolscript(h1, ['restrict protein; cartoon off; wireframe off; measure off; ' ...
... % undo previous selection
'backbone 100; color structure; select Lys; wireframe 100; ' ...
... % select Lysines

3-154
 Visualizing the Three-Dimensional Structure of a Molecule

'select Lys and *.ca; spacefill 300; labels on; ' ...
... % label alpha carbons
'select 72-76; wireframe 100; color cyan; ']);
% select C-terminal tail

Several studies have shown that different roles are played by polyubiquitins when the molecules are
linked together through different Lysines. For example, Lys(11)-, Lys(29)-, and Lys(48)-linked
polyubiquitins target proteins for the proteasome (i.e., for degradation). In contrast, Lys(6)- and
Lys(63)-linked polyubiquitins are associated with reversible modifications, such as protein trafficking
control.

Examining the Isopeptide Bond in Diubiquitin

The crystal structure of a diubiquitin chain consisting of two moieties is represented in the PDB
record 1aar. We can view and label an actual isopeptide bond between the C-terminal tail of one
ubiquitin (labeled as chain A), and Lys(48) of the other ubiquitin (labeled as chain B).

3-155
3 Sequence Analysis

Retrieve the protein 1aar from PDB or load the data from the MAT-file.

aar = getpdb('1aar');

load('ubilikedata.mat','aar')

h2 = molviewer(aar);

evalrasmolscript(h2, ['restrict protein; color chain; ' ...

'spacefill off; wireframe off; ' ...
'cartoon on; select 76:A, 48:B; spacefill; ' ...
... % isopeptide bond
'select 76:A and *.ca; ' ... % select alpha carbon

3-156
 Visualizing the Three-Dimensional Structure of a Molecule

'set labeloffset 40 10; label isopeptide bond; ' ...

'move 0 360 0 -20 0 0 0 0 5; ']); % animate

Aligning Ubiquitin and SUMO Sequences

There is a surprisingly diverse family of ubiquitin-like proteins that display significant structural
similarity to ubiquitin. One of these proteins is SUMO (Small Ubiquitin-like MOdifier), a small protein
involved in a wide spectrum of post-translational modifications, such as transcriptional regulation,
nuclear-cytosolic transport, and protein stability. Similar to ubiquitination, the covalent attachment
and detachment of SUMO occur via a cascade of enzymatic actions. Despite the structural and
operational similarities between ubiquitin and SUMO, these two proteins display quite limited
sequence similarity, as can be seen from their global sequence alignment.

Retrieve the protein SUMO from PDB or load the data from the MAT-file.

aar = getpdb('lwm2');

load('ubilikedata.mat','sumo')

Align the two primary sequences from both compounds.

[score aln] = nwalign(ubi.Sequence.Sequence, sumo.Sequence.Sequence)

score = -3.3333

aln = 3x82 char array

'MQ----I-F-VKTLTGKTITLEVEPSDTIENVKAKIQDKEGIPPDQQRLIFAGKQLEDGRTLSDYNIQKESTLHLVLRLRGG'
' : | : | |::: :::: : :: :::|: | |: | |: ::: | :: ::: |:|: |:: : '
'TENNDHINLKVAGQDGSVVQFKIKRHTPLSKLMKAYCERQGLSMRQIRFRFDGQPINETDTPAQLEMEDEDTID-VFQ-Q--'

Superposing the Structures of Ubiquitin and SUMO

In order to better appreciate the structural similarity between ubiquitin and SUMO, perform a three-
dimensional superposition of the two structures. Using the pdbsuperpose function, we compute and
apply a linear transformation (translation, reflection, orthogonal rotation, and scaling) such that the
atoms of one structure best conform to the atoms of the other structure.

close (h1, h2); % close previous instances of molviewer

pdbsuperpose(ubi, sumo);

3-157
3 Sequence Analysis

h3 = findobj('Tag', 'BioinfoMolviewer'); % retrieve handle for molviewer

evalrasmolscript(h3, ['select all; zoom 200; center selected']);

evalrasmolscript(h3, ['select all; cartoons off; ' ...

'select model = 1; strands on; color red; ' ...% ubiquitin
'select model = 2; strands on; color blue;']); % SUMO

By selecting the appropriate option button in the Models section of the Molecule Viewer window, we
can view the ubiquitin structure (Model = 1) and the SUMO-2 structure (Model = 2) separately or we
can look at them superposed (Model = All). When both models are actively displayed, the structural
similarity between the two folds is striking.

3-158
Visualizing the Three-Dimensional Structure of a Molecule

3-159
3 Sequence Analysis

3-160
 Visualizing the Three-Dimensional Structure of a Molecule

The conservation of the structural fold in the absence of a significant sequence similarity could point
to the occurrence of convergent evolution for these two proteins. However, some of the mechanisms
in ubiquitination and sumoylation have analogies that are not fold-related and could suggest some
deeper, perhaps distant, relationship. More importantly, the fact that the spectrum of functions

3-161
3 Sequence Analysis

performed by ubiquitin and SUMO-2 is so widespread, suggests that the high stability and
compactness of the ubiquitin-like superfold might be the reason behind its conservation.

close all;

3-162
 Calculating and Visualizing Sequence Statistics

Calculating and Visualizing Sequence Statistics

This example shows how to use basic sequence manipulation techniques and computes some useful
sequence statistics. It also illustrates how to look for coding regions (such as proteins) and pursue
further analysis of them.

The Human Mitochondrial Genome

In this example you will explore the DNA sequence of the human mitochondria. Mitochondria are
structures, called organelles, that are found in the cytoplasm of the cell in hundreds to thousands for
each cell. Mitochondria are generally the major energy production center in eukaryotes, they help to
degrade fats and sugars.

The consensus sequence of the human mitochondria genome has accession number NC_012920. You
can getgenbank function to get the latest annotated sequence from GenBank® into the MATLAB®
workspace.

mitochondria_gbk = getgenbank('NC_012920');

For your convenience, previously downloaded sequence is included in a MAT-file. Note that data in
public repositories is frequently curated and updated; therefore the results of this example might be
slightly different when you use up-to-date datasets.

load mitochondria

Copy just the DNA sequence to a new variable mitochondria. You can access parts of the DNA
sequence by using regular MATLAB indexing commands.

mitochondria = mitochondria_gbk.Sequence;
mitochondria_length = length(mitochondria)
first_300_bases = seqdisp(mitochondria(1:300))

mitochondria_length =

16569

first_300_bases =

5×70 char array

' 1 GATCACAGGT CTATCACCCT ATTAACCACT CACGGGAGCT CTCCATGCAT TTGGTATTTT'

' 61 CGTCTGGGGG GTATGCACGC GATAGCATTG CGAGACGCTG GAGCCGGAGC ACCCTATGTC'
'121 GCAGTATCTG TCTTTGATTC CTGCCTCATC CTATTATTTA TCGCACCTAC GTTCAATATT'
'181 ACAGGCGAAC ATACTTACTA AAGTGTGTTA ATTAATTAAT GCTTGTAGGA CATAATAATA'
'241 ACAATTGAAT GTCTGCACAG CCACTTTCCA CACAGACATC ATAACAAAAA ATTTCCACCA'

You can look at the composition of the nucleotides with the ntdensity function.

figure
ntdensity(mitochondria)

3-163
3 Sequence Analysis

This shows that the mitochondria genome is A-T rich. The GC-content is sometimes used to classify
organisms in taxonomy, it may vary between different species from ~30% up to ~70%. Measuring GC
content is also useful for identifying genes and for estimating the annealing temperature of DNA
sequence.

Calculating Sequence Statistics

Now, you will use some of the sequence statistics functions in the Bioinformatics Toolbox™ to look at
various properties of the human mitochondrial genome. You can count the number of bases of the
whole sequence using the basecount function.
bases = basecount(mitochondria)

bases =

struct with fields:

A: 5124
C: 5181
G: 2169
T: 4094

These are on the 5'-3' strand. You can look at the reverse complement case using the
seqrcomplement function.
compBases = basecount(seqrcomplement(mitochondria))

3-164
 Calculating and Visualizing Sequence Statistics

compBases =

struct with fields:

A: 4094
C: 2169
G: 5181
T: 5124

As expected, the base counts on the reverse complement strand are complementary to the counts on
the 5'-3' strand.

You can use the chart option to basecount to display a pie chart of the distribution of the bases.

figure
basecount(mitochondria,'chart','pie');
title('Distribution of Nucleotide Bases for Human Mitochondrial Genome');

Now look at the dimers in the sequence and display the information in a bar chart using
dimercount.

figure
dimers = dimercount(mitochondria,'chart','bar')
title('Mitochondrial Genome Dimer Histogram');

3-165
3 Sequence Analysis

dimers =

struct with fields:

AA: 1604
AC: 1495
AG: 795
AT: 1230
CA: 1534
CC: 1771
CG: 435
CT: 1440
GA: 613
GC: 711
GG: 425
GT: 419
TA: 1373
TC: 1204
TG: 513
TT: 1004

3-166
 Calculating and Visualizing Sequence Statistics

Exploring the Open Reading Frames (ORFs)

In a nucleotide sequence an obvious thing to look for is if there are any open reading frames. An ORF
is any sequence of DNA or RNA that can be potentially translated into a protein. The function
seqshoworfs can be used to visualize ORFs in a sequence.

Note: In the HTML tutorial only the first page of the output is shown, however when running the
example you will be able to inspect the complete mitochondrial genome using the scrollbar on the
figure.

seqshoworfs(mitochondria);

If you compare this output to the genes shown on the NCBI page there seem to be slightly fewer
ORFs, and hence fewer genes, than expected.

Vertebrate mitochondria do not use the Standard genetic code so some codons have different
meaning in mitochondrial genomes. For more information about using different genetic codes in
MATLAB see the help for the function geneticcode. The GeneticCode option to the seqshoworfs
function allows you to look at the ORFs again but this time with the vertebrate mitochondrial genetic
code.

3-167
3 Sequence Analysis

In the human mitochondrial DNA sequence some genes are also started by alternative start codons
[1]. Use the AlternativeStartCodons option to the seqshoworfs function to search also for
these ORFs.

Notice that there are now two much larger ORFs on the third reading frame: One starting at position
4470 and the other starting at 5904. These correspond to the ND2 (NADH dehydrogenase subunit 2)
and COX1 (cytochrome c oxidase subunit I) genes.
orfs = seqshoworfs(mitochondria,'GeneticCode','Vertebrate Mitochondrial',...
'AlternativeStartCodons',true)

orfs =

1×3 struct array with fields:

Start
Stop

Inspecting Annotated Features

You can also look at all the features that have been annotated to the human mitochondrial genome.
Explore the complete GenBank entry mitochondria_gbk with the featureparse function.

3-168
 Calculating and Visualizing Sequence Statistics

Particularly, you can explore the annotated coding sequences (CDS) and compare them with the ORFs
previously found. Use the Sequence option to the featureparse function to extract, when possible,
the DNA sequences respective to each feature. The featureparse function will complement the
pieces of the source sequence when appropriate.

features = featureparse(mitochondria_gbk,'Sequence',true)
coding_sequences = features.CDS;
coding_sequences_id = sprintf('%s ',coding_sequences.gene)

features =

struct with fields:

source: [1×1 struct]

D_loop: [1×1 struct]
gene: [1×37 struct]
tRNA: [1×22 struct]
rRNA: [1×2 struct]
STS: [1×28 struct]
misc_feature: [1×1 struct]
CDS: [1×13 struct]

coding_sequences_id =

'ND1 ND2 COX1 COX2 ATP8 ATP6 COX3 ND3 ND4L ND4 ND5 ND6 CYTB '

ND2CDS = coding_sequences(2) % ND2 is in the 2nd position

COX1CDS = coding_sequences(3) % COX1 is in the 3rd position

ND2CDS =

struct with fields:

Location: '4470..5511'
Indices: [4470 5511]
gene: 'ND2'
gene_synonym: 'MTND2'
note: 'TAA stop codon is completed by the addition of 3' A residues to the mRNA'
codon_start: '1'
transl_except: '(pos:5511,aa:TERM)'
transl_table: '2'
product: 'NADH dehydrogenase subunit 2'
protein_id: 'YP_003024027.1'
db_xref: {'GI:251831108' 'GeneID:4536' 'HGNC:7456' 'MIM:516001'}
translation: 'MNPLAQPVIYSTIFAGTLITALSSHWFFTWVGLEMNMLAFIPVLTKKMNPRSTEAAIKYFLTQATASMILLMAILFN
Sequence: 'attaatcccctggcccaacccgtcatctactctaccatctttgcaggcacactcatcacagcgctaagctcgcactg

COX1CDS =

struct with fields:

Location: '5904..7445'
Indices: [5904 7445]

3-169
3 Sequence Analysis

gene: 'COX1'
gene_synonym: 'COI; MTCO1'
note: 'cytochrome c oxidase I'
codon_start: '1'
transl_except: []
transl_table: '2'
product: 'cytochrome c oxidase subunit I'
protein_id: 'YP_003024028.1'
db_xref: {'GI:251831109' 'GeneID:4512' 'HGNC:7419' 'MIM:516030'}
translation: 'MFADRWLFSTNHKDIGTLYLLFGAWAGVLGTALSLLIRAELGQPGNLLGNDHIYNVIVTAHAFVMIFFMVMPIMIGG
Sequence: 'atgttcgccgaccgttgactattctctacaaaccacaaagacattggaacactatacctattattcggcgcatgagc

Create a map indicating all the features found in this GenBank entry using the featureview
function.

[h,l] = featureview(mitochondria_gbk,{'CDS','tRNA','rRNA','D_loop'},...
[2 1 2 2 2],'Fontsize',9);
legend(h,l,'interpreter','none');
title('Homo sapiens mitochondrion, complete genome')

Extracting and Analyzing the ND2 and COX1 Proteins

You can translate the DNA sequences that code for the ND2 and COX1 proteins by using the nt2aa
function. Again the GeneticCode option must be used to specify the vertebrate mitochondrial
genetic code.

3-170
 Calculating and Visualizing Sequence Statistics

ND2 = nt2aa(ND2CDS,'GeneticCode','Vertebrate Mitochondrial');

disp(seqdisp(ND2))

1 MNPLAQPVIY STIFAGTLIT ALSSHWFFTW VGLEMNMLAF IPVLTKKMNP RSTEAAIKYF

61 LTQATASMIL LMAILFNNML SGQWTMTNTT NQYSSLMIMM AMAMKLGMAP FHFWVPEVTQ
121 GTPLTSGLLL LTWQKLAPIS IMYQISPSLN VSLLLTLSIL SIMAGSWGGL NQTQLRKILA
181 YSSITHMGWM MAVLPYNPNM TILNLTIYII LTTTAFLLLN LNSSTTTLLL SRTWNKLTWL
241 TPLIPSTLLS LGGLPPLTGF LPKWAIIEEF TKNNSLIIPT IMATITLLNL YFYLRLIYST
301 SITLLPMSNN VKMKWQFEHT KPTPFLPTLI ALTTLLLPIS PFMLMIL

COX1 = nt2aa(COX1CDS,'GeneticCode','Vertebrate Mitochondrial');

disp(seqdisp(COX1))

1 MFADRWLFST NHKDIGTLYL LFGAWAGVLG TALSLLIRAE LGQPGNLLGN DHIYNVIVTA

61 HAFVMIFFMV MPIMIGGFGN WLVPLMIGAP DMAFPRMNNM SFWLLPPSLL LLLASAMVEA
121 GAGTGWTVYP PLAGNYSHPG ASVDLTIFSL HLAGVSSILG AINFITTIIN MKPPAMTQYQ
181 TPLFVWSVLI TAVLLLLSLP VLAAGITMLL TDRNLNTTFF DPAGGGDPIL YQHLFWFFGH
241 PEVYILILPG FGMISHIVTY YSGKKEPFGY MGMVWAMMSI GFLGFIVWAH HMFTVGMDVD
301 TRAYFTSATM IIAIPTGVKV FSWLATLHGS NMKWSAAVLW ALGFIFLFTV GGLTGIVLAN
361 SSLDIVLHDT YYVVAHFHYV LSMGAVFAIM GGFIHWFPLF SGYTLDQTYA KIHFTIMFIG
421 VNLTFFPQHF LGLSGMPRRY SDYPDAYTTW NILSSVGSFI SLTAVMLMIF MIWEAFASKR
481 KVLMVEEPSM NLEWLYGCPP PYHTFEEPVY MKS*

You can get a more complete picture of the amino acid content with aacount.

figure
subplot(2,1,1)
ND2aaCount = aacount(ND2,'chart','bar');
title('Histogram of Amino Acid Count for the ND2 Protein');

subplot(2,1,2)
COX1aaCount = aacount(COX1,'chart','bar');
title('Histogram of Amino Acid Count for the COX1 Protein');

3-171
3 Sequence Analysis

Notice the high leucine, threonine and isoleucine content and also the lack of cysteine or aspartic
acid.

You can use the atomiccomp and molweight functions to calculate more properties about the ND2
protein.
ND2AtomicComp = atomiccomp(ND2)
ND2MolWeight = molweight(ND2)

ND2AtomicComp =

struct with fields:

C: 1818
H: 2882
N: 420
O: 471
S: 25

ND2MolWeight =

3.8960e+04

For further investigation of the properties of the ND2 protein, use proteinplot. This is a graphical
user interface (GUI) that allows you to easily create plots of various properties, such as

3-172
 Calculating and Visualizing Sequence Statistics

hydrophobicity, of a protein sequence. Click on the "Edit" menu to create new properties, to modify
existing property values, or, to adjust the smoothing parameters. Click on the "Help" menu in the GUI
for more information on how to use the tool.

proteinplot(ND2)

You can also programmatically create plots of various properties of the sequence using
proteinpropplot.

figure
proteinpropplot(ND2,'PropertyTitle','Parallel beta strand')

3-173
3 Sequence Analysis

Calculating the Codon Frequency using all the Genes in the Human Mitochondrial Genome

The codoncount function counts the number of occurrences of each codon in the sequence and
displays a formatted table of the result.

codoncount(ND2CDS)

AAA - 10 AAC - 14 AAG - 2 AAT - 6

ACA - 11 ACC - 24 ACG - 3 ACT - 5
AGA - 0 AGC - 4 AGG - 0 AGT - 1
ATA - 23 ATC - 24 ATG - 1 ATT - 8
CAA - 8 CAC - 3 CAG - 2 CAT - 1
CCA - 4 CCC - 12 CCG - 2 CCT - 5
CGA - 0 CGC - 3 CGG - 0 CGT - 1
CTA - 26 CTC - 18 CTG - 4 CTT - 7
GAA - 5 GAC - 0 GAG - 1 GAT - 0
GCA - 8 GCC - 7 GCG - 1 GCT - 4
GGA - 5 GGC - 7 GGG - 0 GGT - 1
GTA - 3 GTC - 2 GTG - 0 GTT - 3
TAA - 0 TAC - 8 TAG - 0 TAT - 2
TCA - 7 TCC - 11 TCG - 1 TCT - 4
TGA - 10 TGC - 0 TGG - 1 TGT - 0
TTA - 8 TTC - 7 TTG - 1 TTT - 8

Notice that in the ND2 gene there are more CTA, ATC and ACC codons than others. You can check
what amino acids these codons get translated into using the nt2aa and aminolookup functions.

3-174
 Calculating and Visualizing Sequence Statistics

CTA_aa = aminolookup('code',nt2aa('CTA'))
ATC_aa = aminolookup('code',nt2aa('ATC'))
ACC_aa = aminolookup('code',nt2aa('ACC'))

CTA_aa =

'Leu Leucine
'

ATC_aa =

'Ile Isoleucine
'

ACC_aa =

'Thr Threonine
'

To calculate the codon frequency for all the genes you can concatenate them into a single sequence
before using the function codoncount. You need to ensure that the codons are complete (three
nucleotides each) so the read frame of the sequence is not lost at the concatenation.
numCDS = numel(coding_sequences);
CDS = cell(numCDS,1);
for i = 1:numCDS
seq = coding_sequences(i).Sequence;
CDS{i} = seq(1:3*floor(length(seq)/3));
end
allCDS = [CDS{:}];
codoncount(allCDS)

AAA - 85 AAC - 132 AAG - 10 AAT - 32

ACA - 134 ACC - 155 ACG - 10 ACT - 52
AGA - 1 AGC - 39 AGG - 1 AGT - 14
ATA - 167 ATC - 196 ATG - 40 ATT - 124
CAA - 82 CAC - 79 CAG - 8 CAT - 18
CCA - 52 CCC - 119 CCG - 7 CCT - 41
CGA - 28 CGC - 26 CGG - 2 CGT - 7
CTA - 276 CTC - 167 CTG - 45 CTT - 65
GAA - 64 GAC - 51 GAG - 24 GAT - 15
GCA - 80 GCC - 124 GCG - 8 GCT - 43
GGA - 67 GGC - 87 GGG - 34 GGT - 24
GTA - 70 GTC - 48 GTG - 18 GTT - 31
TAA - 3 TAC - 89 TAG - 2 TAT - 46
TCA - 83 TCC - 99 TCG - 7 TCT - 32
TGA - 93 TGC - 17 TGG - 11 TGT - 5
TTA - 73 TTC - 139 TTG - 18 TTT - 77

Use the figure option to the codoncount function to show a heat map with the codon frequency.
Use the geneticcode option to overlay a grid on the figure that groups the synonymous codons
according with the Vertebrate Mitochondrial genetic code. Observe the particular bias of Leucine
(codons 'CTN').

3-175
3 Sequence Analysis

figure
count = codoncount(allCDS,'figure',true,'geneticcode','Vertebrate Mitochondrial');
title('Human Mitochondrial Genome Codon Frequency')

close all

References

[1] Barrell, B.G., Bankier, A.T. and Drouin, J., "A different genetic code in human mitochondria",
Nature, 282(5735):189-94, 1979.

3-176
 Aligning Pairs of Sequences

Aligning Pairs of Sequences

This example shows how to extract some sequences from GenBank®, find open reading frames
(ORFs), and then align the sequences using global and local alignment algorithms.

Accessing NCBI Data from the MATLAB® Workspace

One of the many fascinating sections of the NCBI web site is the Genes and diseases section. This
section provides a comprehensive introduction to medical genetics.

In this example you will be looking at genes associated with Tay-Sachs Disease. Tay-Sachs is an
autosomal recessive disease caused by mutations in both alleles of a gene (HEXA, which codes for the
alpha subunit of hexosaminidase A) on chromosome 15.

The NCBI reference sequence for HEXA has accession number NM_000520. You can use the
getgenbank function to retrieve the sequence information from the NCBI data repository and load it
into MATLAB®.

humanHEXA = getgenbank('NM_000520');

By doing a BLAST search or by searching in the mouse genome you can find an orthogonal gene,
AK080777 is the accession number for a mouse hexosaminidase A gene.

mouseHEXA = getgenbank('AK080777');

For your convenience, previously downloaded sequences are included in a MAT-file. Note that data in
public repositories is frequently curated and updated; therefore the results of this example might be
slightly different when you use up-to-date datasets.

load('hexosaminidase.mat','humanHEXA','mouseHEXA')

Exploring the Open Reading Frames (ORFs)

You can use the function seqshoworfs to look for ORFs in the sequence for the human HEXA gene.
Notice that the longest ORF is on the first reading frame. The output value in the variable
humanORFs is a structure giving the position of the start and stop codons for all the ORFs on each
reading frame.

humanORFs = seqshoworfs(humanHEXA.Sequence)

3-177
3 Sequence Analysis

humanORFs=1×3 struct array with fields:

Start
Stop

Now look at the ORFs in the mouse HEXA gene. In this case the ORF is also on the first frame.

mouseORFs = seqshoworfs(mouseHEXA.Sequence)

3-178
 Aligning Pairs of Sequences

mouseORFs=1×3 struct array with fields:

Start
Stop

Aligning the Sequences

The first step is to use global sequence alignment to look for similarities between these sequences.
You could look at the alignment between the nucleotide sequences, but it is generally more
instructive to look at the alignment between the protein sequences, in this example we know that the
sequences are coding sequences. Use the nt2aa function to convert the nucleotide sequences into
the corresponding amino acid sequences. Observe that the HEXA gene occurs in the first frame for
both sequences, otherwise you should use the input argument Frame to specify an alternative coding
frame.

humanProtein = nt2aa(humanHEXA.Sequence);
mouseProtein = nt2aa(mouseHEXA.Sequence);

One of the easiest ways to look for similarity between sequences is with a dot plot.

seqdotplot(mouseProtein,humanProtein)

Warning: Match matrix has more points than available screen pixels.
Scaling image by factors of 1 in X and 2 in Y.

3-179
3 Sequence Analysis

xlabel('Human hexosaminidase A');ylabel('Mouse hexosaminidase A');

With the default settings, the dot plot is a little difficult to interpret, so you can try a slightly more
stringent dot plot.
seqdotplot(mouseProtein,humanProtein,4,3)

Warning: Match matrix has more points than available screen pixels.
Scaling image by factors of 1 in X and 2 in Y.

xlabel('Human hexosaminidase A');ylabel('Mouse hexosaminidase A');

The diagonal line indicates that there is probably a good alignment so you can now take a look at the
global alignment using the function nwalign which uses the Needleman-Wunsch algorithm.
[score, globalAlignment] = nwalign(humanProtein,mouseProtein)

score = 634.3333

globalAlignment = 3x812 char array

'SCRRPAQSAARSRSLRSRPEVKGQGVGPPGVAGAEPPLVT*FADKSRGRRSPDQGLTWPAPSERGDQRAMTSSRLWFSLLLAAAFAGRATAL

3-180
 Aligning Pairs of Sequences

' || |: | | | | | ||:: ||| |||||||:| ||||

'--------AA------------GR--------G---------A----G-R-------W----------AMAGCRLWVSLLLAAALACLATAL

Refining With Semi-global Alignment

The alignment is very good except for the terminal segments. For instance, notice the sparse matched
pairs in the first positions. This occurs because a global alignment attempts to force the matching all
the way to the ends and there is point where the penalty for opening new gaps is comparable to the
score of matching residues. In some cases it is desirable to remove the gap penalty added at the ends
of a global alignment; this allows you to better match this pair of sequences. This technique is
commonly known as 'semi-global' alignment or 'glocal' alignment.

[score, globalAlignment] = nwalign(humanProtein,mouseProtein,'glocal',true)

score = 1.0413e+03

globalAlignment = 3x825 char array

'SCRRPAQSAARSRSLRSRPEVKGQGVGPPGVAGAEPPLVT*FADKSRGRRSPDQGLTWPAPSERGDQR-AMTSSRLWFSLLLAAAFAGRATA
' : || | ||:: ||| |||||||:| |||
'------------------------------------------------------------AAGRGAGRWAMAGCRLWVSLLLAAALACLATA

Refining the Alignment by Extracting the Protein Sequence

Another way to refine your alignment is by using only the protein sequences. Notice that the aligned
region is delimited by start ( M-methionine ) and stop ( * ) amino acids in the sequences. If the
sequence is shortened so that only the translated regions are considered, then it seems likely that you
will get a better alignment. Use the find command to look for the index of the start amino acid in
each sequence:

humanStart = find(humanProtein == 'M',1)

humanStart = 70

mouseStart = find(mouseProtein == 'M',1)

mouseStart = 11

Similarly, use the find command to look for the index of the first stop occurring after the start of the
translation. Special care needs to be taken because there is also a stop at the very beginning of the
humanProtein sequence.

humanStop = find(humanProtein(humanStart:end)=='*',1) + humanStart - 1

humanStop = 599

mouseStop = find(mouseProtein(mouseStart:end)=='*',1) + mouseStart - 1

mouseStop = 539

Use these indices to truncate the sequences.

humanSeq = humanProtein(humanStart:humanStop);
humanSeqFormatted = seqdisp(humanSeq)

humanSeqFormatted = 9x70 char array

' 1 MTSSRLWFSL LLAAAFAGRA TALWPWPQNF QTSDQRYVLY PNNFQFQYDV SSAAQPGCSV'
' 61 LDEAFQRYRD LLFGSGSWPR PYLTGKRHTL EKNVLVVSVV TPGCNQLPTL ESVENYTLTI'

3-181
3 Sequence Analysis

'121 NDDQCLLLSE TVWGALRGLE TFSQLVWKSA EGTFFINKTE IEDFPRFPHR GLLLDTSRHY'

'181 LPLSSILDTL DVMAYNKLNV FHWHLVDDPS FPYESFTFPE LMRKGSYNPV THIYTAQDVK'
'241 EVIEYARLRG IRVLAEFDTP GHTLSWGPGI PGLLTPCYSG SEPSGTFGPV NPSLNNTYEF'
'301 MSTFFLEVSS VFPDFYLHLG GDEVDFTCWK SNPEIQDFMR KKGFGEDFKQ LESFYIQTLL'
'361 DIVSSYGKGY VVWQEVFDNK VKIQPDTIIQ VWREDIPVNY MKELELVTKA GFRALLSAPW'
'421 YLNRISYGPD WKDFYIVEPL AFEGTPEQKA LVIGGEACMW GEYVDNTNLV PRLWPRAGAV'
'481 AERLWSNKLT SDLTFAYERL SHFRCELLRR GVQAQPLNVG FCEQEFEQT* '

mouseSeq = mouseProtein(mouseStart:mouseStop);
mouseSeqFormatted = seqdisp(mouseSeq)

mouseSeqFormatted = 9x70 char array

' 1 MAGCRLWVSL LLAAALACLA TALWPWPQYI QTYHRRYTLY PNNFQFRYHV SSAAQAGCVV'
' 61 LDEAFRRYRN LLFGSGSWPR PSFSNKQQTL GKNILVVSVV TAECNEFPNL ESVENYTLTI'
'121 NDDQCLLASE TVWGALRGLE TFSQLVWKSA EGTFFINKTK IKDFPRFPHR GVLLDTSRHY'
'181 LPLSSILDTL DVMAYNKFNV FHWHLVDDSS FPYESFTFPE LTRKGSFNPV THIYTAQDVK'
'241 EVIEYARLRG IRVLAEFDTP GHTLSWGPGA PGLLTPCYSG SHLSGTFGPV NPSLNSTYDF'
'301 MSTLFLEISS VFPDFYLHLG GDEVDFTCWK SNPNIQAFMK KKGFTDFKQL ESFYIQTLLD'
'361 IVSDYDKGYV VWQEVFDNKV KVRPDTIIQV WREEMPVEYM LEMQDITRAG FRALLSAPWY'
'421 LNRVKYGPDW KDMYKVEPLA FHGTPEQKAL VIGGEACMWG EYVDSTNLVP RLWPRAGAVA'
'481 ERLWSSNLTT NIDFAFKRLS HFRCELVRRG IQAQPISVGC CEQEFEQT* '

Align these two sequences.

[score, alignment] = nwalign(humanSeq,mouseSeq)

score = 1.0423e+03

alignment = 3x530 char array

'MTSSRLWFSLLLAAAFAGRATALWPWPQNFQTSDQRYVLYPNNFQFQYDVSSAAQPGCSVLDEAFQRYRDLLFGSGSWPRPYLTGKRHTLEK
'|:: ||| |||||||:| ||||||||| :|| :||:||||||||:| |||||| || ||||||:|||:||||||||||| :::|::|| |
'MAGCRLWVSLLLAAALACLATALWPWPQYIQTYHRRYTLYPNNFQFRYHVSSAAQAGCVVLDEAFRRYRNLLFGSGSWPRPSFSNKQQTLGK

Open reading frame information is also available from the output of the seqshoworfs command, but
the indices are based on the nucleotide sequences. Use these indices to trim the original nucleotide
sequences and then translate them to amino acids.

humanPORF = nt2aa(humanHEXA.Sequence(humanORFs(1).Start(1):humanORFs(1).Stop(1)));
mousePORF = nt2aa(mouseHEXA.Sequence(mouseORFs(1).Start(1):mouseORFs(1).Stop(1)));
[score, ORFAlignment] = nwalign(humanPORF,mousePORF)

score = 1042

ORFAlignment = 3x529 char array

'MTSSRLWFSLLLAAAFAGRATALWPWPQNFQTSDQRYVLYPNNFQFQYDVSSAAQPGCSVLDEAFQRYRDLLFGSGSWPRPYLTGKRHTLEK
'|:: ||| |||||||:| ||||||||| :|| :||:||||||||:| |||||| || ||||||:|||:||||||||||| :::|::|| |
'MAGCRLWVSLLLAAALACLATALWPWPQYIQTYHRRYTLYPNNFQFRYHVSSAAQAGCVVLDEAFRRYRNLLFGSGSWPRPSFSNKQQTLGK

Alternatively, you can use the coding region information (CDS) from the GenBank data structure to
find the coding region of the genes.

idx = humanHEXA.CDS.indices;
humanCodingRegion = humanHEXA.Sequence(idx(1):idx(2));
idx = mouseHEXA.CDS.indices;
mouseCodingRegion = mouseHEXA.Sequence(idx(1):idx(2));

3-182
 Aligning Pairs of Sequences

You can also get the translation of the coding regions from this structure.

humanTranslatedRegion = humanHEXA.CDS.translation;
mouseTranslatedRegion = mouseHEXA.CDS.translation;

Local Alignment

Instead of truncating the sequences to look for better alignment, an alternative approach is to use a
local alignment. The function swalign performs local alignment using the Smith-Waterman
algorithm. This shows a very good alignment for the whole coding region and reasonable similarity
for a few residues beyond at both the ends of the gene.

[score, localAlignment] = swalign(humanProtein,mouseProtein)

score = 1057

localAlignment = 3x547 char array

'RGDQR-AMTSSRLWFSLLLAAAFAGRATALWPWPQNFQTSDQRYVLYPNNFQFQYDVSSAAQPGCSVLDEAFQRYRDLLFGSGSWPRPYLTG
'|| | ||:: ||| |||||||:| ||||||||| :|| :||:||||||||:| |||||| || ||||||:|||:||||||||||| :::
'RGAGRWAMAGCRLWVSLLLAAALACLATALWPWPQYIQTYHRRYTLYPNNFQFRYHVSSAAQAGCVVLDEAFRRYRNLLFGSGSWPRPSFSN

Alignment of Complementary DNA Sequences

All the sequence alignment functions provided in MATLAB can be customized. For example, by
modifying the rows and columns of a scoring matrix you can align sequences by complement and not
by identity. In this case you can reorder the NUC44 scoring matrix; a positive score is given for
complements while a negative score is given otherwise. The first 30 nucleotides from the mouse
HEXA gene will be aligned to its complement.

[M, info] = nuc44;

map = nt2int(seqcomplement(info.Order))

map = 1x15 uint8 row vector

4 3 2 1 6 5 8 7 9 10 14 13 12 11 15

Mc = M(:,map)

Mc = 15×15

-4 -4 -4 5 -4 1 1 -4 -4 1 -1 -1 -1 -4 -2
-4 -4 5 -4 1 -4 1 -4 1 -4 -1 -1 -4 -1 -2
-4 5 -4 -4 -4 1 -4 1 1 -4 -1 -4 -1 -1 -2
5 -4 -4 -4 1 -4 -4 1 -4 1 -4 -1 -1 -1 -2
-4 1 -4 1 -4 -1 -2 -2 -2 -2 -1 -3 -1 -3 -1
1 -4 1 -4 -1 -4 -2 -2 -2 -2 -3 -1 -3 -1 -1
1 1 -4 -4 -2 -2 -4 -1 -2 -2 -3 -3 -1 -1 -1
-4 -4 1 1 -2 -2 -1 -4 -2 -2 -1 -1 -3 -3 -1
-4 1 1 -4 -2 -2 -2 -2 -1 -4 -1 -3 -3 -1 -1
1 -4 -4 1 -2 -2 -2 -2 -4 -1 -3 -1 -1 -3 -1
⋮

[score, compAlignment] = nwalign(mouseHEXA.Sequence(1:30), ...

seqcomplement(mouseHEXA.Sequence(1:30)), 'SCORINGMATRIX', ...
Mc, 'ALPHABET', 'NT')

3-183
3 Sequence Analysis

score = 150

compAlignment = 3x30 char array

'GCTGCTGGAAGGGGAGCTGGCCGGTGGGCC'
'::::::::::::::::::::::::::::::'
'CGACGACCTTCCCCTCGACCGGCCACCCGG'

close all;

3-184
 Assessing the Significance of an Alignment

Assessing the Significance of an Alignment

This example shows a method that can be used to investigate the significance of sequence
alignments. The number of identities or positives in an alignment is not a clear indicator of a
significant alignment. A permutation of a sequence from an alignment will have similar percentages
of positives and identities when aligned against the original sequence. The score from an alignment is
a better indicator of the significance of an alignment. This example uses the same Tay-Sachs disease
related genes and proteins analyzed in “Aligning Pairs of Sequences” on page 3-177.

Accessing NCBI Data from the MATLAB® Workspace

In this example, you will work directly with protein data so use getgenpept instead of getgenbank
to download the data from the NCBI site. First read the human protein information into MATLAB®.
humanProtein = getgenpept('NP_000511');

Results from a BLASTX search performed with this sequence showed that a Drosophila protein,
GenPept accession number AAM29423, has some similarity to the human HEXA sequence. Use
getgenpept to download this sequence.
flyProtein = getgenpept('AAM29423');

For your convenience, previously downloaded sequences are included in a MAT-file. Note that data in
public repositories is frequently curated and updated; therefore the results of this example might be
slightly different when you use up-to-date datasets.
load('flyandhumanproteins.mat','humanProtein','flyProtein')
seqdisp(humanProtein)
seqdisp(flyProtein)

ans =

10x70 char array

'>gi|189181666|gb|NP_000511.2| beta-hexosaminidase subunit alpha pre...'

' 1 MTSSRLWFSL LLAAAFAGRA TALWPWPQNF QTSDQRYVLY PNNFQFQYDV SSAAQPGCSV'
' 61 LDEAFQRYRD LLFGSGSWPR PYLTGKRHTL EKNVLVVSVV TPGCNQLPTL ESVENYTLTI'
'121 NDDQCLLLSE TVWGALRGLE TFSQLVWKSA EGTFFINKTE IEDFPRFPHR GLLLDTSRHY'
'181 LPLSSILDTL DVMAYNKLNV FHWHLVDDPS FPYESFTFPE LMRKGSYNPV THIYTAQDVK'
'241 EVIEYARLRG IRVLAEFDTP GHTLSWGPGI PGLLTPCYSG SEPSGTFGPV NPSLNNTYEF'
'301 MSTFFLEVSS VFPDFYLHLG GDEVDFTCWK SNPEIQDFMR KKGFGEDFKQ LESFYIQTLL'
'361 DIVSSYGKGY VVWQEVFDNK VKIQPDTIIQ VWREDIPVNY MKELELVTKA GFRALLSAPW'
'421 YLNRISYGPD WKDFYIVEPL AFEGTPEQKA LVIGGEACMW GEYVDNTNLV PRLWPRAGAV'
'481 AERLWSNKLT SDLTFAYERL SHFRCELLRR GVQAQPLNVG FCEQEFEQT '

ans =

12x70 char array

'>gi|21064387|gb|AAM29423.1| RE17456p [Drosophila melanogaster]. '

' 1 MSLAVSLRRA LLVLLTGAIF ILTVLYWNQG VTKAQAYNEA LERPHSHHDA SGFPIPVEKS'
' 61 WTYKCENDRC MRVGHHGKSA KRVSFISCSM TCGDVNIWPH PTQKFLLSSQ THSFSVEDVQ'
'121 LHVDTAHREV RKQLQLAFDW FLKDLRLIQR LDYGGSSSEP TVSESSSKSR HHADLEPAAT'
'181 LFGATFGVKK AGDLTSVQVK ISVLKSGDLN FSLDNDETYQ LSTQTEGHRL QVEIIANSYF'

3-185
3 Sequence Analysis

'241 GARHGLSTLQ QLIWFDDEDH LLHTYANSKV KDAPKFRYRG LMLDTSRHFF SVESIKRTIV'

'301 GMGLAKMNRF HWHLTDAQSF PYISRYYPEL AVHGAYSESE TYSEQDVREV AEFAKIYGVQ'
'361 VIPEIDAPAH AGNGWDWGPK RGMGELAMCI NQQPWSFYCG EPPCGQLNPK NNYTYLILQR'
'421 IYEELLQHTG PTDFFHLGGD EVNLDCWAQY FNDTDLRGLW CDFMLQAMAR LKLANNGVAP'
'481 KHVAVWSSAL TNTKRLPNSQ FTVQVWGGST WQENYDLLDN GYNVIFSHVD AWYLDCGFGS'
'541 WRATGDAACA QYRTWQNVYK HRPWERMRLD KKRKKQVLGG EVCMWTEQVD ENQLDNRLWP'
'601 RTAALAERLW TDPSDDHDMD IVPPDVFRRI SLFRNRLVEL GIRAEALFPK YCAQNPGECI'

A First Comparison and Global Alignment

The first thing to do is to use seqdotplot to see if there are any areas that are clearly aligned. This
doesn't show any obvious alignments, but there are some areas of interest.

seqdotplot(humanProtein,flyProtein,3,2)
title('Dot Plot of Two HexA-like Proteins');
ylabel('Human Protein');xlabel('Drosophila Protein');

3-186
 Assessing the Significance of an Alignment

Notice that there are a few diagonal stretches in the dot plot. This is not particularly good evidence of
a significant global alignment, but you can try a global alignment using the function nwalign. The
BLOSUM50 scoring matrix is used by default.

[sc50,globAlig50] = nwalign(humanProtein,flyProtein)
fprintf('Score = %g \n',sc50)

sc50 =

49.6667

globAlig50 =

3x670 char array

'MT-S-S--R----LW----F--SLL-----LA-A-AF--A-GR------ATAL-WP----W--P-QN---FQT-----SDQR--Y-------
'|: : | | | | ::| :: | |: | | |::: | | :| ::: | :| :
'MSLAVSLRRALLVLLTGAIFILTVLYWNQGVTKAQAYNEALERPHSHHDASGFPIPVEKSWTYKCENDRCMRVGHHGKSAKRVSFISCSMTC

Score = 49.6667

The sequence similarity is fairly low, so BLOSUM30 might be a more appropriate scoring matrix.

[sc30,globAlig30] = nwalign(humanProtein,flyProtein,'scoringmatrix','blosum30')
fprintf('Score = %g \n',sc30)

sc30 =

82

globAlig30 =

3x670 char array

'MT-S-S--R-----L-W--F--S-LL----L--AAAF--A-GR------ATAL-WP----W--P-QN-F--QT-----SDQR--Y-------
'|: : | | | : | : |: : |:|: | | |::: | | : :| :: |::| :
'MSLAVSLRRALLVLLTGAIFILTVLYWNQGVTKAQAYNEALERPHSHHDASGFPIPVEKSWTYKCENDRCMRVGHHGKSAKRVSFISCSMTC

Score = 82

This gives an alignment that has some areas of fairly strong similarity, but is this alignment
statistically significant? One way to investigate whether this score is significant is to use Monte Carlo
techniques. Given that the fly sequence was found using a BLAST search, there is some evidence that
there is similarity between the two sequences. It is reasonable to expect the score for this alignment
to be higher than the scores obtained from aligning random sequences of amino acids to the protein.

Assessing the Significance of the Score

To assess if the score is significant the first step is to make some random sequences that are similar
to that of the fly protein. One way to do this is to take random permutations of the fly sequence. This
can be done with the randperm function. Then calculate the global alignment of these random
sequences against the human protein and look at the statistical significance of the scores.

3-187
3 Sequence Analysis

Initialize the state of the default random number generators to ensure that the figures and results
generated match the ones in the HTML version of this example.
rng(0,'twister')
n = 50;
globalscores = zeros(n,1);
flyLen = length(flyProtein.Sequence);
for i = 1:n
perm = randperm(flyLen);
permutedSequence = flyProtein.Sequence(perm);
globalscores(i) = nwalign(humanProtein,permutedSequence,'scoringmatrix','blosum30');
end

Now plot the scores as a bar chart. Note that because you are using randomly generated sequences.
figure
buckets = ceil(n/5);
hist(globalscores,buckets)
hold on;
stem(sc30,1,'k')
title('Determining Alignment Significance using Monte Carlo Techniques');
xlabel('Score'); ylabel('Number of Sequences');

The scores of the alignments to the random sequences can be approximated by the type 1 extreme
value distribution. Use the evfit function from the Statistics and Machine Learning Toolbox™ to
estimate the parameters of this distribution.
parmhat = evfit(globalscores)

3-188
 Assessing the Significance of an Alignment

parmhat =

-31.7597 6.6440

Overlay a plot of the probability density function of the estimated distribution.

x = min(globalscores):max([globalscores;sc30]);
y = evpdf(x,parmhat(1),parmhat(2));
[v, c] = hist(globalscores,buckets);
binWidth = c(2) - c(1);
scaleFactor = n*binWidth;
plot(x,scaleFactor*y,'r');
hold off;

From this plot you can see that the global alignment (globAlig30) is clearly statistically significant.

An Example Where the Score is Not Statistically Significant

In FLYBASE web site you can search for all Drosophila beta-N-acetylhexosaminidase genes. The gene
that you have been looking at so far is referenced as CG8824. Now you want to take a look at another
similar gene, for instance Hexo1.

flyHexo1 = getgenpept('AAL28566');

The fly Hexo1 aminoacid sequence is also provided in the MAT-file flyandhumanproteins.mat.

3-189
3 Sequence Analysis

load('flyandhumanproteins.mat','flyHexo1')
seqdisp(humanProtein)

ans =

10x70 char array

'>gi|189181666|gb|NP_000511.2| beta-hexosaminidase subunit alpha pre...'

' 1 MTSSRLWFSL LLAAAFAGRA TALWPWPQNF QTSDQRYVLY PNNFQFQYDV SSAAQPGCSV'
' 61 LDEAFQRYRD LLFGSGSWPR PYLTGKRHTL EKNVLVVSVV TPGCNQLPTL ESVENYTLTI'
'121 NDDQCLLLSE TVWGALRGLE TFSQLVWKSA EGTFFINKTE IEDFPRFPHR GLLLDTSRHY'
'181 LPLSSILDTL DVMAYNKLNV FHWHLVDDPS FPYESFTFPE LMRKGSYNPV THIYTAQDVK'
'241 EVIEYARLRG IRVLAEFDTP GHTLSWGPGI PGLLTPCYSG SEPSGTFGPV NPSLNNTYEF'
'301 MSTFFLEVSS VFPDFYLHLG GDEVDFTCWK SNPEIQDFMR KKGFGEDFKQ LESFYIQTLL'
'361 DIVSSYGKGY VVWQEVFDNK VKIQPDTIIQ VWREDIPVNY MKELELVTKA GFRALLSAPW'
'421 YLNRISYGPD WKDFYIVEPL AFEGTPEQKA LVIGGEACMW GEYVDNTNLV PRLWPRAGAV'
'481 AERLWSNKLT SDLTFAYERL SHFRCELLRR GVQAQPLNVG FCEQEFEQT '

Repeat the process of generating a global alignment and then using random permutations of the
amino acids to estimate the significance of the global alignment.

[Hexo1score,Hexo1Alignment] = nwalign(humanProtein,flyHexo1,'scoringmatrix','blosum30')
fprintf('Score = %g \n',Hexo1score)
Hexo1globalscores = zeros(n,1);
flyLen = length(flyHexo1.Sequence);
for i = 1:n
perm = randperm(flyLen);
permutedSequence = flyHexo1.Sequence(perm);
Hexo1globalscores(i) = nwalign(humanProtein,permutedSequence,'scoringmatrix','blosum30');
end

Hexo1score =

-72.2000

Hexo1Alignment =

3x534 char array

'MTSSRL-WFSLLLAAAFA-GRATALWPWPQNFQTSDQRYVLYPNNFQFQYDVSSAAQPGCSVLDEAFQRYRDLLFGSGSWPRPYLTGKRHTL
'|: :| | :::: : ::: :| ::: : :|| | | : : : ||::::: | | :: |:: | : : | :: | :|
'MALVKLNTFHWHITDSHSFPLEVKKRPELHKLGAYSQRQV-Y--T-R-R-DVAEVVEYG-RV--RGI-RVMP-EF-D-A-PAHVGEGWQH--

Score = -72.2

Plot the scores, calculate the parameters of the distribution and overlay the PDF on the bar chart.

figure
buckets = ceil(n/5);
hist(Hexo1globalscores,buckets)
title('Determining Alignment Significance using Monte Carlo Techniques');
xlabel('Score');
ylabel('Number of Sequences');
hold on;

3-190
 Assessing the Significance of an Alignment

stem(Hexo1score,1,'c')
parmhat = evfit(Hexo1globalscores)
x = min(Hexo1globalscores):max([Hexo1globalscores;Hexo1score]);
y = evpdf(x,parmhat(1),parmhat(2));
[v, c] = hist(Hexo1globalscores,buckets);
binWidth = c(2) - c(1);
scaleFactor = n*binWidth;
plot(x,scaleFactor*y,'r');
hold off;

parmhat =

-70.6926 7.0619

In this case it appears that the alignment is not statistically significant. Higher scoring alignments
can easily be generated from a random permutation of the amino acids in the sequence. You can
calculate an approximate p-value from the estimated extreme value CDF: However, far more than 50
random permutations are needed to get a reliable estimate of the extreme value pdf parameters from
which to calculate a reasonably accurate p-value.

p = 1 - evcdf(Hexo1score,parmhat(1),parmhat(2))

p =

3-191
3 Sequence Analysis

0.4458

One thing to notice is that the lengths of the two sequences are very different. The human HEXA1 is
529 residues long and the fly Hexo1 protein is only 383 residues in length. When you try to align
these two sequences globally this difference in length means that a large number of gaps will have to
be introduced into the sequence. This means that the significance of the scores will be heavily
dependent on the GAPOPEN and EXTENDGP parameters. (See the help for nwalign for more details.)
Instead of using global alignment, in this case a better approach might be to look at the local
alignment between the two sequences.

Using Local Alignment and Randseq

You will now repeat the process of estimating the significance of an alignment this time using local
alignment and a slightly different method of generating the random sequences. Instead of simply
permuting the letters in the sequence, an alternative is to draw a sequence from a multinomial
distribution which is estimated from the fly protein sequence. You can do this using the aacount and
randseq functions; the first estimates the amino acid frequencies of the query sequence and the
later randomly creates new sequences based on this distribution.

[lscore,locAlig] = swalign(humanProtein,flyHexo1,'scoringmatrix','blosum30')
fprintf('Score = %g \n',lscore)

localscores = zeros(n,1);
aas = aacount(flyHexo1);
for i = 1:n
randProtein = randseq(flyLen,'FROMSTRUCTURE',aas);
localscores(i) = swalign(humanProtein,randProtein,'scoringmatrix','blosum30');
end

lscore =

152

locAlig =

3x361 char array

'MAYNKLNVFHWHLVDDPSFPYESFTFPELMRKGSYNPVTHIYTAQDVKEVIEYARLRGIRVLAEFDTPGHT-LSWG-PGIPGLL-TPCYSGS
'||: |||:||||::|::|||:| |||:: |:|:: ::|| :||:||:||:|:|||||: |||:|:|: :| ::::::: :: :::
'MALVKLNTFHWHITDSHSFPLEVKKRPELHKLGAYSQR-QVYTRRDVAEVVEYGRVRGIRVMPEFDAPAHVGEGWQHKNMTACFNAQPWKSF

Score = 152

Plot the scores, calculate the parameters of the distribution and overlay the PDF on the bar chart.

figure
hist(localscores,buckets)
title('Determining Alignment Significance using Monte Carlo Techniques');
xlabel('Score');
ylabel('Number of Sequences');
hold on;
stem(lscore,1,'r')
parmhat = evfit(localscores)
x = min(localscores):max([localscores;lscore]);

3-192
 Assessing the Significance of an Alignment

y = evpdf(x,parmhat(1),parmhat(2));
[v, c] = hist(localscores,buckets);
binWidth = c(2) - c(1);
scaleFactor = n*binWidth;
plot(x,scaleFactor*y,'r');
hold off;

parmhat =

40.8331 3.9312

You might like to experiment to see if there are significant differences in the distribution of scores
generated with randperm and randseq.

With the local alignment it appears that the alignment is statistically significant. In fact, looking at
the local alignment shows a very good alignment for the full length of the Hexo1 sequence.

close all;

3-193
3 Sequence Analysis

Using Scoring Matrices to Measure Evolutionary Distance

This example shows how to handle Scoring Matrices with the sequence alignment tools. The example
uses proteins associated with retinoblastoma, a disease caused by a tumor which develops from the
immature retina.

Accessing the NCBI Website and Database

More information on retinoblastoma can be found at the Genes and diseases section of the NCBI web
site.

The "BLink" link on this page shows related sequences in different organisms. These links can change
frequently, so for this example you can load a set of previously saved data from a MAT-file.
load retinoblastoma

You can also use the getgenpept function to retrieve the sequence information from the NCBI data
repository and load it into MATLAB.
human = getgenpept('AAA69808','SequenceOnly',true);
chicken = getgenpept('NP_989750','SequenceOnly',true);
trout = getgenpept('AAD13390','SequenceOnly',true);
xenopus = getgenpept('A44879','SequenceOnly',true);

Aligning CAH72243 (Human Protein) to CAA51019 (Chicken Protein)

One approach to study the relationship between these two proteins is to use a global alignment with
the nwalign function.
[sc,hvc] = nwalign(human,chicken)

sc = 1.4543e+03

hvc = 3x938 char array

'MPPKTPRKTAATAAAAAAEPPAPPPPPPPEEDPEQDSGPEDLPLVRLEFEETEEPDFTALCQKLKIPDHVRERAWLTWEKVSSVDGVLGGYI
'|||| | : |::| : : | : | | :| :||| |: | |:|||: || || |||:||:|:::::::||: ::|
'MPPK-PLRRAGAARSQRTSPEGGAGTASP---P---GG------TRLEVGEA-E--FVALCDALKAPDSVREKAWMTYQSLAAADGA-SAYN

In this alignment the function used the default scoring matrix, BLOSUM62. Different scoring matrices
can give different alignments. How can you find the best alignment? One approach is to try different
scoring matrices and look for the highest score. When the score from the alignment functions is in
the same scale (in this case, bits) you can compare different alignments to see which gives the
highest score.

This example uses the PAM family of matrices, though the approach used could also be used with the
BLOSUM family of scoring matrices. The PAM family of matrices in the Bioinformatics Toolbox™
consists of 50 matrices, PAM10, PAM20,..., PAM490, PAM500.

Take the two sequences (CAH72243 and CAA51019) and align them with each member of the PAM
family and then look for the highest score.
score = zeros(1,50);
fprintf('Trying different PAM matrices ')

Trying different PAM matrices

3-194
 Using Scoring Matrices to Measure Evolutionary Distance

for step = 1:50

fprintf('.')
PamNumber = step * 10;
[matrix,info] = pam(PamNumber);
score(step) = nwalign(human,chicken,'scoringmatrix',matrix,'scale',info.Scale);
end

..................................................

Plotting the Scores

You can use the plot function to create a graph of the results.

x = 10:10:500;
plot(x,score)
legend('Human vs. Chicken');
title('Global Alignment Scores for Different PAM Scoring Matrices');
xlabel('PAM matrix');ylabel('Score (bits)');

Finding the Best Score

You can use max with two outputs to find the highest score and the index in the results vector where
the highest value occurred. In this case the highest score occurred with the third matrix, that is
PAM30.

[bestScore, idx] = max(score)

bestScore = 2.2605e+03

3-195
3 Sequence Analysis

idx = 3

Aligning to Other Organisms

Repeat this with different organisms: xenopus and rainbow trout.

xenopusScore = zeros(1,50);
troutScore = zeros(1,50);
fprintf('Trying different PAM matrices ')

Trying different PAM matrices

for step = 1:50

fprintf('.')
PamNumber = step * 10;
[matrix,info] = pam(PamNumber);
xenopusScore(step) = nwalign(human,xenopus,'scoringmatrix',matrix,'scale',info.Scale);
troutScore(step) = nwalign(human,trout,'scoringmatrix',matrix,'scale',info.Scale);
end

..................................................

Adding More Lines to the Same Plot

You can use the command hold on to tell MATLAB® to add new plots to the existing figure. Once
you have finished doing this you must remember to disable this feature by using hold off.

hold on
plot(x,xenopusScore,'g')
plot(x,troutScore,'r')
legend({'Human vs. Chicken','Human vs. Xenopus','Human vs. Trout'});box on
title('Global Alignment Scores for Different PAM Scoring Matrices');
xlabel('PAM matrix');ylabel('Score (bits)');
hold off

3-196
 Using Scoring Matrices to Measure Evolutionary Distance

Finding the Best Scores

You will see that different matrices give the highest scores for the different organisms. For human
and xenopus, the best score is with PAM40 and for human and trout the best score is PAM50.

[bestXScore, Xidx] = max(xenopusScore)

bestXScore = 1607

Xidx = 4

[bestTScore, Tidx] = max(troutScore)

bestTScore = 1484

Tidx = 5

The PAM scoring matrix giving the best alignment for two sequences is an indicator of the relative
evolutionary interval since the organisms diverged: The smaller the PAM number, the more closely
related the organisms. Since organisms, and protein families across organisms, evolve at widely
varying rates, there is no simple correlation between PAM distance and evolutionary time. However,
for an analysis of a specific protein family across multiple species, the corresponding PAM matrices
will provide a relative evolutionary distance between the species and allow accurate phylogenetic
mapping. In this example, the results indicate that the human sequence is more closely related to the
chicken sequence than to the frog sequence, which in turn is more closely related than the trout
sequence.

3-197
3 Sequence Analysis

Calling Bioperl Functions from MATLAB

This example shows the interoperability between MATLAB® and Bioperl - passing arguments from
MATLAB to Perl scripts and pulling BLAST search data back to MATLAB.

NOTE: Perl and the Bioperl modules must be installed to run the Perl scripts in this example. Since
version 1.4, Bioperl modules have a warnings.pm dependency requiring at least version 5.6 of Perl. If
you have difficulty running the Perl scripts, make sure your PERL5LIB environment variable includes
the path to your Bioperl installation or try running from the Bioperl installation directory. See the
links at https://www.perl.com and https://bioperl.org/ for current release files and complete
installation instructions.

Introduction

Gleevec™ (STI571 or imatinib mesylate) was the first approved drug to specifically turn off the signal
of a known cancer-causing protein. Initially approved to treat chronic myelogenous leukemia (CML),
it is also effective for treatment of gastrointestinal stromal tumors (GIST).

Research has identified several gene targets for Gleevec including: Proto-oncogene tyrosine-protein
kinase ABL1 (NP_009297), Proto-oncogene tyrosine-protein kinase Kit (NP_000213), and Platelet-
derived growth factor receptor alpha precursor (NP_006197).

target_ABL1 = 'NP_009297';
target_Kit = 'NP_000213';
target_PDGFRA = 'NP_006197';

Accessing Sequence Information

You can load the sequence information for these proteins from local GenPept text files using
genpeptread.

ABL1_seq = getfield(genpeptread('ABL1_gp.txt'), 'Sequence');

Kit_seq = getfield(genpeptread('Kit_gp.txt'), 'Sequence');
PDGFRA_seq = getfield(genpeptread('PDGFRA_gp.txt'), 'Sequence');

Alternatively, you can obtain protein information directly from the online GenPept database
maintained by the National Center for Biotechnology Information (NCBI).

Run these commands to download data from NCBI:

% ABL1_seq = getgenpept(target_ABL1, 'SequenceOnly', true);

% Kit_seq = getgenpept(target_Kit, 'SequenceOnly', true);
% PDGFRA_seq = getgenpept(target_PDGFRA, 'SequenceOnly', true);

The MATLAB whos command gives information about the size of these sequences.

whos ABL1_seq
whos Kit_seq
whos PDGFRA_seq

Name Size Bytes Class Attributes

ABL1_seq 1x1149 2298 char

Name Size Bytes Class Attributes

3-198
 Calling Bioperl Functions from MATLAB

Kit_seq 1x976 1952 char

Name Size Bytes Class Attributes

PDGFRA_seq 1x1089 2178 char

Calling Perl Programs from MATLAB

From MATLAB, you can harness existing Bioperl modules to run a BLAST search on these sequences.
MW_BLAST.pl is a Perl program based on the RemoteBlast Bioperl module. It reads sequences from
FASTA files, so start by creating a FASTA file for each sequence.

fastawrite('ABL1.fa', 'ABL1 Proto-oncogene tyrosine-protein kinase (NP_009297)', ABL1_seq);

fastawrite('Kit.fa', 'Kit Proto-oncogene tyrosine-protein kinase (NP_000213)', Kit_seq);
fastawrite('PDGFRA.fa', 'PDGFRA alpha precursor (NP_006197)', PDGFRA_seq);

Warning: ABL1.fa already exists. The data will be appended to the file.
Warning: Kit.fa already exists. The data will be appended to the file.
Warning: PDGFRA.fa already exists. The data will be appended to the file.

BLAST searches can take a long time to return results, and the Perl program MW_BLAST includes a
repeating sleep state to await the report. Sample results have been included with this example, but if
you want to try running the BLAST search with the three sequences, uncomment the following
commands. MW_BLAST.pl will save the BLAST results in three files on your disk, ABL1.out, Kit.out
and PDGFRA.out. The process can take 15 minutes or more.

% try
% perl('MW_BLAST.pl','blastp','pdb','1e-10','ABL1.fa','Kit.fa','PDGFRA.fa');
% catch
% error(message('bioinfo:bioperldemo:PerlError'))
% end

Here is the Perl code for MW_BLAST:

type MW_BLAST.pl

#!/usr/bin/perl -w
use Bio::Tools::Run::RemoteBlast;
use strict;
use 5.006;

# A sample Blast program based on the RemoteBlast.pm Bioperl module. Takes

# parameters for the BLAST search program, the database, and the expectation
# or E-value (defaults: blastp, pdb, 1e-10), followed by a list of FASTA files
# containing sequences to search.

# Copyright 2003-2004 The MathWorks, Inc.

# Retrieve arguments and set parameters

my $prog = shift @ARGV;
my $db = shift @ARGV;
my $e_val= shift @ARGV;

my @params = ('-prog' => $prog,

3-199
3 Sequence Analysis

'-data' => $db,

'-expect' => $e_val,
'-readmethod' => 'SearchIO' );

# Create a remote BLAST factory

my $factory = Bio::Tools::Run::RemoteBlast->new(@params);

# Change a parameter in RemoteBlast

$Bio::Tools::Run::RemoteBlast::HEADER{'ENTREZ_QUERY'} = 'Homo sapiens [ORGN]';

# Remove a parameter from RemoteBlast

delete $Bio::Tools::Run::RemoteBlast::HEADER{'FILTER'};

# Submit each file

while ( defined($ARGV[0])) {
my $fa_file = shift @ARGV;
my $str = Bio::SeqIO->new(-file=>$fa_file, '-format' => 'fasta' );
my $r = $factory->submit_blast($fa_file);

# Wait for the reply and save the output file

while ( my @rids = $factory->each_rid ) {
foreach my $rid ( @rids ) {
my $rc = $factory->retrieve_blast($rid);
if( !ref($rc) ) {
if( $rc < 0 ) {
$factory->remove_rid($rid);
}
sleep 5;
} else {
my $result = $rc->next_result();
my $filename = $result->query_name()."\.out";
$factory->save_output($filename);
$factory->remove_rid($rid);
}
}
}
}

The next step is to parse the output reports and find scores >= 100. You can then identify hits found
by more than one protein for further research, possibly identifying new targets for drug therapy.

try
protein_list = perl('MW_parse.pl', which('ABL1.out'), which('Kit.out'), which('PDGFRA.out'))
catch
error(message('bioinfo:bioperldemo:PerlError'))
end

protein_list =

'
/home/Data/ABL1.out
1OPL, 2584, 0.0, Chain A, Structural Basis For The Auto-Inhibition Of C-Abl...
1FMK, 923, 1e-100, Crystal Structure Of Human Tyrosine-Protein Kinase C-Src p...
1QCF, 919, 1e-100, Chain A, Crystal Structure Of Hck In Complex With A Src Fa...
1KSW, 916, 1e-100, Chain A, Structure Of Human C-Src Tyrosine Kinase (Thr338g...
1AD5, 883, 6e-96, Chain A, Src Family Kinase Hck-Amp-Pnp Complex pdb|1AD5|B ...
2ABL, 866, 5e-94, Sh3-Sh2 Domain Fragment Of Human Bcr-Abl Tyrosine Kinase

3-200
 Calling Bioperl Functions from MATLAB

3LCK, 666, 9e-71, The Kinase Domain Of Human Lymphocyte Kinase (Lck), Activa...
1QPE, 666, 9e-71, Chain A, Structural Analysis Of The Lymphocyte-Specific Ki...
1QPD, 656, 1e-69, Chain A, Structural Analysis Of The Lymphocyte-Specific Ki...
1K2P, 620, 2e-65, Chain A, Crystal Structure Of Bruton's Tyrosine Kinase Dom...
1BYG, 592, 3e-62, Chain A, Kinase Domain Of Human C-Terminal Src Kinase (Csk...
1M7N, 561, 1e-58, Chain A, Crystal Structure Of Unactivated Apo Insulin-Like...
1JQH, 560, 2e-58, Chain A, Igf-1 Receptor Kinase Domain pdb|1JQH|B Chain B, ...
1P4O, 560, 2e-58, Chain A, Structure Of Apo Unactivated Igf-1r Kinase Domain...
1K3A, 553, 1e-57, Chain A, Structure Of The Insulin-Like Growth Factor 1 Rec...
1GJO, 550, 2e-57, Chain A, The Fgfr2 Tyrosine Kinase Domain
1FVR, 540, 3e-56, Chain A, Tie2 Kinase Domain pdb|1FVR|B Chain B, Tie2 Kinas...
1AB2, 528, 9e-55, Proto-Oncogene Tyrosine Kinase (E.C.2.7.1.112) (Src Homolo...
1IRK, 525, 2e-54, Insulin Receptor (Tyrosine Kinase Domain) Mutant With Cys ...
1I44, 523, 3e-54, Chain A, Crystallographic Studies Of An Activation Loop Mu...
1IR3, 522, 4e-54, Chain A, Phosphorylated Insulin Receptor Tyrosine Kinase I...
1FGK, 522, 4e-54, Chain A, Crystal Structure Of The Tyrosine Kinase Domain O...
1P14, 521, 6e-54, Chain A, Crystal Structure Of A Catalytic-Loop Mutant Of T...
1M14, 496, 4e-51, Chain A, Tyrosine Kinase Domain From Epidermal Growth Fact...
1PKG, 496, 4e-51, Chain A, Structure Of A C-Kit Kinase Product Complex pdb|1...
1VR2, 463, 3e-47, Chain A, Human Vascular Endothelial Growth Factor Receptor...
1JU5, 330, 8e-32, Chain C, Ternary Complex Of An Crk Sh2 Domain, Crk-Derived...
1BBZ, 317, 3e-30, Chain A, Crystal Structure Of The Abl-Sh3 Domain Complexed...
1AWO, 303, 1e-28, The Solution Nmr Structure Of Abl Sh3 And Its Relationship...
1BBZ, 303, 1e-28, Chain E, Crystal Structure Of The Abl-Sh3 Domain Complexed...
1G83, 287, 8e-27, Chain A, Crystal Structure Of Fyn Sh3-Sh2 pdb|1G83|B Chain...
1LCK, 270, 7e-25, Chain A, Sh3-Sh2 Domain Fragment Of Human P56-Lck Tyrosine...
1MUO, 233, 1e-20, Chain A, Crystal Structure Of Aurora-2, An Oncogenic Serin...
1GRI, 232, 2e-20, Chain A, Grb2 pdb|1GRI|B Chain B, Grb2
1A9U, 220, 4e-19, The Complex Structure Of The Map Kinase P38SB203580 pdb|1B...
1BMK, 213, 3e-18, Chain A, The Complex Structure Of The Map Kinase P38SB2186...
1IAN, 209, 8e-18, Human P38 Map Kinase Inhibitor Complex
1GZ8, 208, 1e-17, Chain A, Human Cyclin Dependent Kinase 2 Complexed With Th...
1OVE, 208, 1e-17, Chain A, The Structure Of P38 Alpha In Complex With A Dihy...
1OIT, 207, 1e-17, Chain A, Imidazopyridines: A Potent And Selective Class Of...
1B38, 206, 2e-17, Chain A, Human Cyclin-Dependent Kinase 2 pdb|1B39|A Chain ...
1OGU, 206, 2e-17, Chain A, Structure Of Human Thr160-Phospho Cdk2CYCLIN A CO...
1E9H, 206, 2e-17, Chain A, Thr 160 Phosphorylated Cdk2 - Human Cyclin A3 Com...
1JST, 206, 2e-17, Chain A, Phosphorylated Cyclin-Dependent Kinase-2 Bound To...
1WFC, 206, 2e-17, Structure Of Apo, Unphosphorylated, P38 Mitogen Activated ...
1QMZ, 206, 2e-17, Chain A, Phosphorylated Cdk2-Cyclyin A-Substrate Peptide C...
1DI8, 206, 2e-17, Chain A, The Structure Of Cyclin-Dependent Kinase 2 (Cdk2)...
1H1P, 206, 2e-17, Chain A, Structure Of Human Thr160-Phospho Cdk2CYCLIN A CO...
1DI9, 205, 2e-17, Chain A, The Structure Of P38 Mitogen-Activated Protein Ki...
1H4L, 202, 5e-17, Chain A, Structure And Regulation Of The Cdk5-P25(Nck5a) C...

--------------------- WARNING ---------------------

MSG: No HSPs for this minimal Hit (pdb|1H01|A)
If using NCBI BLAST, check bits() instead
---------------------------------------------------

--------------------- WARNING ---------------------

MSG: No HSPs for this minimal Hit (pdb|1OIR|A)
If using NCBI BLAST, check bits() instead
---------------------------------------------------

--------------------- WARNING ---------------------

MSG: No HSPs for this minimal Hit (pdb|1GII|A)
If using NCBI BLAST, check bits() instead

3-201
3 Sequence Analysis

---------------------------------------------------

--------------------- WARNING ---------------------

MSG: No HSPs for this minimal Hit (pdb|1CSY|A)
If using NCBI BLAST, check bits() instead
---------------------------------------------------

--------------------- WARNING ---------------------

MSG: No HSPs for this minimal Hit (pdb|1F3M|C)
If using NCBI BLAST, check bits() instead
---------------------------------------------------

--------------------- WARNING ---------------------

MSG: No HSPs for this minimal Hit (pdb|1A81|A)
If using NCBI BLAST, check bits() instead
---------------------------------------------------

--------------------- WARNING ---------------------

MSG: No HSPs for this minimal Hit (pdb|1H1W|A)
If using NCBI BLAST, check bits() instead
---------------------------------------------------

--------------------- WARNING ---------------------

MSG: No HSPs for this minimal Hit (pdb|1B6C|B)
If using NCBI BLAST, check bits() instead
---------------------------------------------------

--------------------- WARNING ---------------------

MSG: No HSPs for this minimal Hit (pdb|1IG1|A)
If using NCBI BLAST, check bits() instead
---------------------------------------------------

--------------------- WARNING ---------------------

MSG: No HSPs for this minimal Hit (pdb|1JKK|A)
If using NCBI BLAST, check bits() instead
---------------------------------------------------

--------------------- WARNING ---------------------

MSG: No HSPs for this minimal Hit (pdb|1JOW|B)
If using NCBI BLAST, check bits() instead
---------------------------------------------------

--------------------- WARNING ---------------------

MSG: No HSPs for this minimal Hit (pdb|1BI8|A)
If using NCBI BLAST, check bits() instead
---------------------------------------------------

--------------------- WARNING ---------------------

MSG: No HSPs for this minimal Hit (pdb|1O6K|A)
If using NCBI BLAST, check bits() instead
---------------------------------------------------

--------------------- WARNING ---------------------

MSG: No HSPs for this minimal Hit (pdb|1GZK|A)
If using NCBI BLAST, check bits() instead
---------------------------------------------------

--------------------- WARNING ---------------------

3-202
 Calling Bioperl Functions from MATLAB

MSG: No HSPs for this minimal Hit (pdb|1GZN|A)

If using NCBI BLAST, check bits() instead
---------------------------------------------------

--------------------- WARNING ---------------------

MSG: No HSPs for this minimal Hit (pdb|1O6L|A)
If using NCBI BLAST, check bits() instead
---------------------------------------------------

--------------------- WARNING ---------------------

MSG: No HSPs for this minimal Hit (pdb|1BHF|A)
If using NCBI BLAST, check bits() instead
---------------------------------------------------

--------------------- WARNING ---------------------

MSG: No HSPs for this minimal Hit (pdb|1LCJ|A)
If using NCBI BLAST, check bits() instead
---------------------------------------------------

--------------------- WARNING ---------------------

MSG: No HSPs for this minimal Hit (pdb|1PME|)
If using NCBI BLAST, check bits() instead
---------------------------------------------------

--------------------- WARNING ---------------------

MSG: No HSPs for this minimal Hit (pdb|1CM8|A)
If using NCBI BLAST, check bits() instead
---------------------------------------------------

--------------------- WARNING ---------------------

MSG: No HSPs for this minimal Hit (pdb|1A1A|A)
If using NCBI BLAST, check bits() instead
---------------------------------------------------

--------------------- WARNING ---------------------

MSG: No HSPs for this minimal Hit (pdb|3HCK|)
If using NCBI BLAST, check bits() instead
---------------------------------------------------

--------------------- WARNING ---------------------

MSG: No HSPs for this minimal Hit (pdb|1AOT|F)
If using NCBI BLAST, check bits() instead
---------------------------------------------------

--------------------- WARNING ---------------------

MSG: No HSPs for this minimal Hit (pdb|1PMQ|A)
If using NCBI BLAST, check bits() instead
---------------------------------------------------

--------------------- WARNING ---------------------

MSG: No HSPs for this minimal Hit (pdb|1LKK|A)
If using NCBI BLAST, check bits() instead
---------------------------------------------------

--------------------- WARNING ---------------------

MSG: No HSPs for this minimal Hit (pdb|1JNK|)
If using NCBI BLAST, check bits() instead
---------------------------------------------------

3-203
3 Sequence Analysis

--------------------- WARNING ---------------------

MSG: No HSPs for this minimal Hit (pdb|1SHD|A)
If using NCBI BLAST, check bits() instead
---------------------------------------------------

--------------------- WARNING ---------------------

MSG: No HSPs for this minimal Hit (pdb|1LKL|A)
If using NCBI BLAST, check bits() instead
---------------------------------------------------

--------------------- WARNING ---------------------

MSG: No HSPs for this minimal Hit (pdb|1BM2|A)
If using NCBI BLAST, check bits() instead
---------------------------------------------------

--------------------- WARNING ---------------------

MSG: No HSPs for this minimal Hit (pdb|1BMB|A)
If using NCBI BLAST, check bits() instead
---------------------------------------------------

--------------------- WARNING ---------------------

MSG: No HSPs for this minimal Hit (pdb|1CWD|L)
If using NCBI BLAST, check bits() instead
---------------------------------------------------

--------------------- WARNING ---------------------

MSG: No HSPs for this minimal Hit (pdb|1BHH|B)
If using NCBI BLAST, check bits() instead
---------------------------------------------------

--------------------- WARNING ---------------------

MSG: No HSPs for this minimal Hit (pdb|1IA8|A)
If using NCBI BLAST, check bits() instead
---------------------------------------------------

--------------------- WARNING ---------------------

MSG: No HSPs for this minimal Hit (pdb|1FBZ|A)
If using NCBI BLAST, check bits() instead
---------------------------------------------------

--------------------- WARNING ---------------------

MSG: No HSPs for this minimal Hit (pdb|1IJR|A)
If using NCBI BLAST, check bits() instead
---------------------------------------------------

/home/Data/Kit.out
1PKG, 974, 1e-106, Chain A, Structure Of A C-Kit Kinase Product Complex pdb|1...
1VR2, 805, 6e-87, Chain A, Human Vascular Endothelial Growth Factor Receptor...
1GJO, 730, 3e-78, Chain A, The Fgfr2 Tyrosine Kinase Domain
1FGK, 700, 8e-75, Chain A, Crystal Structure Of The Tyrosine Kinase Domain O...
1OPL, 410, 4e-41, Chain A, Structural Basis For The Auto-Inhibition Of C-Abl...
1FVR, 405, 1e-40, Chain A, Tie2 Kinase Domain pdb|1FVR|B Chain B, Tie2 Kinas...
1M7N, 383, 5e-38, Chain A, Crystal Structure Of Unactivated Apo Insulin-Like...
1P4O, 383, 5e-38, Chain A, Structure Of Apo Unactivated Igf-1r Kinase Domain...
1JQH, 381, 8e-38, Chain A, Igf-1 Receptor Kinase Domain pdb|1JQH|B Chain B, ...
1QCF, 377, 2e-37, Chain A, Crystal Structure Of Hck In Complex With A Src Fa...
1K3A, 371, 1e-36, Chain A, Structure Of The Insulin-Like Growth Factor 1 Rec...

3-204
 Calling Bioperl Functions from MATLAB

1I44, 368, 3e-36, Chain A, Crystallographic Studies Of An Activation Loop Mu...

1IRK, 367, 3e-36, Insulin Receptor (Tyrosine Kinase Domain) Mutant With Cys ...
1P14, 361, 2e-35, Chain A, Crystal Structure Of A Catalytic-Loop Mutant Of T...
1IR3, 361, 2e-35, Chain A, Phosphorylated Insulin Receptor Tyrosine Kinase I...
3LCK, 354, 1e-34, The Kinase Domain Of Human Lymphocyte Kinase (Lck), Activa...
1QPE, 354, 1e-34, Chain A, Structural Analysis Of The Lymphocyte-Specific Ki...
1QPD, 354, 1e-34, Chain A, Structural Analysis Of The Lymphocyte-Specific Ki...
1AD5, 348, 6e-34, Chain A, Src Family Kinase Hck-Amp-Pnp Complex pdb|1AD5|B ...
1KSW, 344, 2e-33, Chain A, Structure Of Human C-Src Tyrosine Kinase (Thr338g...
1FMK, 344, 2e-33, Crystal Structure Of Human Tyrosine-Protein Kinase C-Src p...
1BYG, 342, 3e-33, Chain A, Kinase Domain Of Human C-Terminal Src Kinase (Csk...
1M14, 335, 2e-32, Chain A, Tyrosine Kinase Domain From Epidermal Growth Fact...
1K2P, 294, 1e-27, Chain A, Crystal Structure Of Bruton's Tyrosine Kinase Dom...
1H4L, 167, 5e-13, Chain A, Structure And Regulation Of The Cdk5-P25(Nck5a) C...
1PME, 158, 6e-12, Structure Of Penta Mutant Human Erk2 Map Kinase Complexed ...
1F3M, 156, 1e-11, Chain C, Crystal Structure Of Human SerineTHREONINE KINASE...

/home/Data/PDGFRA.out
1PKG, 625, 5e-66, Chain A, Structure Of A C-Kit Kinase Product Complex pdb|1...
1VR2, 550, 2e-57, Chain A, Human Vascular Endothelial Growth Factor Receptor...
1FGI, 500, 1e-51, Chain A, Crystal Structure Of The Tyrosine Kinase Domain O...
1GJO, 492, 1e-50, Chain A, The Fgfr2 Tyrosine Kinase Domain
1FVR, 419, 4e-42, Chain A, Tie2 Kinase Domain pdb|1FVR|B Chain B, Tie2 Kinas...
1QCF, 380, 1e-37, Chain A, Crystal Structure Of Hck In Complex With A Src Fa...
1QPE, 364, 9e-36, Chain A, Structural Analysis Of The Lymphocyte-Specific Ki...
1QPD, 364, 9e-36, Chain A, Structural Analysis Of The Lymphocyte-Specific Ki...
3LCK, 360, 2e-35, The Kinase Domain Of Human Lymphocyte Kinase (Lck), Activa...
1OPL, 358, 4e-35, Chain A, Structural Basis For The Auto-Inhibition Of C-Abl...
1FMK, 354, 1e-34, Crystal Structure Of Human Tyrosine-Protein Kinase C-Src p...
1KSW, 353, 2e-34, Chain A, Structure Of Human C-Src Tyrosine Kinase (Thr338g...
1AD5, 353, 2e-34, Chain A, Src Family Kinase Hck-Amp-Pnp Complex pdb|1AD5|B ...
1BYG, 352, 2e-34, Chain A, Kinase Domain Of Human C-Terminal Src Kinase (Csk...
1I44, 351, 3e-34, Chain A, Crystallographic Studies Of An Activation Loop Mu...
1IRK, 350, 4e-34, Insulin Receptor (Tyrosine Kinase Domain) Mutant With Cys ...
1M7N, 349, 5e-34, Chain A, Crystal Structure Of Unactivated Apo Insulin-Like...
1JQH, 349, 5e-34, Chain A, Igf-1 Receptor Kinase Domain pdb|1JQH|B Chain B, ...
1P4O, 349, 5e-34, Chain A, Structure Of Apo Unactivated Igf-1r Kinase Domain...
1P14, 344, 2e-33, Chain A, Crystal Structure Of A Catalytic-Loop Mutant Of T...
1IR3, 343, 2e-33, Chain A, Phosphorylated Insulin Receptor Tyrosine Kinase I...
1K3A, 338, 9e-33, Chain A, Structure Of The Insulin-Like Growth Factor 1 Rec...
1M14, 332, 4e-32, Chain A, Tyrosine Kinase Domain From Epidermal Growth Fact...
1K2P, 315, 4e-30, Chain A, Crystal Structure Of Bruton's Tyrosine Kinase Dom...
1PME, 167, 6e-13, Structure Of Penta Mutant Human Erk2 Map Kinase Complexed ...
1JOW, 155, 1e-11, Chain B, Crystal Structure Of A Complex Of Human Cdk6 And ...
1BI8, 155, 1e-11, Chain A, Mechanism Of G1 Cyclin Dependent Kinase Inhibitio...
1F3M, 150, 6e-11, Chain C, Crystal Structure Of Human SerineTHREONINE KINASE...
'

This is the code for MW_parse:

type MW_parse.pl

#!/usr/bin/perl
use Bio::SearchIO;
use strict;
use 5.006;

3-205
3 Sequence Analysis

# A sample BLAST parsing program based on the SearchIO.pm Bioperl module. Takes
# a list of BLAST report files and prints a list of the top hits from each
# report based on an arbitrary minimum score.

# Copyright 2003-2012 The MathWorks, Inc.

# Set a cutoff value for the raw score.

my $min_score = 100;

# Take each report name and print information about the top hits.
my $seq_count = 0;
while ( defined($ARGV[0])) {
my $breport = shift @ARGV;
print "\n$breport\n";
my $in = new Bio::SearchIO(-format => 'blast',
-file => $breport);
my $num_hit = 0;
my $short_desc;
while ( my $result = $in->next_result) {
while ( my $curr_hit = $result->next_hit ) {
if ( $curr_hit->raw_score >= $min_score ) {
if (length($curr_hit->description) >= 60) {
$short_desc = substr($curr_hit->description, 0, 58)."...";
} else {
$short_desc = $curr_hit->description;
}
print $curr_hit->accession, ", ",
$curr_hit->raw_score, ", ",
$curr_hit->significance, ", ",
$short_desc, "\n";
}
$num_hit++;
}
}
$seq_count++;
}

Calling MATLAB Functions within Perl Programs

If you are running on Windows®, it is also possible to call MATLAB functions from Perl. You can
launch MATLAB in an Automation Server mode by using the /Automation switch in the MATLAB
startup command (e.g. D:\applications\matlab7x\bin\matlab.exe /Automation).

Here's a script to illustrate the process of launching an automation server, calling MATLAB functions
and passing variables between Perl and MATLAB.

type MATLAB_from_Perl.pl

#!/usr/bin/perl -w
use Win32::OLE;
use Win32::OLE::Variant;

# Simple perl script to execute commands in Matlab.

# Note the name Win32::OLE is misleading and this actually uses COM!
#

3-206
 Calling Bioperl Functions from MATLAB

# Use existing instance if Matlab is already running.

eval {$matlabApp = Win32::OLE->GetActiveObject('Matlab.Application')};
die "Matlab not installed" if $@;
unless (defined $matlabApp) {
$matlabApp = Win32::OLE->new('Matlab.Application')
or die "Oops, cannot start Matlab";
}

# Examples of executing MATLAB commands - these functions execute in

# MATLAB and return the status.

@exe_commands = ("IRK = pdbread('pdb1irk.ent');",

"LCK = pdbread('pdb3lck.ent');",
"seqdisp(IRK)",
"seqdisp(LCK)",
"[Score, Alignment] = swalign(IRK, LCK,'showscore',1);");

# send the commands to Matlab

foreach $exe_command (@exe_commands)
{ $status = &send_to_matlab('Execute', $exe_command);
print "Matlab status = ", $status, "\n";
}

sub send_to_matlab
{ my ($call, @command) = @_;
my $status = 0;
print "\n>> $call( @command )\n";
$result = $matlabApp->Invoke($call, @command);
if (defined($result))
{ unless ($result =~ s/^.\?{3}/Error:/)
{ print "$result\n" unless ($result eq "");
}
else
{ print "$result\n";
$status = -1;
}
}
return $status;
}

# Examples of passing variables between MATLAB and Perl.

#
# MATLAB supoprts passing character arrays directly with the following syntax:
#
# PutCharArray([in] BSTR name, [in] BSTR workspace, [in] BSTR string);
# GetCharArray([in] BSTR name, [in] BSTR workspace, [out] BSTR string);

&send_to_matlab('PutCharArray', 'centralDogma', 'base', 'DNA->RNA->Protein.');

&send_to_matlab('GetCharArray', 'centralDogma', 'base');

# Numeric arrays can be passed by reference in a SAFEARRAY using the

# PutFullMatrix and GetFullMatrix functions.
#
# PutFullMatrix([in] BSTR name, [in] BSTR workspace, [in] BSTR data);
# GetFullMatrix([in] BSTR varname, [in] BSTR workspace, [out] BSTR retdata);

3-207
3 Sequence Analysis

$mReal = Variant(VT_ARRAY|VT_R8, 4, 4);

$mImag = Variant(VT_ARRAY|VT_R8, 4, 4);

$mReal->Put([[0,0,0,0], [0,0,0,0], [0,0,0,0], [0,0,0,0]]);

print "\n>> PutFullMatrix( 'magicArray', 'base', ",'$mReal, $mImag'," )\n";
$matlabApp->PutFullMatrix('magicArray', 'base', $mReal, $mImag);
$matlabApp->Execute('magicArray = magic(4)');

$m2Real = Variant(VT_ARRAY|VT_R8|VT_BYREF,4,4);
$m2Imag = Variant(VT_ARRAY|VT_R8|VT_BYREF,4,4);
print "\n>> GetFullMatrix( 'magicArray', 'base', ",'$m2Real, $m2Imag'," )\n";
$matlabApp->GetFullMatrix('magicArray', 'base', $m2Real, $m2Imag);

for ($i = 0; $i < 4; $i++) {

printf "%3d %3d %3d %3d\n", $m2Real->Get($i,0), $m2Real->Get($i,1),
$m2Real->Get($i,2), $m2Real->Get($i,3);
}

# Additionally, you can use Variants to send scalar variables by reference

# to MATLAB for all data types except sparse arrays and function handles through
# PutWorkspaceData:
# PutWorkspaceData([in] BSTR name, [in] BSTR workspace, [in] BSTR data);
#
# Results are passed back to Perl directly with GetVariable:
# HRESULT = GetVariable([in] BSTR Name, [in] BSTR Workspace);

# Create and initialize a date Variant.

$dnaDate = Variant->new(VT_DATE|VT_BYREF, 'Feb 28, 1953');

&send_to_matlab('PutWorkspaceData', 'dnaDate', 'base', $dnaDate);

&send_to_matlab('Execute', 'dnaDate');

# Create and initialize a new string Variant.

$aminoString = Variant->new(VT_BSTR|VT_BYREF, 'matlap');
&send_to_matlab('PutWorkspaceData', 'aminoAcids', 'base', $aminoString);

# Change the value in MATLAB

&send_to_matlab('Execute', "aminoAcids = 'ARNDCQEGHILKMFPSTWYV';");

# Bring the new value back

$aa = $matlabApp->GetVariable('aminoAcids', 'base');
printf "Amino acid codes: %s\n", $aa;

undef $matlabApp; # close Matlab if we opened it

Protein Analysis Tools in Bioinformatics Toolbox™

MATLAB offers additional tools for protein analysis and further research with these proteins. For
example, to access the sequences and run a full Smith-Waterman alignment on the tyrosine kinase
domain of the human insulin receptor (pdb 1IRK) and the kinase domain of the human lymphocyte
kinase (pdb 3LCK), load the sequence data:
IRK = pdbread('pdb1irk.ent');
LCK = pdbread('pdb3lck.ent');

% Run these commands to bring the data from the Internet:

% IRK = getpdb('1IRK');
% LCK = getpdb('3LCK');

3-208
 Calling Bioperl Functions from MATLAB

Now perform a local alignment with the Smith-Waterman algorithm. MATLAB uses BLOSUM 50 as
the default scoring matrix for AA strings with a gap penalty of 8. Of course, you can change any of
these parameters.

[Score, Alignment] = swalign(IRK, LCK, 'showscore', true);

MATLAB and the Bioinformatics Toolbox™ offer additional tools for investigating nucleotide and
amino acid sequences. For example, pdbdistplot displays the distances between atoms and amino
acids in a PDB structure, while ramachandran generates a plot of the torsion angle PHI and the
torsion angle PSI of the protein sequence. The toolbox function proteinplot provides a graphical
user interface (GUI) to easily import sequences and plot various properties such as hydrophobicity.

3-209
3 Sequence Analysis

Accessing NCBI Entrez Databases with E-Utilities

This example shows how to programmatically search and retrieve data from NCBI's Entrez databases
using NCBI's Entrez Utilities (E-Utilities).

Using NCBI E-Utilities to Retrieve Biological Data

E-Utilities (eUtils) are server-side programs (e.g. ESearch, ESummary, EFetch, etc.,) developed and
maintained by NCBI for searching and retrieving data from most Entpwdrez Databases. You access
tools via URLs with a strict syntax of a specific base URL, a call to the eUtil's script and its associated
parameters. For more details on eUtils, see E-Utilities Help.

Searching Nucleotide Database with ESearch

In this example, we consider the genes sequenced from the H5N1 virus, isolated in 1997 from a
chicken in Hong Kong as a starting point for our analysis. This particular virus jumped from chickens
to humans, killing six people before the spread of the disease was brought under control by
destroying all poultry in Hong Kong [1]. You can use ESearch to find the sequence data needed for
the analysis. ESearch requires input of a database (db) and search term (term). Optionally, you can
request for ESearch to store your search results on the NCBI history server through the usehistory
parameter.

baseURL = 'http://eutils.ncbi.nlm.nih.gov/entrez/eutils/';
eutil = 'esearch.fcgi?';
dbParam = 'db=nuccore';
termParam = '&term=A/chicken/Hong+Kong/915/97+OR+A/chicken/Hong+Kong/915/1997';
usehistoryParam = '&usehistory=y';
esearchURL = [baseURL, eutil, dbParam, termParam, usehistoryParam]

esearchURL =

1×145 char array

http://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi?db=nuccore&term=A/chicken/Hong+Kong/915

The term parameter can be any valid Entrez query. Note that there cannot be any spaces in the URL,
so parameters are separated by '&' and any spaces in a query term need to be replaced with '+' (e.g.
'Hong+Kong').

You can use webread to send the URL and return the results from ESearch as a character array.

searchReport = webread(esearchURL)

searchReport =

1×1714 char array

<?xml version="1.0" encoding="UTF-8"?>

<!DOCTYPE eSearchResult PUBLIC "-//NLM//DTD esearch 20060628//EN" "http://eutils.ncbi.nlm.nih.gov
<eSearchResult><Count>8</Count><RetMax>8</RetMax><RetStart>0</RetStart><QueryKey>1</QueryKey><Web
<Id>6048802</Id>
<Id>6048927</Id>

3-210
 Accessing NCBI Entrez Databases with E-Utilities

<Id>6048903</Id>
<Id>6048875</Id>
<Id>6048849</Id>
<Id>6048829</Id>
<Id>6048770</Id>
<Id>3421265</Id>
</IdList><TranslationSet/><TranslationStack> <TermSet> <Term>A/chicken/Hong[All Fields]</Ter

ESearch returns the search results in XML. The report contains information about the query
performed, which database was searched and UIDs (unique IDs) to the records that match the query.
If you use the history server, the report contains two additional IDs, WebEnv and query_key, for
accessing the results. WebEnv is the location of the results on the server, and query_key is a number
indexing the queries performed. Since WebEnv and query_key are query dependent they will
change every time the search is executed. Either the UIDs or WebEnv and query_key can be parsed
out of the XML report then passed to other eUtils. You can use regexp to do the parsing and store
the tokens in the structure with fieldnames WebEnv and QueryKey.

ncbi = regexp(searchReport,...
'<QueryKey>(?<QueryKey>\w+)</QueryKey>\s*<WebEnv>(?<WebEnv>\S+)</WebEnv>',...
'names')

ncbi =

struct with fields:

QueryKey: '1'
WebEnv: 'NCID_1_3777459_130.14.22.215_9001_1464976330_1306835914_0Me...'

Getting GenBank® File Summaries with ESummary

To get a quick overview of sequences that matched the query you can use ESummary. ESummary
retrieves a brief summary, or Document Summary (DocSum), for each record. ESummary requires an
input of which database to access and which records to retrieve, identified either by a list of UIDs
passed through id parameter or by the WebEnv and query_key parameters. ESummary returns a
report in XML that contains the summary information for each record. Use websave with ESummary
to perform the record summary retrieval and write out the XML report to a file.

tmpDirectory = tempdir;
summaryFname = fullfile(tmpDirectory,'summaryReport.xml');

websave(summaryFname, [baseURL...
'esummary.fcgi?db=nuccore&WebEnv=',ncbi.WebEnv,...
'&query_key=',ncbi.QueryKey]);

You can create an XSL stylesheet to view information from the ESummary XML report in a web
browser. For more information on writing XSL stylesheets, see W3C® XSL. An XSL stylesheet was
created for this example to view the sequence summary information and provide links to their full
GenBank® files. Xslt can be used to view the XML report in a Web browser from MATLAB®.

xslt(summaryFname,'genbankSummary.xsl','-web');

3-211
3 Sequence Analysis

Retrieving Full GenBank Files with EFetch

To perform the sequence analysis, you need to get the full GenBank record for each sequence. EFetch
retrieves full records from Entrez databases. EFetch requires an input of a database and a list of
UIDs or WebEnv and query_key. Additionally, EFetch can return the output in different formats. You
can specify which output format (i.e. GenBank (gb), FASTA) and file format (i.e. text, ASN.1, XML)
you want through the rettype and retmode parameters, respectively. Rettype equals gb for
GenBank file format and retmode equals text for this query. Genbankread can be used directly
with the EFetch URL to retrieve all the GenBank records and read them into a structure array. This
structure can then be used as input to seqviewer to visualize the sequences.

ch97struct = genbankread([baseURL...
'efetch.fcgi?db=nuccore&rettype=gb&retmode=text&WebEnv=',ncbi.WebEnv,...
'&query_key=',ncbi.QueryKey]);
seqviewer(ch97struct)

3-212
 Accessing NCBI Entrez Databases with E-Utilities

Finding Links Between Databases with ELink

It might be useful to have PubMed articles related to these genes records. ELink provides this
functionality. It finds associations between records within or between databases. You can give ELink
the query_key and WebEnv IDs from above and tell it to find records in the PubMed Database (db
parameter) associated with your records from the Nucleotide (nuccore) Database (dbfrom
parameter). ELink returns an XML report with the UIDs for the records in PubMed. These UIDs can

3-213
3 Sequence Analysis

be parsed out of the report and passed to other eUtils (e.g. ESummary). Use the stylesheet created
for viewing ESummary reports to view the results of ELink.

elinkReport = webread([baseURL...
'elink.fcgi?dbfrom=nuccore&db=pubmed&WebEnv=', ncbi.WebEnv,...
'&query_key=',ncbi.QueryKey]);

Extract the PubMed UIDs from the ELink report.

pubmedIDs = regexp(elinkReport,'<Link>\s+<Id>(\w*)</Id>\s+</Link>','tokens');
NumberOfArticles = numel(pubmedIDs)

% Put PubMed UIDs into a string that can be read by EPost URL.

pubmed_str = [];
for ii = 1:NumberOfArticles
pubmed_str = sprintf([pubmed_str '%s,'],char(pubmedIDs{ii}));
end

NumberOfArticles =

Posting UIDs to NCBI History Server with EPost

You can use EPost to posts UIDs to the history server. It returns an XML report with a query_key
and WebEnv IDs pointing to the location of the history server. Again, these can be parsed out of the
report and used with other eUtils calls.

epostReport = webread([baseURL 'epost.fcgi?db=pubmed&id=',pubmed_str(1:end-1)]);

epostKeys = regexp(epostReport,...
'<QueryKey>(?<QueryKey>\w+)</QueryKey>\s*<WebEnv>(?<WebEnv>\S+)</WebEnv>','names')

epostKeys =

struct with fields:

QueryKey: '1'
WebEnv: 'NCID_1_3778415_130.14.22.215_9001_1464976335_906725031_0Met...'

Using ELink to Find Associated Files within the Same Database

ELink can do "within" database searches. For example, you can query for a nucleotide sequence
within Nucleotide (nuccore) database to find similar sequences, essentially performing a BLAST
search. For "within" database searches, ELink returns an XML report containing the related records,
along with a score ranking its relationship to the query record. From the above PubMed search, you
might be interested in finding all articles related to those articles in PubMed. This is easy to do with
ELink. To do a "within" database search, set db and dbfrom to PubMed. You can use the query_key
and WebEnv from the EPost call.

pm2pmReport = webread([baseURL...
'elink.fcgi?dbfrom=pubmed&db=pubmed&query_key=',epostKeys.QueryKey,...
'&WebEnv=',epostKeys.WebEnv]);
pubmedIDs = regexp(pm2pmReport,'(?<=<Id>)\w*(?=</Id>)','match');

3-214
 Accessing NCBI Entrez Databases with E-Utilities

NumberOfArticles = numel(unique(pubmedIDs))

pubmed_str = [];
for ii = 1:NumberOfArticles
pubmed_str = sprintf([pubmed_str '%s,'],char(pubmedIDs{ii}));
end

NumberOfArticles =

526

Use websave with EFetch to retrieve full abstracts for the articles and write out the returned XML
report to a file. An XSL stylesheet is provided with this example for viewing the results of the EFetch
query. The XML report can be transformed using the stylesheet and opened in a Web browser from
MATLAB using xslt.

fullFname = fullfile(tmpDirectory,'H5N1_relatedArticles.xml');
websave(fullFname, [baseURL 'efetch.fcgi?db=pubmed&retmode=xml&id=',...
pubmed_str(1:end-1)]);
xslt(fullFname,'pubmedFullReport.xsl','-web');

3-215
3 Sequence Analysis

Using EGQuery to get a Global View of H5N1 Related-Records in Entrez

To see what other Entrez databases contain information about the H5N1 virus, use EGQuery.
EGQuery performs a text search across all available Entrez databases and returns the number of hits
in each. EGQuery accepts any valid Entrez text query as input through the term parameter.
entrezSearch = webread([baseURL,'egquery.fcgi?term=H5N1+AND+virus']);
entrezResults = regexp(entrezSearch,...
'<DbName>(?<DB>\w+\s*\w*)</DbName>.*?(<Count>)(?<Count>\d+)</Count>',...
'names');

entrezDBs = {entrezResults(:).DB};
dbCounts = str2double({entrezResults(:).Count});
entrezDBs = entrezDBs(logical(dbCounts)); % remove databases with no records
[dbCounts,sortInd] = sort(dbCounts(logical(dbCounts)));
entrezDBs = entrezDBs(sortInd);
numDBs = numel(entrezDBs);

barh(log10(dbCounts));
ylim([.5 numDBs+.5])
ax = gca;

3-216
 Accessing NCBI Entrez Databases with E-Utilities

ax.YTick = 1:numDBs;
ax.YTickLabel = entrezDBs;
xlabel('Log(Number of Records)');
title('Number of H5N1 Related-Records Per Entrez Database');

References

[1] Cristianini, N. and Hahn, M.W. "Introduction to Computational Genomics: A Case Studies
Approach", Cambridge University Press, 2007.

3-217
 4

Microarray Analysis

• “Managing Gene Expression Data in Objects” on page 4-2

• “Representing Expression Data Values in DataMatrix Objects” on page 4-5
• “Representing Expression Data Values in ExptData Objects” on page 4-9
• “Representing Sample and Feature Metadata in MetaData Objects” on page 4-12
• “Representing Experiment Information in a MIAME Object” on page 4-16
• “Representing All Data in an ExpressionSet Object” on page 4-19
• “Analyzing Illumina Bead Summary Gene Expression Data” on page 4-23
• “Detecting DNA Copy Number Alteration in Array-Based CGH Data” on page 4-44
• “Analyzing Array-Based CGH Data Using Bayesian Hidden Markov Modeling” on page 4-60
• “Visualizing Microarray Data” on page 4-74
• “Gene Expression Profile Analysis” on page 4-95
• “Working with Affymetrix Data” on page 4-112
• “Preprocessing Affymetrix Microarray Data at the Probe Level” on page 4-131
• “Analyzing Affymetrix SNP Arrays for DNA Copy Number Variants” on page 4-142
• “Working with GEO Series Data” on page 4-162
• “Identifying Biomolecular Subgroups Using Attractor Metagenes” on page 4-173
• “Working with the Clustergram Function” on page 4-185
• “Working with Objects for Microarray Experiment Data” on page 4-203
4 Microarray Analysis

Managing Gene Expression Data in Objects

Microarray gene expression experiments are complex, containing data and information from various
sources. The data and information from such an experiment is typically subdivided into four
categories:

• Measured expression data values

• Sample metadata
• Microarray feature metadata
• Descriptions of experiment methods and conditions

In MATLAB, you can represent all the previous data and information in an ExpressionSet object,
which typically contains the following objects:

• One ExptData object containing expression values from a microarray experiment in one or more
DataMatrix objects
• One MetaData object containing sample metadata in two dataset arrays
• One MetaData object containing feature metadata in two dataset arrays
• One MIAME object containing experiment descriptions

The following graphic illustrates a typical ExpressionSet object and its component objects.

4-2
 Managing Gene Expression Data in Objects

Each element (DataMatrix object) in the ExpressionSet object has an element name. Also, there is
always one DataMatrix object whose element name is Expressions.

An ExpressionSet object lets you store, manage, and subset the data from a microarray gene
expression experiment. An ExpressionSet object includes properties and methods that let you access,
retrieve, and change data, metadata, and other information about the microarray experiment. These
properties and methods are useful to view and analyze the data. For a list of the properties and
methods, see ExpressionSet class.

To learn more about constructing and using objects for microarray gene expression data and
information, see:

• “Representing Expression Data Values in DataMatrix Objects” on page 4-5

• “Representing Expression Data Values in ExptData Objects” on page 4-9

4-3
4 Microarray Analysis

• “Representing Sample and Feature Metadata in MetaData Objects” on page 4-12

• “Representing Experiment Information in a MIAME Object” on page 4-16
• “Representing All Data in an ExpressionSet Object” on page 4-19

4-4
 Representing Expression Data Values in DataMatrix Objects

Representing Expression Data Values in DataMatrix Objects

In this section...
“Overview of DataMatrix Objects” on page 4-5
“Constructing DataMatrix Objects” on page 4-5
“Getting and Setting Properties of a DataMatrix Object” on page 4-6
“Accessing Data in DataMatrix Objects” on page 4-6

Overview of DataMatrix Objects

The toolbox includes functions, objects, and methods for creating, storing, and accessing microarray
data.

The object constructor function, DataMatrix, lets you create a DataMatrix object to encapsulate
data and metadata (row and column names) from a microarray experiment. A DataMatrix object
stores experimental data in a matrix, with rows typically corresponding to gene names or probe
identifiers, and columns typically corresponding to sample identifiers. A DataMatrix object also stores
metadata, including the gene names or probe identifiers (as the row names) and sample identifiers
(as the column names).

You can reference microarray expression values in a DataMatrix object the same way you reference
data in a MATLAB array, that is, by using linear or logical indexing. Alternately, you can reference this
experimental data by gene (probe) identifiers and sample identifiers. Indexing by these identifiers lets
you quickly and conveniently access subsets of the data without having to maintain additional index
arrays.

Many MATLAB operators and arithmetic functions are available to DataMatrix objects by means of
methods. These methods let you modify, combine, compare, analyze, plot, and access information
from DataMatrix objects. Additionally, you can easily extend the functionality by using general
element-wise functions, dmarrayfun and dmbsxfun, and by manually accessing the properties of a
DataMatrix object.

Note For tables describing the properties and methods of a DataMatrix object, see the DataMatrix
object reference page.

Constructing DataMatrix Objects

1 Load the MAT-file, provided with the Bioinformatics Toolbox software, that contains yeast data.
This MAT-file includes three variables: yeastvalues, a 614-by-7 matrix of gene expression data,
genes, a cell array of 614 GenBank accession numbers for labeling the rows in yeastvalues,
and times, a 1-by-7 vector of time values for labeling the columns in yeastvalues.

load filteredyeastdata
2 Create variables to contain a subset of the data, specifically the first five rows and first four
columns of the yeastvalues matrix, the genes cell array, and the times vector.

yeastvalues = yeastvalues(1:5,1:4);
genes = genes(1:5,:);
times = times(1:4);

4-5
4 Microarray Analysis

3 Import the microarray object package so that the DataMatrix constructor function will be
available.

import bioma.data.*
4 Use the DataMatrix constructor function to create a small DataMatrix object from the gene
expression data.

dmo = DataMatrix(yeastvalues,genes,times)

dmo =

0 9.5 11.5 13.5

SS DNA -0.131 1.699 -0.026 0.365
YAL003W 0.305 0.146 -0.129 -0.444
YAL012W 0.157 0.175 0.467 -0.379
YAL026C 0.246 0.796 0.384 0.981
YAL034C -0.235 0.487 -0.184 -0.669

Getting and Setting Properties of a DataMatrix Object

You use the get and set methods to retrieve and set properties of a DataMatrix object.

1 Use the get method to display the properties of the DataMatrix object, dmo.

get(dmo)
Name: ''
RowNames: {5x1 cell}
ColNames: {' 0' ' 9.5' '11.5' '13.5'}
NRows: 5
NCols: 4
NDims: 2
ElementClass: 'double'
2 Use the set method to specify a name for the DataMatrix object, dmo.

dmo = set(dmo,'Name','MyDMObject');
3 Use the get method again to display the properties of the DataMatrix object, dmo.

get(dmo)
Name: 'MyDMObject'
RowNames: {5x1 cell}
ColNames: {' 0' ' 9.5' '11.5' '13.5'}
NRows: 5
NCols: 4
NDims: 2
ElementClass: 'double'

Note For a description of all properties of a DataMatrix object, see the DataMatrix object reference
page.

Accessing Data in DataMatrix Objects

DataMatrix objects support the following types of indexing to extract, assign, and delete data:

4-6
 Representing Expression Data Values in DataMatrix Objects

• Parenthesis ( ) indexing
• Dot . indexing

Parentheses () Indexing

Use parenthesis indexing to extract a subset of the data in dmo and assign it to a new DataMatrix
object dmo2:

dmo2 = dmo(1:5,2:3)
dmo2 =
9.5 11.5
SS DNA 1.699 -0.026
YAL003W 0.146 -0.129
YAL012W 0.175 0.467
YAL026C 0.796 0.384
YAL034C 0.487 -0.184

Use parenthesis indexing to extract a subset of the data using row names and column names, and
assign it to a new DataMatrix object dmo3:

dmo3 = dmo({'SS DNA','YAL012W','YAL034C'},'11.5')

dmo3 =

11.5
SS DNA -0.026
YAL012W 0.467
YAL034C -0.184

Note If you use a cell array of row names or column names to index into a DataMatrix object, the
names must be unique, even though the row names or column names within the DataMatrix object
are not unique.

Use parenthesis indexing to assign new data to a subset of the elements in dmo2:

dmo2({'SS DNA', 'YAL003W'}, 1:2) = [1.700 -0.030; 0.150 -0.130]

dmo2 =

9.5 11.5
SS DNA 1.7 -0.03
YAL003W 0.15 -0.13
YAL012W 0.175 0.467
YAL026C 0.796 0.384
YAL034C 0.487 -0.184

Use parenthesis indexing to delete a subset of the data in dmo2:

dmo2({'SS DNA', 'YAL003W'}, :) = []

dmo2 =

9.5 11.5
YAL012W 0.175 0.467
YAL026C 0.796 0.384
YAL034C 0.487 -0.184

4-7
4 Microarray Analysis

Dot . Indexing

Note In the following examples, notice that when using dot indexing with DataMatrix objects, you
specify all rows or all columns using a colon within single quotation marks, (':').

Use dot indexing to extract the data from the 11.5 column only of dmo:

timeValues = dmo.(':')('11.5')
timeValues =

-0.0260
-0.1290
0.4670
0.3840
-0.1840

Use dot indexing to assign new data to a subset of the elements in dmo:

dmo.(1:2)(':') = 7
dmo =

0 9.5 11.5 13.5

SS DNA 7 7 7 7
YAL003W 7 7 7 7
YAL012W 0.157 0.175 0.467 -0.379
YAL026C 0.246 0.796 0.384 0.981
YAL034C -0.235 0.487 -0.184 -0.669

Use dot indexing to delete an entire variable from dmo:

dmo.YAL034C = []
dmo =

0 9.5 11.5 13.5

SS DNA 7 7 7 7
YAL003W 7 7 7 7
YAL012W 0.157 0.175 0.467 -0.379
YAL026C 0.246 0.796 0.384 0.981

Use dot indexing to delete two columns from dmo:

dmo.(':')(2:3)=[]

dmo =

0 13.5
SS DNA 7 7
YAL003W 7 7
YAL012W 0.157 -0.379
YAL026C 0.246 0.981

4-8
 Representing Expression Data Values in ExptData Objects

Representing Expression Data Values in ExptData Objects

In this section...
“Overview of ExptData Objects” on page 4-9
“Constructing ExptData Objects” on page 4-9
“Using Properties of an ExptData Object” on page 4-10
“Using Methods of an ExptData Object” on page 4-10
“References” on page 4-11

Overview of ExptData Objects

You can use an ExptData object to store expression values from a microarray experiment. An
ExprData object stores the data values in one or more DataMatrix objects, each having the same row
names (feature names) and column names (sample names). Each element (DataMatrix object) in the
ExptData object has an element name.

The following illustrates a small DataMatrix object containing expression values from three samples
(columns) and seven features (rows):

A B C
100001_at 2.26 20.14 31.66
100002_at 158.86 236.25 206.27
100003_at 68.11 105.45 82.92
100004_at 74.32 96.68 84.87
100005_at 75.05 53.17 57.94
100006_at 80.36 42.89 77.21
100007_at 216.64 191.32 219.48

An ExptData object lets you store, manage, and subset the data values from a microarray experiment.
An ExptData object includes properties and methods that let you access, retrieve, and change data
values from a microarray experiment. These properties and methods are useful to view and analyze
the data. For a list of the properties and methods, see ExptData class.

Constructing ExptData Objects

The mouseExprsData.txt file used in this example contains data from Hovatta et al., 2005.

1 Import the bioma.data package so that the DataMatrix and ExptData constructor functions
are available.

import bioma.data.*
2 Use the DataMatrix constructor function to create a DataMatrix object from the gene
expression data in the mouseExprsData.txt file. This file contains a table of expression values
and metadata (sample and feature names) from a microarray experiment done using the
Affymetrix MGU74Av2 GeneChip array. There are 26 sample names (A through Z), and 500
feature names (probe set names).

dmObj = DataMatrix('File', 'mouseExprsData.txt');

3 Use the ExptData constructor function to create an ExptData object from the DataMatrix object.

EDObj = ExptData(dmObj);

4-9
4 Microarray Analysis

4 Display information about the ExptData object, EDObj.

EDObj

Experiment Data:
500 features, 26 samples
1 elements
Element names: Elmt1

Note For complete information on constructing ExptData objects, see ExptData class.

Using Properties of an ExptData Object

To access properties of an ExptData object, use the following syntax:
objectname.propertyname

For example, to determine the number of elements (DataMatrix objects) in an ExptData object:
EDObj.NElements

ans =

To set properties of an ExptData object, use the following syntax:

objectname.propertyname = propertyvalue

For example, to set the Name property of an ExptData object:

EDObj.Name = 'MyExptDataObject'

Note Property names are case sensitive. For a list and description of all properties of an ExptData
object, see ExptData class.

Using Methods of an ExptData Object

To use methods of an ExptData object, use either of the following syntaxes:
objectname.methodname

or
methodname(objectname)

For example, to retrieve the sample names from an ExptData object:

EDObj.sampleNames

Columns 1 through 9

'A' 'B' 'C' 'D' 'E' 'F' 'G' 'H' 'I' ...

To return the size of an ExptData object:

4-10
 Representing Expression Data Values in ExptData Objects

size(EDObj)

ans =

500 26

Note For a complete list of methods of an ExptData object, see ExptData class.

References

[1] Hovatta, I., Tennant, R S., Helton, R., et al. (2005). Glyoxalase 1 and glutathione reductase 1
regulate anxiety in mice. Nature 438, 662–666.

4-11
4 Microarray Analysis

Representing Sample and Feature Metadata in MetaData

Objects
In this section...
“Overview of MetaData Objects” on page 4-12
“Constructing MetaData Objects” on page 4-13
“Using Properties of a MetaData Object” on page 4-15
“Using Methods of a MetaData Object” on page 4-15

Overview of MetaData Objects

You can store either sample or feature metadata from a microarray gene expression experiment in a
MetaData object. The metadata consists of variable names, for example, related to either samples or
microarray features, along with descriptions and values for the variables.

A MetaData object stores the metadata in two dataset arrays:

• Values dataset array — A dataset array containing the measured value of each variable per
sample or feature. In this dataset array, the columns correspond to variables and rows correspond
to either samples or features. The number and names of the columns in this dataset array must
match the number and names of the rows in the Descriptions dataset array. If this dataset array
contains sample metadata, then the number and names of the rows (samples) must match the
number and names of the columns in the DataMatrix objects in the same ExpressionSet object. If
this dataset array contains feature metadata, then the number and names of the rows (features)
must match the number and names of the rows in the DataMatrix objects in the same
ExpressionSet object.
• Descriptions dataset array — A dataset array containing a list of the variable names and their
descriptions. In this dataset array, each row corresponds to a variable. The row names are the
variable names, and a column, named VariableDescription, contains a description of the
variable. The number and names of the rows in the Descriptions dataset array must match the
number and names of the columns in the Values dataset array.

The following illustrates a dataset array containing the measured value of each variable per sample
or feature:
Gender Age Type Strain Source
A 'Male' 8 'Wild type' '129S6/SvEvTac' 'amygdala'
B 'Male' 8 'Wild type' '129S6/SvEvTac' 'amygdala'
C 'Male' 8 'Wild type' '129S6/SvEvTac' 'amygdala'
D 'Male' 8 'Wild type' 'A/J ' 'amygdala'
E 'Male' 8 'Wild type' 'A/J ' 'amygdala'
F 'Male' 8 'Wild type' 'C57BL/6J ' 'amygdala'

The following illustrates a dataset array containing a list of the variable names and their descriptions:
VariableDescription
id 'Sample identifier'
Gender 'Gender of the mouse in study'
Age 'The number of weeks since mouse birth'
Type 'Genetic characters'
Strain 'The mouse strain'
Source 'The tissue source for RNA collection'

4-12
 Representing Sample and Feature Metadata in MetaData Objects

A MetaData object lets you store, manage, and subset the metadata from a microarray experiment. A
MetaData object includes properties and methods that let you access, retrieve, and change metadata
from a microarray experiment. These properties and methods are useful to view and analyze the
metadata. For a list of the properties and methods, see MetaData class

Constructing MetaData Objects

Constructing a MetaData Object from Two dataset Arrays

1 Import the bioma.data package so that the MetaData constructor function is available.

import bioma.data.*
2 Load some sample data, which includes Fisher’s iris data of 5 measurements on a sample of 150
irises.

load fisheriris
3 Create a dataset array from some of Fisher's iris data. The dataset array will contain 750
measured values, one for each of 150 samples (iris replicates) at five variables (species, SL, SW,
PL, PW). In this dataset array, the rows correspond to samples, and the columns correspond to
variables.

irisValues = dataset({nominal(species),'species'}, ...

{meas, 'SL', 'SW', 'PL', 'PW'});
4 Create another dataset array containing a list of the variable names and their descriptions. This
dataset array will contain five rows, each corresponding to the five variables: species, SL, SW, PL,
and PW. The first column will contain the variable name. The second column will have a column
header of VariableDescription and contain a description of the variable.

% Create 5-by-1 cell array of description text for the variables

varDesc = {'Iris species', 'Sepal Length', 'Sepal Width', ...
'Petal Length', 'Petal Width'}';
% Create the dataset array from the variable descriptions
irisVarDesc = dataset(varDesc, ...
'ObsNames', {'species','SL','SW','PL','PW'}, ...
'VarNames', {'VariableDescription'})

irisVarDesc =

VariableDescription
species 'Iris species'
SL 'Sepal Length'
SW 'Sepal Width'
PL 'Petal Length'
PW 'Petal Width'
5 Create a MetaData object from the two dataset arrays.

MDObj1 = MetaData(irisValues, irisVarDesc);

Constructing a MetaData Object from a Text File

1 Import the bioma.datapackage so that the MetaData constructor function is available.

import bioma.data.*
2 View the mouseSampleData.txt file included with the Bioinformatics Toolbox software.

4-13
4 Microarray Analysis

Note that this text file contains two tables. One table contains 130 measured values, one for each
of 26 samples (A through Z) at five variables (Gender, Age, Type, Strain, and Source). In this
table, the rows correspond to samples, and the columns correspond to variables. The second
table has lines prefaced by the # symbol. It contains five rows, each corresponding to the five
variables: Gender, Age, Type, Strain, and Source. The first column contains the variable name.
The second column has a column header of VariableDescription and contains a description
of the variable.

# id: Sample identifier

# Gender: Gender of the mouse in study
# Age: The number of weeks since mouse birth
# Type: Genetic characters
# Strain: The mouse strain
# Source: The tissue source for RNA collection
ID Gender Age Type Strain Source
A Male 8 Wild type 129S6/SvEvTac amygdala
B Male 8 Wild type 129S6/SvEvTac amygdala
C Male 8 Wild type 129S6/SvEvTac amygdala
D Male 8 Wild type A/J amygdala
E Male 8 Wild type A/J amygdala
F Male 8 Wild type C57BL/6J amygdala
G Male 8 Wild type C57BL/6J amygdala
H Male 8 Wild type 129S6/SvEvTac cingulate cortex
I Male 8 Wild type 129S6/SvEvTac cingulate cortex
J Male 8 Wild type A/J cingulate cortex
K Male 8 Wild type A/J cingulate cortex
L Male 8 Wild type A/J cingulate cortex
M Male 8 Wild type C57BL/6J cingulate cortex
N Male 8 Wild type C57BL/6J cingulate cortex
O Male 8 Wild type 129S6/SvEvTac hippocampus
P Male 8 Wild type 129S6/SvEvTac hippocampus
Q Male 8 Wild type A/J hippocampus
R Male 8 Wild type A/J hippocampus
S Male 8 Wild type C57BL/6J hippocampus
T Male 8 Wild type C57BL/6J4 hippocampus
U Male 8 Wild type 129S6/SvEvTac hypothalamus
V Male 8 Wild type 129S6/SvEvTac hypothalamus
W Male 8 Wild type A/J hypothalamus
X Male 8 Wild type A/J hypothalamus
Y Male 8 Wild type C57BL/6J hypothalamus
Z Male 8 Wild type C57BL/6J hypothalamus
3 Create a MetaData object from the metadata in the mouseSampleData.txt file.
MDObj2 = MetaData('File', 'mouseSampleData.txt', 'VarDescChar', '#')

Sample Names:
A, B, ...,Z (26 total)
Variable Names and Meta Information:

VariableDescription
Gender ' Gender of the mouse in study'
Age ' The number of weeks since mouse birth'
Type ' Genetic characters'
Strain ' The mouse strain'
Source ' The tissue source for RNA collection'

For complete information on constructing MetaData objects, see MetaData class.

4-14
 Representing Sample and Feature Metadata in MetaData Objects

Using Properties of a MetaData Object

To access properties of a MetaData object, use the following syntax:

objectname.propertyname

For example, to determine the number of variables in a MetaData object:

MDObj2.NVariables

ans =

To set properties of a MetaData object, use the following syntax:

objectname.propertyname = propertyvalue

For example, to set the Description property of a MetaData object:

MDObj1.Description = 'This is my MetaData object for my sample metadata'

Note Property names are case sensitive. For a list and description of all properties of a MetaData
object, see MetaData class.

Using Methods of a MetaData Object

To use methods of a MetaData object, use either of the following syntaxes:

objectname.methodname

or

methodname(objectname)

For example, to access the dataset array in a MetaData object that contains the variable values:

MDObj2.variableValues;

To access the dataset array of a MetaData object that contains the variable descriptions:

variableDesc(MDObj2)

ans =

VariableDescription
Gender ' Gender of the mouse in study'
Age ' The number of weeks since mouse birth'
Type ' Genetic characters'
Strain ' The mouse strain'
Source ' The tissue source for RNA collection'

Note For a complete list of methods of a MetaData object, see MetaData class.

4-15
4 Microarray Analysis

Representing Experiment Information in a MIAME Object

In this section...
“Overview of MIAME Objects” on page 4-16
“Constructing MIAME Objects” on page 4-16
“Using Properties of a MIAME Object” on page 4-17
“Using Methods of a MIAME Object” on page 4-18

Overview of MIAME Objects

You can store information about experimental methods and conditions from a microarray gene
expression experiment in a MIAME object. It loosely follows the Minimum Information About a
Microarray Experiment (MIAME) specification. It can include information about:

• Experiment design
• Microarrays used
• Samples used
• Sample preparation and labeling
• Hybridization procedures and parameters
• Normalization controls
• Preprocessing information
• Data processing specifications

A MIAME object includes properties and methods that let you access, retrieve, and change
experiment information related to a microarray experiment. These properties and methods are useful
to view and analyze the information. For a list of the properties and methods, see MIAME class.

Constructing MIAME Objects

For complete information on constructing MIAME objects, see MIAME class.

Constructing a MIAME Object from a GEO Structure

1 Import the bioma.data package so that the MIAME constructor function is available.

import bioma.data.*
2 Use the getgeodata function to return a MATLAB structure containing Gene Expression
Omnibus (GEO) Series data related to accession number GSE4616.

geoStruct = getgeodata('GSE4616')

geoStruct =

Header: [1x1 struct]

Data: [12488x12 bioma.data.DataMatrix]
3 Use the MIAME constructor function to create a MIAME object from the structure.

MIAMEObj1 = MIAME(geoStruct);

4-16
 Representing Experiment Information in a MIAME Object

4 Display information about the MIAME object, MIAMEObj.

MIAMEObj1

MIAMEObj1 =

Experiment Description:
Author name: Mika,,Silvennoinen
Riikka,,Kivelä
Maarit,,Lehti
Anna-Maria,,Touvras
Jyrki,,Komulainen
Veikko,,Vihko
Heikki,,Kainulainen
Laboratory: LIKES - Research Center
Contact information: Mika,,Silvennoinen
URL:
PubMedIDs: 17003243
Abstract: A 90 word abstract is available. Use the Abstract property.
Experiment Design: A 234 word summary is available. Use the ExptDesign property.
Other notes:
[1x80 char]

Constructing a MIAME Object from Properties

1 Import the bioma.data package so that theMIAME constructor function is available.

import bioma.data.*
2 Use the MIAME constructor function to create a MIAME object using individual properties.
MIAMEObj2 = MIAME('investigator', 'Jane Researcher',...
'lab', 'One Bioinformatics Laboratory',...
'contact', '[email protected]',...
'url', 'www.lab.not.exist',...
'title', 'Normal vs. Diseased Experiment',...
'abstract', 'Example of using expression data',...
'other', {'Notes:Created from a text file.'});

3 Display information about the MIAME object, MIAMEObj2.

MIAMEObj2

MIAMEObj2 =

Experiment Description:
Author name: Jane Researcher
Laboratory: One Bioinformatics Laboratory
Contact information: [email protected]
URL: www.lab.not.exist
PubMedIDs:
Abstract: A 4 word abstract is available. Use the Abstract property.
No experiment design summary available.
Other notes:
'Notes:Created from a text file.'

Using Properties of a MIAME Object

To access properties of a MIAME object, use the following syntax:

objectname.propertyname

For example, to retrieve the PubMed identifier of publications related to a MIAME object:

MIAMEObj1.PubMedID

ans =

4-17
4 Microarray Analysis

17003243

To set properties of a MIAME object, use the following syntax:

objectname.propertyname = propertyvalue

For example, to set the Laboratory property of a MIAME object:

MIAMEObj1.Laboratory = 'XYZ Lab'

Note Property names are case sensitive. For a list and description of all properties of a MIAME
object, see MIAME class.

Using Methods of a MIAME Object

To use methods of a MIAME object, use either of the following syntaxes:

objectname.methodname

or

methodname(objectname)

For example, to determine if a MIAME object is empty:

MIAMEObj1.isempty

ans =

Note For a complete list of methods of a MIAME object, see MIAME class.

4-18
 Representing All Data in an ExpressionSet Object

Representing All Data in an ExpressionSet Object

In this section...
“Overview of ExpressionSet Objects” on page 4-19
“Constructing ExpressionSet Objects” on page 4-20
“Using Properties of an ExpressionSet Object” on page 4-21
“Using Methods of an ExpressionSet Object” on page 4-21

Overview of ExpressionSet Objects

You can store all microarray experiment data and information in one object by assembling the
following into an ExpressionSet object:

• One ExptData object containing expression values from a microarray experiment in one or more
DataMatrix objects
• One MetaData object containing sample metadata in two dataset arrays
• One MetaData object containing feature metadata in two dataset arrays
• One MIAME object containing experiment descriptions

The following graphic illustrates a typical ExpressionSet object and its component objects.

4-19
4 Microarray Analysis

Each element (DataMatrix object) in the ExpressionSet object has an element name. Also, there is
always one DataMatrix object whose element name is Expressions.

An ExpressionSet object lets you store, manage, and subset the data from a microarray gene
expression experiment. An ExpressionSet object includes properties and methods that let you access,
retrieve, and change data, metadata, and other information about the microarray experiment. These
properties and methods are useful to view and analyze the data. For a list of the properties and
methods, see ExpressionSet class.

Constructing ExpressionSet Objects

Note The following procedure assumes you have executed the example code in the previous sections:

4-20
 Representing All Data in an ExpressionSet Object

• “Representing Expression Data Values in ExptData Objects” on page 4-9

• “Representing Sample and Feature Metadata in MetaData Objects” on page 4-12
• “Representing Experiment Information in a MIAME Object” on page 4-16

1 Import the bioma package so that the ExpressionSet constructor function is available.

import bioma.*
2 Construct an ExpressionSet object from EDObj, an ExptData object, MDObj2, a MetaData object
containing sample variable information, and MIAMEObj, a MIAME object.

ESObj = ExpressionSet(EDObj, 'SData', MDObj2, 'EInfo', MIAMEObj1);

3 Display information about the ExpressionSet object, ESObj.

ESObj

ExpressionSet
Experiment Data: 500 features, 26 samples
Element names: Expressions
Sample Data:
Sample names: A, B, ...,Z (26 total)
Sample variable names and meta information:
Gender: Gender of the mouse in study
Age: The number of weeks since mouse birth
Type: Genetic characters
Strain: The mouse strain
Source: The tissue source for RNA collection
Feature Data: none
Experiment Information: use 'exptInfo(obj)'

For complete information on constructing ExpressionSet objects, see ExpressionSet class.

Using Properties of an ExpressionSet Object

To access properties of an ExpressionSet object, use the following syntax:

objectname.propertyname

For example, to determine the number of samples in an ExpressionSet object:

ESObj.NSamples

ans =

26

Note Property names are case sensitive. For a list and description of all properties of an
ExpressionSet object, see ExpressionSet class.

Using Methods of an ExpressionSet Object

To use methods of an ExpressionSet object, use either of the following syntaxes:

4-21
4 Microarray Analysis

objectname.methodname

or

methodname(objectname)

For example, to retrieve the sample variable names from an ExpressionSet object:

ESObj.sampleVarNames

ans =

'Gender' 'Age' 'Type' 'Strain' 'Source'

To retrieve the experiment information contained in an ExpressionSet object:

exptInfo(ESObj)

ans =

Experiment description
Author name: Mika,,Silvennoinen
Riikka,,Kivelä
Maarit,,Lehti
Anna-Maria,,Touvras
Jyrki,,Komulainen
Veikko,,Vihko
Heikki,,Kainulainen
Laboratory: XYZ Lab
Contact information: Mika,,Silvennoinen
URL:
PubMedIDs: 17003243
Abstract: A 90 word abstract is available Use the Abstract property.
Experiment Design: A 234 word summary is available Use the ExptDesign property.
Other notes:
[1x80 char]

Note For a complete list of methods of an ExpressionSet object, see ExpressionSet class.

4-22
 Analyzing Illumina Bead Summary Gene Expression Data

Analyzing Illumina Bead Summary Gene Expression Data

This example shows how to analyze Illumina® BeadChip gene expression summary data using
MATLAB® and Bioinformatics Toolbox™ functions.

Introduction

This example shows how to import and analyze gene expression data from the Illumina BeadChip
microarray platform. The data set in the example is from the study of gene expression profiles of
human spermatogenesis by Platts et al. [1]. The expression levels were measured on Illumina Sentrix
Human 6 (WG6) BeadChips.

The data from most microarray gene expression experiments generally consists of four components:
experiment data values, sample information, feature annotations, and information about the
experiment. This example uses microarray data, constructs each of the four components, assembles
them into an ExpressionSet object, and finds the differentially expressed genes. For more
examples about the ExpressionSet class, see “Working with Objects for Microarray Experiment
Data” on page 4-203.

Importing Experimental Data from the GEO Database

Samples were hybridized on three Illumina Sentrix Human 6 (WG6) BeadChips. Retrieve the GEO
Series data GSE6967 using getgeodata function.
TNGEOData = getgeodata('GSE6967')

TNGEOData =

struct with fields:

Header: [1×1 struct]

Data: [47293×13 bioma.data.DataMatrix]

The TNGEOData structure contains Header and Data fields. The Header field has two fields, Series
and Samples, containing a description of the experiment and samples respectively. The Data field
contains a DataMatrix of normalized summary expression levels from the experiment.

Determine the number of samples in the experiment.

nSamples = numel(TNGEOData.Header.Samples.geo_accession)

nSamples =

13

Inspect the sample titles from the Header.Samples field.

TNGEOData.Header.Samples.title'

ans =

4-23
4 Microarray Analysis

13×1 cell array

{'Teratozoospermic individual: Sample T2'}

{'Teratozoospermic individual: Sample T1'}
{'Teratozoospermic individual: Sample T6'}
{'Teratozoospermic individual: Sample T4'}
{'Teratozoospermic individual: Sample T8'}
{'Normospermic individual: Sample N11' }
{'Teratozoospermic individual: Sample T3'}
{'Teratozoospermic individual: Sample T7'}
{'Teratozoospermic individual: Sample T5'}
{'Normospermic individual: Sample N6' }
{'Normospermic individual: Sample N12' }
{'Normospermic individual: Sample N5' }
{'Normospermic individual: Sample N1' }

For simplicity, extract the sample labels from the sample titles.
sampleLabels = cellfun(@(x) char(regexp(x, '\w\d+', 'match')),...
TNGEOData.Header.Samples.title, 'UniformOutput',false)

sampleLabels =

1×13 cell array

Columns 1 through 7

{'T2'} {'T1'} {'T6'} {'T4'} {'T8'} {'N11'} {'T3'}

Columns 8 through 13

{'T7'} {'T5'} {'N6'} {'N12'} {'N5'} {'N1'}

Importing Expression Data from Illumina BeadStudio Summary Files

Download the supplementary file GSE6967_RAW.tar and unzip the file to access the 13 text files
produced by the BeadStudio software, which contain the raw, non-normalized bead summary values.

The raw data files are named with their GSM accession numbers. For this example, construct the file
names of the text data files using the path where the text files are located.
rawDataFiles = cell(1,nSamples);
for i = 1:nSamples
rawDataFiles {i} = [TNGEOData.Header.Samples.geo_accession{i} '.txt'];
end

Set the variable rawDataPath to the path and directory to which you extracted the data files.
rawDataPath = 'C:\Examples\illuminagedemo\GSE6967';

Use the ilmnbsread function to read the first of the summary files and store the content in a
structure.
rawData = ilmnbsread(fullfile(rawDataPath, rawDataFiles{1}))

rawData =

4-24
 Analyzing Illumina Bead Summary Gene Expression Data

struct with fields:

Header: [1×1 struct]

TargetID: {47293×1 cell}
ColumnNames: {1×8 cell}
Data: [47293×8 double]
TextColumnNames: {}
TextData: {}

Inspect the column names in the rawData structure.

rawData.ColumnNames'

ans =

8×1 cell array

{'MIN_Signal-1412091085_A' }
{'AVG_Signal-1412091085_A' }
{'MAX_Signal-1412091085_A' }
{'NARRAYS-1412091085_A' }
{'ARRAY_STDEV-1412091085_A'}
{'BEAD_STDEV-1412091085_A' }
{'Avg_NBEADS-1412091085_A' }
{'Detection-1412091085_A' }

Determine the number of target probes.

nTargets = size(rawData.Data, 1)

nTargets =

47293

Read the non-normalized expression values (Avg_Signal), the detection confidence limits and the
Sentrix chip IDs from the summary data files. The gene expression values are identified with Illumina
probe target IDs. You can specify the columns to read from the data file.

rawMatrix = bioma.data.DataMatrix(zeros(nTargets, nSamples),...

rawData.TargetID, sampleLabels);
detectionConf = bioma.data.DataMatrix(zeros(nTargets, nSamples),...
rawData.TargetID, sampleLabels);
chipIDs = categorical([]);

for i = 1:nSamples
rawData =ilmnbsread(fullfile(rawDataPath, rawDataFiles{i}),...
'COLUMNS', {'AVG_Signal', 'Detection'});
chipIDs(i) = regexp(rawData.ColumnNames(1), '\d*', 'match', 'once');
rawMatrix(:, i) = rawData.Data(:, 1);
detectionConf(:,i) = rawData.Data(:,2);
end

4-25
4 Microarray Analysis

There are three Sentrix BeadChips used in the experiment. Inspect the Illumina Sentrix BeadChip IDs
in chipIDs to determine the number of samples hybridized on each chip.

summary(chipIDs)

samplesChip1 = sampleLabels(chipIDs=='1412091085')
samplesChip2 = sampleLabels(chipIDs=='1412091086')
samplesChip3 = sampleLabels(chipIDs=='1477791158')

1412091085 1412091086 1477791158

6 4 3

samplesChip1 =

1×6 cell array

{'T2'} {'T1'} {'T6'} {'T4'} {'T8'} {'N11'}

samplesChip2 =

1×4 cell array

{'T3'} {'T7'} {'T5'} {'N6'}

samplesChip3 =

1×3 cell array

{'N12'} {'N5'} {'N1'}

Six samples (T2, T1, T6, T4, T8, and N11) were hybridized to six arrays on the first chip, four samples
(T3, T7, T5, and N6) on the second chip, and three samples (N12, N5, and N1) on the third chip.

Normalizing the Expression Data

Use a boxplot to view the raw expression levels of each sample in the experiment.

logRawExprs = log2(rawMatrix);

maboxplot(logRawExprs, 'Orientation', 'horizontal')

ylabel('Sample Labels')
xlabel('log2(Expression Value)')
title('Before Normalization')

4-26
 Analyzing Illumina Bead Summary Gene Expression Data

The difference in intensities between samples on the same chip and samples on different chips does
not seem too large. The first BeadChip, containing samples T2, T1, T6, T4, T8 and N11, seems to be
slightly more variable than others.

Using MA and XY plots to do a pairwise comparison of the arrays on a BeadChip can be informative.
On an MA plot, the average (A) of the expression levels of two arrays are plotted on the x axis, and
the difference (M) in the measurement on the y axis. An XY plot is a scatter plot of one array against
another. This example uses the helper function maxyplot to plot MAXY plots for a pairwise
comparison of the three arrays on the first chip hybridized with teratozoospermic samples (T2, T1
and T6).

Note: You can also use the mairplot function to create the MA or IR (Intensity/Ratio) plots for
comparison of specific arrays.

inspectIdx = 1:3;
maxyplot(rawMatrix, inspectIdx)
sgtitle('Before Normalization')

4-27
4 Microarray Analysis

In an MAXY plot, the MA plots for all pairwise comparisons are in the upper-right corner. In the
lower-left corner are the XY plots of the comparisons. The MAXY plot shows the two arrays, T1 and
T2, to be quite similar, while different from the other array, T6.

The expression box plots and MAXY plots reveal that there are differences in expression levels within
chips and between chips; hence, the data requires normalization. Use the quantilenorm function to
apply quantile normalization to the raw data.

Note: You can also try invariant set normalization using the mainvarsetnorm function.

normExprs = rawMatrix;
normExprs(:, :) = quantilenorm(rawMatrix.(':')(':'));

log2NormExprs = log2(normExprs);

Display and inspect the normalized expression levels in a box plot.

figure;
maboxplot(log2NormExprs, 'ORIENTATION', 'horizontal')
ylabel('Sample Labels')
xlabel('log2(Expression Value)')
title('After Quantile Normalization')

4-28
 Analyzing Illumina Bead Summary Gene Expression Data

Display and inspect the MAXY plot of the three arrays (T2, T1 and T6) on the first chip after the
normalization.

maxyplot(normExprs, inspectIdx)
sgtitle('After Quantile Normalization')

4-29
4 Microarray Analysis

Many of the genes in this study are not expressed, or have only small variability across the samples.

First, you can remove genes with very low absolute expression values by using genelowvalfilter.

[mask, log2NormExprs] = genelowvalfilter(log2NormExprs);

detectionConf = detectionConf(mask, :);

Second, filter out genes with a small variance across samples using genevarfilter.

[mask, log2NormExprs] = genevarfilter(log2NormExprs);

detectionConf = detectionConf(mask, :);

Importing Feature Metadata from a BeadChip Annotation File

Microarray manufacturers usually provide annotations of a collection of features for each type of
chip. The chip annotation files contain metadata such as the gene name, symbol, NCBI accession
number, chromosome location and pathway information. Before assembling an ExpressionSet
object for the experiment, obtain the annotations about the features or probes on the BeadChip. You
can download the Human_WG-6.csv annotation file for Sentrix Human 6 (WG6) BeadChips from the
Support page at the Illumina web site and save the file locally. Read the annotation file into MATLAB
as a dataset array. Set the variable annotPath to the path and directory to which you downloaded
the annotation file.

annotPath = fullfile('C:\Examples\illuminagedemo\Annotation');
WG6Annot = dataset('xlsfile', fullfile(annotPath, 'Human_WG-6.csv'));

Inspect the properties of this dataset array.

4-30
 Analyzing Illumina Bead Summary Gene Expression Data

get(WG6Annot)

Description: ''
VarDescription: {}
Units: {}
DimNames: {'Observations' 'Variables'}
UserData: []
ObsNames: {}
VarNames: {1×13 cell}

Get the names of variables describing the features on the chip.

fDataVariables = get(WG6Annot, 'VarNames')

fDataVariables =

1×13 cell array

Columns 1 through 5

{'Search_key'} {'Target'} {'ProbeId'} {'Gid'} {'Transcript'}

Columns 6 through 10

{'Accession'} {'Symbol'} {'Type'} {'Start'} {'Probe_Sequence'}

Columns 11 through 13

{'Definition'} {'Ontology'} {'Synonym'}

Check the number of probe target IDs in the annotation file.

numel(WG6Annot.Target)

ans =

47296

Because the expression data in this example is only a small set of the full expression values, you will
work with only the features in the DataMatrix object log2NormExprs. Find the matching features
in log2NormExprs and WG6Annot.Target.
[commTargets, fI, WGI] = intersect(rownames(log2NormExprs), WG6Annot.Target);

Building an ExpressionSet Object for Experimental Data

You can store the preprocessed expression values and detection limits of the annotated probe targets
as an ExptData object.
fNames = rownames(log2NormExprs);
TNExptData = bioma.data.ExptData({log2NormExprs(fI, :), 'ExprsValues'},...
{detectionConf(fI, :), 'DetectionConfidences'})

TNExptData =

4-31
4 Microarray Analysis

Experiment Data:
42313 features, 13 samples
2 elements
Element names: ExprsValues, DetectionConfidences

Building an ExpressionSet Object for Sample Information

The sample data in the Header.Samples field of the TNGEOData structure can be overwhelming and
difficult to navigate through. From the data in Header.Samples field, you can gather the essential
information about the samples, such as the sample titles, GEO sample accession numbers, etc., and
store the sample data as a MetaData object.

Retrieve the descriptions about sample characteristics.

sampleChars = cellfun(@(x) char(regexp(x, '\w*tile', 'match')),...

TNGEOData.Header.Samples.characteristics_ch1, 'UniformOutput', false)

sampleChars =

1×13 cell array

Columns 1 through 4

{'Infertile'} {'Infertile'} {'Infertile'} {'Infertile'}

Columns 5 through 8

{'Infertile'} {'Fertile'} {'Infertile'} {'Infertile'}

Columns 9 through 13

{'Infertile'} {'Fertile'} {'Fertile'} {'Fertile'} {'Fertile'}

Create a dataset array to store the sample data you just extracted.

sampleDS = dataset({TNGEOData.Header.Samples.geo_accession', 'GSM'},...

{strtok(TNGEOData.Header.Samples.title)', 'Type'},...
{sampleChars', 'Characteristics'}, 'obsnames', sampleLabels')

sampleDS =

GSM Type Characteristics

T2 {'GSM160620'} {'Teratozoospermic'} {'Infertile'}
T1 {'GSM160621'} {'Teratozoospermic'} {'Infertile'}
T6 {'GSM160622'} {'Teratozoospermic'} {'Infertile'}
T4 {'GSM160623'} {'Teratozoospermic'} {'Infertile'}
T8 {'GSM160624'} {'Teratozoospermic'} {'Infertile'}
N11 {'GSM160625'} {'Normospermic' } {'Fertile' }
T3 {'GSM160626'} {'Teratozoospermic'} {'Infertile'}
T7 {'GSM160627'} {'Teratozoospermic'} {'Infertile'}
T5 {'GSM160628'} {'Teratozoospermic'} {'Infertile'}
N6 {'GSM160629'} {'Normospermic' } {'Fertile' }
N12 {'GSM160630'} {'Normospermic' } {'Fertile' }
N5 {'GSM160631'} {'Normospermic' } {'Fertile' }

4-32
 Analyzing Illumina Bead Summary Gene Expression Data

N1 {'GSM160632'} {'Normospermic' } {'Fertile' }

Store the sample metadata as an object of the MetaData class, including a short description for each
variable.

TNSData = bioma.data.MetaData(sampleDS,...
{'Sample GEO accession number',...
'Spermic type',...
'Fertility characteristics'})

TNSData =

Sample Names:
T2, T1, ...,N1 (13 total)
Variable Names and Meta Information:
VariableDescription
GSM {'Sample GEO accession number'}
Type {'Spermic type' }
Characteristics {'Fertility characteristics' }

Building an ExpressionSet Object for Feature Annotations

The collection of feature metadata for Sentrix Human 6 BeadChips is large and diverse. Select
information about features that are unique to the experiment and save the information as a
MetaData object. Extract annotations of interest, for example, Accession and Symbol.

fIdx = ismember(fDataVariables, {'Accession', 'Symbol'});

featureAnnot = WG6Annot(WGI, fDataVariables(fIdx));

featureAnnot = set(featureAnnot, 'ObsNames', WG6Annot.Target(WGI));

Create a MetaData object for the feature annotation information with brief descriptions about the
two variables of the metadata.

WG6FData = bioma.data.MetaData(featureAnnot, ...

{'Accession number of probe target', 'Gene Symbol of probe target'})

WG6FData =

Sample Names:
GI_10047089-S, GI_10047091-S, ...,hmm9988-S (42313 total)
Variable Names and Meta Information:
VariableDescription
Accession {'Accession number of probe target'}
Symbol {'Gene Symbol of probe target' }

Building an ExpressionSet Object for Experiment Information

Most of the experiment descriptions in the Header.Series field of the TNGEOData structure can be
reorganized and stored as a MIAME object, which you will use to assemble the ExpressionSet object
for the experiment.

TNExptInfo = bioma.data.MIAME(TNGEOData)

4-33
4 Microarray Analysis

TNExptInfo =

Experiment Description:
Author name: Adrian,E,Platts
David,J,Dix
Hector,E,Chemes
Kary,E,Thompson
Robert,,Goodrich
John,C,Rockett
Vanesa,Y,Rawe
Silvina,,Quintana
Michael,P,Diamond
Lillian,F,Strader
Stephen,A,Krawetz
Laboratory: Wayne State University
Contact information: Stephen,A,Krawetz
URL: http://compbio.med.wayne.edu
PubMedIDs: 17327269
Abstract: A 82 word abstract is available. Use the Abstract property.
Experiment Design: A 61 word summary is available. Use the ExptDesign property.
Other notes:
{'ftp://ftp.ncbi.nlm.nih.gov/geo/series/GSE6nnn/GSE6967/suppl/GSE6967_RAW.tar'}

Assembling an ExpressionSet Object

Now that you've created all the components, you can create an object of the ExpressionSet class to
store the expression values, sample information, chip feature annotations and description information
about this experiment.

TNExprSet = bioma.ExpressionSet(TNExptData, 'sData', TNSData,...

'fData', WG6FData,...
'eInfo', TNExptInfo)

TNExprSet =

ExpressionSet
Experiment Data: 42313 features, 13 samples
Element names: Expressions, DetectionConfidences
Sample Data:
Sample names: T2, T1, ...,N1 (13 total)
Sample variable names and meta information:
GSM: Sample GEO accession number
Type: Spermic type
Characteristics: Fertility characteristics
Feature Data:
Feature names: GI_10047089-S, GI_10047091-S, ...,hmm9988-S (42313 total)
Feature variable names and meta information:
Accession: Accession number of probe target
Symbol: Gene Symbol of probe target
Experiment Information: use 'exptInfo(obj)'

Note: The ExprsValues element in the ExptData object, TNExptData, is renamed to

Expressions in TNGeneExprSet.

You can save an object of ExpressionSet class as a MAT file for further data analysis.

4-34
 Analyzing Illumina Bead Summary Gene Expression Data

save TNGeneExprSet TNExprSet

Profiling Gene Expression by Using Permutation T-tests

Load the experiment data saved from the previous section. You will use this datas to find differentially
expressed genes between the teratozoospermia and normal samples.

load TNGeneExprSet

To identify the differential changes in the levels of transcripts in normospermic Ns and

teratozoospermic Tz samples, compare the gene expression values between the two groups of data:
Tz and Ns.

TNSamples = sampleNames(TNExprSet);
Tz = strncmp(TNSamples, 'T', 1);
Ns = strncmp(TNSamples, 'N', 1);
nTz = sum(Tz)
nNs = sum(Ns)

TNExprs = expressions(TNExprSet);
TzData = TNExprs(:,Tz);
NsData = TNExprs(:,Ns);
meanTzData = mean(TzData,2);
meanNsData = mean(NsData,2);
groupLabels = [TNSamples(Tz), TNSamples(Ns)];

nTz =

nNs =

Perform a permutation t-test using the mattest function to permute the columns of the gene
expression data matrix of TzData and NsData. Note: Depending on the sample size, it may not be
feasible to consider all possible permutations. Usually, a random subset of permutations are
considered in the case of a large sample size.

Use the nchoosek function in Statistics and Machine Learning Toolbox™ to determine the number of
all possible permutations of the samples in this example.

perms = nchoosek(1:nTz+nNs, nTz);

nPerms = size(perms,1)

nPerms =

1287

Use the PERMUTE option of the mattest function to compute the p-values of all the permutations.

pValues = mattest(TzData, NsData, 'Permute', nPerms);

4-35
4 Microarray Analysis

You can also compute the differential score from the p-values using the following anonymous function
[1].

diffscore = @(p, avgTest, avgRef) -10*sign(avgTest - avgRef).*log10(p);

A differential score of 13 corresponds to a p-value of 0.05, a differential score of 20 corresponds to a

p-value of 0.01, and a differential score of 30 corresponds to a p-value of 0.001. A positive differential
score represents up-regulation, while a negative score represents down-regulation of genes.

diffScores = diffscore(pValues, meanTzData, meanNsData);

Determine the number of genes considered to have a differential score greater than 20. Note: You
may get a different number of genes due to the permutation test outcome.

up = sum(diffScores > 20)

down = sum(diffScores < -20)

up =

3741

down =

3033

Estimating False Discovery Rate (FDR)

In multiple hypothesis testing, where we simultaneously tests the null hypothesis of thousands of
genes, each test has a specific false positive rate, or a false discovery rate (FDR) [2]. Estimate the
FDR and q-values for each test using the mafdr function.

figure;
[pFDR, qValues] = mafdr(pValues, 'showplot', true);
diffScoresFDRQ = diffscore(qValues, meanTzData, meanNsData);

4-36
 Analyzing Illumina Bead Summary Gene Expression Data

Determine the number of genes with an absolute differential score greater than 20. Note: You may
get a different number of genes due to the permutation test and the bootstrap outcomes.

sum(abs(diffScoresFDRQ)>=20)

ans =

3122

Identifying Genes that Are Differentially Expressed

Plot the -log10 of p-values against fold changes in a volcano plot.

diffStruct = mavolcanoplot(TzData, NsData, qValues,...

'pcutoff', 0.01, 'foldchange', 5);

4-37
4 Microarray Analysis

Note: From the volcano plot UI, you can interactively change the p-value cutoff and fold-change limit,
and export differentially expressed genes.

Determine the number of differentially expressed genes.

nDiffGenes = numel(diffStruct.GeneLabels)

nDiffGenes =

451

Get the list of up-regulated genes for the Tz samples compared to the Ns samples.

up_genes = diffStruct.GeneLabels(diffStruct.FoldChanges > 0);

nUpGenes = length(up_genes)

nUpGenes =

223

Get the list of down-regulated genes for the Tz samples compared to the Ns samples.

4-38
 Analyzing Illumina Bead Summary Gene Expression Data

down_genes = diffStruct.GeneLabels(diffStruct.FoldChanges < 0);

nDownGenes = length(down_genes)

nDownGenes =

228

Extract a list of differentially expressed genes.

diff_geneidx = zeros(nDiffGenes, 1);

for i = 1:nDiffGenes
diff_geneidx(i) = find(strncmpi(TNExprSet.featureNames, ...
diffStruct.GeneLabels{i}, length(diffStruct.GeneLabels{i})), 1);
end

You can get the subset of experiment data containing only the differentially expressed genes.

TNDiffExprSet = TNExprSet(diff_geneidx, groupLabels);

Performing PCA and Clustering Analysis of Significant Gene Profiles

Principal component analysis (PCA) on differentially expressed genes shows linear separability of the
Tz samples from the Ns samples.

PCAScore = pca(TNDiffExprSet.expressions);

Display the coefficients of the first and sixth principal components.

figure;
plot(PCAScore(:,1), PCAScore(:,6), 's', 'MarkerSize',10, 'MarkerFaceColor','g');
hold on
text(PCAScore(:,1)+0.02, PCAScore(:,6), TNDiffExprSet.sampleNames)
plot([0,0], [-0.5 0.5], '--r')
ax = gca;
ax.XTick = [];
ax.YTick = [];
ax.YTickLabel = [];
title('PCA Mapping')
xlabel('Principal Component 1')
ylabel('Principal Component 6')

4-39
4 Microarray Analysis

You can also use the interactive tool created by the mapcaplot function to perform principal
component analysis.

mapcaplot((TNDiffExprSet.expressions)')

4-40
 Analyzing Illumina Bead Summary Gene Expression Data

Perform unsupervised hierarchical clustering of the significant gene profiles from the Tz and Ns
groups using correlation as the distance metric to cluster the samples.
sampleDist = pdist(TNDiffExprSet.expressions','correlation');
sampleLink = linkage(sampleDist);

figure;
dendrogram(sampleLink, 'labels', TNDiffExprSet.sampleNames,'ColorThreshold',0.5)

4-41
4 Microarray Analysis

ax = gca;
ax.YTick = [];
ax.Box = 'on';
title('Hierarchical Sample Clustering')

Use the clustergram function to create the hierarchical clustering of differentially expressed genes,
and apply the colormap redbluecmap to the clustergram.

cmap = redbluecmap(9);
cg = clustergram(TNDiffExprSet.expressions,'Colormap',cmap,'Standardize',2);
addTitle(cg,'Hierarchical Gene Clustering')

4-42
 Analyzing Illumina Bead Summary Gene Expression Data

Clustering of the most differentially abundant transcripts clearly partitions teratozoospermic (Tz) and
normospermic (Ns) spermatozoal RNAs.

References

[1] Platts, A.E., et al., "Success and failure in human spermatogenesis as revealed by
teratozoospermic RNAs", Human Molecular Genetics, 16(7):763-73, 2007.

[2] Storey, J.D. and Tibshirani, R., "Statistical significance for genomewide studies", PNAS,
100(16):9440-5, 2003.

4-43
4 Microarray Analysis

Detecting DNA Copy Number Alteration in Array-Based CGH

Data

This example shows how to detect DNA copy number alterations in genome-wide array-based
comparative genomic hybridization (CGH) data.

Introduction

Copy number changes or alterations is a form of genetic variation in the human genome [1]. DNA
copy number alterations (CNAs) have been linked to the development and progression of cancer and
many diseases.

DNA microarray based comparative genomic hybridization (CGH) is a technique allows simultaneous
monitoring of copy number of thousands of genes throughout the genome [2,3]. In this technique,
DNA fragments or "clones" from a test sample and a reference sample differentially labeled with dyes
(typically, Cy3 and Cy5) are hybridized to mapped DNA microarrays and imaged. Copy number
alterations are related to the Cy3 and Cy5 fluorescence intensity ratio of the targets hybridized to
each probe on a microarray. Clones with normalized test intensities significantly greater than
reference intensities indicate copy number gains in the test sample at those positions. Similarly,
significantly lower intensities in the test sample are signs of copy number loss. BAC (bacterial
artificial chromosome) clone based CGH arrays have a resolution in the order of one million base
pairs (1Mb) [3]. Oligonucleotide and cDNA arrays provide a higher resolution of 50-100kb [2].

Array CGH log2-based intensity ratios provide useful information about genome-wide CNAs. In
humans, the normal DNA copy number is two for all the autosomes. In an ideal situation, the normal
clones would correspond to a log2 ratio of zero. The log2 intensity ratios of a single copy loss would
be -1, and a single copy gain would be 0.58. The goal is to effectively identify locations of gains or
losses of DNA copy number.

The data in this example is the Coriell cell line BAC array CGH data analyzed by Snijders et al.(2001).
The Coriell cell line data is widely regarded as a "gold standard" data set. You can download this data
of normalized log2-based intensity ratios and the supplemental table of known karyotypes from
https://www.nature.com/articles/ng754#supplementary-information. You will compare these
cytogenically mapped alterations with the locations of gains or losses identified with various functions
of MATLAB and its toolboxes.

For this example, the Coriell cell line data are provided in a MAT file. The data file
coriell_baccgh.mat contains coriell_data, a structure containing of the normalized average of
the log2-based test to reference intensity ratios of 15 fibroblast cell lines and their genomic positions.
The BAC targets are ordered by genome position beginning at 1p and ending at Xq.
load coriell_baccgh
coriell_data

coriell_data =

struct with fields:

Sample: {1x15 cell}

Chromosome: [2285x1 int8]
GenomicPosition: [2285x1 int32]
Log2Ratio: [2285x15 double]

4-44
 Detecting DNA Copy Number Alteration in Array-Based CGH Data

FISHMap: {2285x1 cell}

Visualizing the Genome Profile of the Array CGH Data Set

You can plot the genome wide log2-based test/reference intensity ratios of DNA clones. In this
example, you will display the log2 intensity ratios for cell line GM03576 for chromosomes 1 through
23.

Find the sample index for the CM03576 cell line.

sample = find(strcmpi(coriell_data.Sample, 'GM03576'))

sample =

To label chromosomes and draw the chromosome borders, you need to find the number of data points
of in each chromosome.

chr_nums = zeros(1, 23);

chr_data_len = zeros(1,23);
for c = 1:23
tmp = coriell_data.Chromosome == c;
chr_nums(c) = find(tmp, 1, 'last');
chr_data_len(c) = length(find(tmp));
end

% Draw a vertical bar at the end of a chromosome to indicate the border

x_vbar = repmat(chr_nums, 3, 1);
y_vbar = repmat([2;-2;NaN], 1, 23);

% Label the autosomes with their chromosome numbers, and the sex chromosome
% with X.
x_label = chr_nums - ceil(chr_data_len/2);
y_label = zeros(1, length(x_label)) - 1.6;
chr_labels = num2str((1:1:23)');
chr_labels = cellstr(chr_labels);
chr_labels{end} = 'X';

figure
hold on
h_ratio = plot(coriell_data.Log2Ratio(:,sample), '.');
h_vbar = line(x_vbar, y_vbar, 'color', [0.8 0.8 0.8]);
h_text = text(x_label, y_label, chr_labels,...
'fontsize', 8, 'HorizontalAlignment', 'center');

h_axis = h_ratio.Parent;
h_axis.XTick = [];
h_axis.YGrid = 'on';
h_axis.Box = 'on';
xlim([0 chr_nums(23)])
ylim([-1.5 1.5])

title(coriell_data.Sample{sample})
xlabel({'', 'Chromosome'})

4-45
4 Microarray Analysis

ylabel('Log2(T/R)')
hold off

In the plot, borders between chromosomes are indicated by grey vertical bars. The plot indicates that
the GM03576 cell line is trisomic for chromosomes 2 and 21 [3].

You can also plot the profile of each chromosome in a genome. In this example, you will display the
log2 intensity ratios for each chromosome in cell line GM05296 individually.

sample = find(strcmpi(coriell_data.Sample, 'GM05296'));

figure;
for c = 1:23
idx = coriell_data.Chromosome == c;
chr_y = coriell_data.Log2Ratio(idx, sample);
subplot(5,5,c);

hp = plot(chr_y, '.');
line([0, chr_data_len(c)], [0,0], 'color', 'r');

h_axis = hp.Parent;
h_axis.XTick = [];
h_axis.Box = 'on';
xlim([0 chr_data_len(c)])
ylim([-1.5 1.5])
xlabel(['chr ' chr_labels{c}], 'FontSize', 8)
end
sgtitle('GM05296');

4-46
 Detecting DNA Copy Number Alteration in Array-Based CGH Data

The plot indicates the GM05296 cell line has a partial trisomy at chromosome 10 and a partial
monosomy at chromosome 11.

Observe that the gains and losses of copy number are discrete. These alterations occur in contiguous
regions of a chromosome that cover several clones to entitle chromosome.

The array-based CGH data can be quite noisy. Therefore, accurate identification of chromosome
regions of equal copy number that accounts for the noise in the data requires robust computational
methods. In the rest of this example, you will work with the data of chromosomes 9, 10 and 11 of the
GM05296 cell line.

Initialize a structure array for the data of these three chromosomes.

GM05296_Data = struct('Chromosome', {9 10 11},...
'GenomicPosition', {[], [], []},...
'Log2Ratio', {[], [], []},...
'SmoothedRatio', {[], [], []},...
'DiffRatio', {[], [], []},...
'SegIndex', {[], [], []});

Filtering and Smoothing Data

A simple approach to perform high-level smoothing is to use a nonparametric filter. The function
mslowess implements a linear fit to samples within a shifting window, is this example you use a SPAN
of 15 samples.
for iloop = 1:length(GM05296_Data)
idx = coriell_data.Chromosome == GM05296_Data(iloop).Chromosome;

4-47
4 Microarray Analysis

chr_x = coriell_data.GenomicPosition(idx);
chr_y = coriell_data.Log2Ratio(idx, sample);

% Remove NaN data points

idx = ~isnan(chr_y);
GM05296_Data(iloop).GenomicPosition = double(chr_x(idx));
GM05296_Data(iloop).Log2Ratio = chr_y(idx);

% Smoother
GM05296_Data(iloop).SmoothedRatio = ...
mslowess(GM05296_Data(iloop).GenomicPosition,...
GM05296_Data(iloop).Log2Ratio,...
'SPAN',15);

% Find the derivative of the smoothed ratio

GM05296_Data(iloop).DiffRatio = ...
diff([0; GM05296_Data(iloop).SmoothedRatio]);
end

To better visualize and later validate the locations of copy number changes, we need cytoband
information. Read the human cytoband information from the hs_cytoBand.txt data file using the
cytobandread function. It returns a structure of human cytoband information [4].

hs_cytobands = cytobandread('hs_cytoBand.txt')

% Find the centromere positions for the chromosomes.

acen_idx = strcmpi(hs_cytobands.GieStains, 'acen');
acen_ends = hs_cytobands.BandEndBPs(acen_idx);

% Convert the cytoband data from bp to kilo bp because the genomic

% positions in Coriell Cell Line data set are in kilo base pairs.
acen_pos = acen_ends(1:2:end)/1000;

hs_cytobands =

struct with fields:

ChromLabels: {862x1 cell}

BandStartBPs: [862x1 int32]
BandEndBPs: [862x1 int32]
BandLabels: {862x1 cell}
GieStains: {862x1 cell}

You can inspect the data by plotting the log2-based ratios, the smoothed ratios and the derivative of
the smoothed ratios together. You can also display the centromere position of a chromosome in the
data plots. The magenta vertical bar marks the centromere of the chromosome.

for iloop = 1:length(GM05296_Data)

chr = GM05296_Data(iloop).Chromosome;
chr_x = GM05296_Data(iloop).GenomicPosition;
figure
hold on
plot(chr_x, GM05296_Data(iloop).Log2Ratio, '.');
line(chr_x, GM05296_Data(iloop).SmoothedRatio,...
'Color', 'r', 'LineWidth', 2);
line(chr_x, GM05296_Data(iloop).DiffRatio,...

4-48
 Detecting DNA Copy Number Alteration in Array-Based CGH Data

'Color', 'k', 'LineWidth', 2);

line([acen_pos(chr), acen_pos(chr)], [-1, 1],...
'Color', 'm', 'LineWidth', 2, 'LineStyle', '-.');
if iloop == 1
legend('Raw','Smoothed','Diff', 'Centromere');
end
ylim([-1, 1])
xlabel('Genomic Position')
ylabel('Log2(T/R)')
title(sprintf('GM05296: Chromosome %d ', chr))
hold off
end

4-49
4 Microarray Analysis

4-50
 Detecting DNA Copy Number Alteration in Array-Based CGH Data

Detecting Change-Points

The derivatives of the smoothed ratio over a certain threshold usually indicate substantial changes
with large peaks, and provide the estimate of the change-point indices. For this example you will
select a threshold of 0.1.

thrd = 0.1;

for iloop = 1:length(GM05296_Data)

idx = find(abs(GM05296_Data(iloop).DiffRatio) > thrd );
N = numel(GM05296_Data(iloop).SmoothedRatio);
GM05296_Data(iloop).SegIndex = [1;idx;N];

% Number of possible segments found

fprintf('%d segments initially found on Chromosome %d.\n',...
numel(GM05296_Data(iloop).SegIndex) - 1,...
GM05296_Data(iloop).Chromosome)
end

1 segments initially found on Chromosome 9.

4 segments initially found on Chromosome 10.
5 segments initially found on Chromosome 11.

Optimizing Change-Points by GM Clustering

Gaussian Mixture (GM) or Expectation-Maximization (EM) clustering can provide fine adjustments to
the change-point indices [5]. The convergence to statistically optimal change-point indices can be

4-51
4 Microarray Analysis

facilitated by surrounding each index with equal-length set of adjacent indices. Thus each edge is
associated with left and right distributions. The GM clustering learns the maximum-likelihood
parameters of the two distributions. It then optimally adjusts the indices given the learned
parameters.

You can set the length for the set of adjacent positions distributed around the change-point indices.
For this example, you will select a length of 5. You can also inspect each change-point by plotting its
GM clusters. In this example, you will plot the GM clusters for the Chromosome 10 data.

len = 5;
for iloop = 1:length(GM05296_Data)
seg_num = numel(GM05296_Data(iloop).SegIndex) - 1;
if seg_num > 1
% Plot the data points in chromosome 10 data
if GM05296_Data(iloop).Chromosome == 10
figure
hold on;
plot(GM05296_Data(iloop).GenomicPosition,...
GM05296_Data(iloop).Log2Ratio, '.')
ylim([-0.5, 1])
xlabel('Genomic Position')
ylabel('Log2(T/R)')
title(sprintf('Chromosome %d - GM05296', ...
GM05296_Data(iloop).Chromosome))
end

segidx = GM05296_Data(iloop).SegIndex;
segidx_emadj = GM05296_Data(iloop).SegIndex;

for jloop = 2:seg_num

ileft = min(segidx(jloop) - len, segidx(jloop));
iright = max(segidx(jloop) + len, segidx(jloop));
gmx = GM05296_Data(iloop).GenomicPosition(ileft:iright);
gmy = GM05296_Data(iloop).SmoothedRatio(ileft:iright);

% Select initial guess for the cluster index for each point.
gmpart = (gmy > (min(gmy) + range(gmy)/2)) + 1;

% Create a Gaussian mixture model object

gm = gmdistribution.fit(gmy, 2, 'start', gmpart);
gmid = cluster(gm,gmy);

segidx_emadj(jloop) = find(abs(diff(gmid))==1) + ileft;

% Plot GM clusters for the change-points in chromosome 10 data

if GM05296_Data(iloop).Chromosome == 10
plot(gmx(gmid==1),gmy(gmid==1), 'g.',...
gmx(gmid==2), gmy(gmid==2), 'r.')
end
end

% Remove repeat indices

zeroidx = [diff(segidx_emadj) == 0; 0];
GM05296_Data(iloop).SegIndex = segidx_emadj(~zeroidx);
end

% Number of possible segments found

4-52
 Detecting DNA Copy Number Alteration in Array-Based CGH Data

fprintf('%d segments found on Chromosome %d after GM clustering adjustment.\n',...

numel(GM05296_Data(iloop).SegIndex) - 1,...
GM05296_Data(iloop).Chromosome)
end
hold off;

1 segments found on Chromosome 9 after GM clustering adjustment.

3 segments found on Chromosome 10 after GM clustering adjustment.
5 segments found on Chromosome 11 after GM clustering adjustment.

Testing Change-Point Significance

Once you determine the optimal change-point indices, you also need to determine if each segment
represents a statistically significant changes in DNA copy number. You will perform permutation t-
tests to assess the significance of the segments identified. A segment includes all the data points from
one change-point to the next change-point or the chromosome end. In this example, you will perform
10,000 permutations of the data points on two consecutive segments along the chromosome at the
significance level of 0.01.

alpha = 0.01;
for iloop = 1:length(GM05296_Data)
seg_num = numel(GM05296_Data(iloop).SegIndex) - 1;
seg_index = GM05296_Data(iloop).SegIndex;
if seg_num > 1
ppvals = zeros(seg_num+1, 1);

for sloop = 1:seg_num-1

4-53
4 Microarray Analysis

seg1idx = seg_index(sloop):seg_index(sloop+1)-1;

if sloop== seg_num-1
seg2idx = seg_index(sloop+1):(seg_index(sloop+2));
else
seg2idx = seg_index(sloop+1):(seg_index(sloop+2)-1);
end

seg1 = GM05296_Data(iloop).SmoothedRatio(seg1idx);
seg2 = GM05296_Data(iloop).SmoothedRatio(seg2idx);

n1 = numel(seg1);
n2 = numel(seg2);
N = n1+n2;
segs = [seg1;seg2];

% Compute observed t statistics

t_obs = mean(seg1) - mean(seg2);

% Permutation test
iter = 10000;
t_perm = zeros(iter,1);
for i = 1:iter
randseg = segs(randperm(N));
t_perm(i) = abs(mean(randseg(1:n1))-mean(randseg(n1+1:N)));
end
ppvals(sloop+1) = sum(t_perm >= abs(t_obs))/iter;
end

sigidx = ppvals < alpha;

GM05296_Data(iloop).SegIndex = seg_index(sigidx);
end

% Number segments after significance tests

fprintf('%d segments found on Chromosome %d after significance tests.\n',...
numel(GM05296_Data(iloop).SegIndex) - 1, GM05296_Data(iloop).Chromosome)
end

1 segments found on Chromosome 9 after significance tests.

3 segments found on Chromosome 10 after significance tests.
4 segments found on Chromosome 11 after significance tests.

Assessing Copy Number Alterations

Cytogenetic study indicates cell line GM05296 has a trisomy at 10q21-10q24 and a monosomy at
11p12-11p13 [3]. Plot the segment means of the three chromosomes over the original data with bold
red lines, and add the chromosome ideograms to the plots using the chromosomeplot function. Note
that the genomic positions in the Coriell cell line data set are in kilo base pairs. Therefore, you will
need to convert cytoband data from bp to kilo bp when adding the ideograms to the plot.
for iloop = 1:length(GM05296_Data)
figure;
seg_num = numel(GM05296_Data(iloop).SegIndex) - 1;
seg_mean = ones(seg_num,1);
chr_num = GM05296_Data(iloop).Chromosome;
for jloop = 2:seg_num+1
idx = GM05296_Data(iloop).SegIndex(jloop-1):GM05296_Data(iloop).SegIndex(jloop);
seg_mean(idx) = mean(GM05296_Data(iloop).Log2Ratio(idx));

4-54
 Detecting DNA Copy Number Alteration in Array-Based CGH Data

line(GM05296_Data(iloop).GenomicPosition(idx), seg_mean(idx),...
'color', 'r', 'linewidth', 3);
end
line(GM05296_Data(iloop).GenomicPosition, GM05296_Data(iloop).Log2Ratio,...
'linestyle', 'none', 'Marker', '.');
line([acen_pos(chr_num), acen_pos(chr_num)], [-1, 1],...
'linewidth', 2,...
'color', 'm',...
'linestyle', '-.');

ylabel('Log2(T/R)')
ax = gca;
ax.Box = 'on';
ylim([-1, 1])
title(sprintf('Chromosome %d - GM05296', chr_num));
chromosomeplot(hs_cytobands, chr_num, 'addtoplot', gca, 'unit', 2)

end

4-55
4 Microarray Analysis

4-56
 Detecting DNA Copy Number Alteration in Array-Based CGH Data

As shown in the plots, no copy number alterations were found on chromosome 9, there is copy
number gain span from 10q21 to 10q24, and a copy number loss region from 11p12 to 11p13. The
CNAs found match the known results in cell line GM05296 determined by cytogenetic analysis.

You can also display the CNAs of the GM05296 cell line align to the chromosome ideogram summary
view using the chromosomeplot function. Determine the genomic positions for the CNAs on
chromosomes 10 and 11.

chr10_idx = GM05296_Data(2).SegIndex(2):GM05296_Data(2).SegIndex(3)-1;
chr10_cna_start = GM05296_Data(2).GenomicPosition(chr10_idx(1))*1000;
chr10_cna_end = GM05296_Data(2).GenomicPosition(chr10_idx(end))*1000;

chr11_idx = GM05296_Data(3).SegIndex(2):GM05296_Data(3).SegIndex(3)-1;
chr11_cna_start = GM05296_Data(3).GenomicPosition(chr11_idx(1))*1000;
chr11_cna_end = GM05296_Data(3).GenomicPosition(chr11_idx(end))*1000;

Create a structure containing the copy number alteration data from the GM05296 cell line data
according to the input requirements of the chromosomeplot function.

cna_struct = struct('Chromosome', [10 11],...

'CNVType', [2 1],...
'Start', [chr10_cna_start, chr11_cna_start],...
'End', [chr10_cna_end, chr11_cna_end])

cna_struct =

4-57
4 Microarray Analysis

struct with fields:

Chromosome: [10 11]

CNVType: [2 1]
Start: [69209000 34420000]
End: [105905000 35914000]

chromosomeplot(hs_cytobands, 'cnv', cna_struct, 'unit', 2)

title('Human Karyogram with Copy Number Alterations of GM05296')

This example shows how MATLAB and its toolboxes provide tools for the analysis and visualization of
copy-number alterations in array-based CGH data.

References

[1] Redon, R., et al., "Global variation in copy number in the human genome", Nature,
444(7118):444-54, 2006.

[2] Pinkel, D., et al., "High resolution analysis of DNA copy number variations using comparative
genomic hybridization to microarrays", Nature Genetics, 20(2):207-11, 1998.

[3] Snijders, A.M., et al., "Assembly of microarrays for genome-wide measurement of DNA copy
number", Nature Genetics, 29(3):263-4, 2001.

[4] Human Genome NCBI Build 36.

4-58
 Detecting DNA Copy Number Alteration in Array-Based CGH Data

[5] Myers, C.L., et al., "Accurate detection of aneuploidies in array CGH and gene expression
microarray data", Bioinformatics, 20(18):3533-43, 2004.

4-59
4 Microarray Analysis

Analyzing Array-Based CGH Data Using Bayesian Hidden

Markov Modeling

This example shows how to use a Bayesian hidden Markov model (HMM) technique to identify copy
number alteration in array-based comparative genomic hybridization (CGH) data.

Introduction

Array-based CGH is a high-throughput technique to measure DNA copy number change across the
genome. The DNA fragments or "clones" of test and reference samples are hybridized to mapped
array fragments. Log2 intensity ratios of test to reference provide useful information about genome-
wide profiles in copy number. In an ideal situation, the log2 ratio of normal (copy-neutral) clones is
log2(2/2) = 0, single copy losses is log2(1/2) = -1, and single copy gains is log2(3/2) = 0.58. Multiple
copy gains or amplifications would have values of log2(4/2), log2(5/2),.... Loss of both copies, or a
deletion would correspond to the value of -inf. In real applications, even after accounting for
measurement error, the log2 ratios differ considerably from the theoretical values. The ratios
typically shrink towards zero. One main factor is contamination of the tumor samples with normal
cells. There is also a dependence between the intensity ratios of neighboring clones. These issues
necessitate the use of efficient statistical algorithms characterizing the genomic profiles.

Bayesian HMM

Guha et al., (2006) proposed a Bayesian statistical approach depending on a hidden Markov model
(HMM) for analyzing array CGH data. The hidden Markov model accounts for the dependence
between neighboring clones. The intensity ratios are generated by hidden copy number states.
Bayesian learning is used to identify genome-wide changes in copy number from the data. Posterior
inferences are made about the copy number gains and losses.

In this Bayesian HMM algorithm, there are four states, defined as copy number loss state (1), copy
number neutral state (2), single copy gain state (3), and amplification (multiple gain) state (4). A copy
number state is associated with each clone. The normalized log2 ratios are assumed to be distributed
as

The mu is a unknown parameter for each state with this constraint:

The priors for mean copy number changes are:

Guha et al., (2006) also described an Metropolis-within-Gibbs algorithm to generate posterior

samples. The MCMC algorithm is independently run for each chromosome to generate an MCMC

4-60
 Analyzing Array-Based CGH Data Using Bayesian Hidden Markov Modeling

sample for the chromosome parameters. The starting values of the parameters are generated from
the priors. The generated copy number states represent draws from the marginal posterior of
interest, For each MCMC draw, the generated states are inspected and classified as focal ablations,
transition points, amplifications, outliers and whole chromosomal changes.

In this example, you will apply the Bayesian HMM algorithm to analyze the array CGH profiles of
some pancreatic cancer samples [2].

Loading the Data

The data in this example is the array CGH profiles of 24 pancreatic adenocarcinoma cell lines and 13
primary tumor specimens from Alguirre et al.,(2004). Labeled DNA fragments were hybridized to
Agilent® human cDNA microarrays containing 14,160 cDNA clones. About 9,420 clones have unique
map positions with a median interval between mapped elements of 100.1 kb. More details of the data
and experiment can be found in [2]. For convenience, the data has already been normalized and the
log2 based intensity ratios are provided by the MAT file pancrea_oligocgh.mat.

You will apply the Bayesian HMM algorithm to analyze chromosome 12 of sample 6 of the pancreatic
adenocarcinoma data, and compare the results with the segments found by the circular binary
segmentation (CBS) algorithm of Olshen et al.,(2004).

Load the MAT file containing the log2 intensity ratios and mapped genomic positions of the 37
samples.

load pancrea_oligocgh
pancrea_data

pancrea_data =

struct with fields:

Sample: {37x1 cell}

Chromosome: [13446x1 int8]
GenomicPosition: [13446x1 int32]
Log2Ratio: [13446x37 double]
Log2RatioMed: [13446x37 double]
Log2RatioSeg: [13446x37 double]
CloneIDs: [13446x1 int32]

Specify the chromosome number and sample to analyze.

sampleIndex = 6;
chromID = 12;
sample = pancrea_data.Sample{sampleIndex}

sample =

'PA.C.Dan.G'

Load and plot the log2 ratio data of chromosome 12 from sample PA.C.Dan.G.

idx = pancrea_data.Chromosome == chromID;

X = double(pancrea_data.GenomicPosition(idx));

4-61
4 Microarray Analysis

Y = pancrea_data.Log2Ratio(idx, sampleIndex);

% Remove NaN data points

idx = ~isnan(Y);
X = X(idx);
Y = Y(idx);

% Plot the data

figure;
plot(X, Y, '.', 'color', [0.6 0.6 1])

ylims = [-1.5, 3.5];

ylim(gca, ylims)
title(sprintf('%s - Chromosome %d', sample, chromID))
xlabel('Genomic Position');
ylabel('Log2(Ratio)')

Number of clones on chromosome 12 to be analyzed

N = numel(Y)

N =

437

4-62
 Analyzing Array-Based CGH Data Using Bayesian Hidden Markov Modeling

Performing Circular Binary Segmentation

You can start the analysis by performing chromosomal segmentation using the CBS algorithm [3],
which is implemented in the cghcbs function. The process will take several seconds. You can view
the plot of the segment means over the original data by specifying the SHOWPLOT parameter. Note:
You can type doc cghcbs for more details on this function.
PS = cghcbs(pancrea_data, 'SampleInd', sampleIndex, ...
'Chromosome', chromID, 'ShowPlot', chromID);
ylim(gca, ylims)

Analyzing: PA.C.Dan.G. Current chromosome 12

As shown in the figure, the CBS procedure declared the set of high intensity ratios as two separate
segments. The CBS procedure also found a region with copy number losses.

Initializing Parameters

The Bayesian HMM approach uses a Metropolis-within-Gibbs algorithm to generate posterior samples
of the parameters [1]. The model parameters are grouped into four blocks. The algorithm iteratively
generates each of the four blocks conditional on the remaining blocks and the data.

To analyze the data with the Bayesian HMM algorithm, you need to initialize the parameters. More
details on prior parameters can be found in references [1] and [4].

Initialize the state of the random number generator to ensure that the figures generated by these
command match the figures in the HTML version of this example.

4-63
4 Microarray Analysis

rng('default');

Define the number of states

NS = 4;

Define the number of MCMC iterations

NMC = 100;

Determine the hyperparameters of the prior distributions for the four states.

mus_hyper = [-1, 0, 0.58, 1];

taus_hyper = [1, 1, 1, 2];

Set the parameter epsilon which determines the constrains of the means.

eps = 0.1;

Set the bounds of the prior means of each state.

mu_low_bounds = [-Inf, -eps, eps, 0.58];

mu_up_bounds = [-eps, eps, 0.58, Inf];

Guha et al., (2006) assumes the inverse of the prior error variances (sigma^2) as gamma
distributions with lower bounds of 0.41 for states 1, 2 and 3. Set the scale parameters for the gamma
distributions for each state.

sg_alpha = [1 1 1 1];
sg_beta = [1, 1, 1, 1];
sg_bounds = [0.41 0.41 0.41 1];

Define a variable states to store the copy number state sequences of the clones for each MCMC
iteration.

states = zeros(N, NMC);

Define a variable st_counts to hold the state transition counts for each copy number state.

st_counts = zeros(NS, NS);

Determining the Prior Distributions

The MCMC iteration starts at

iloop = 1;

Determine sigmas for the four states by sampling from gamma distribution with prior scale
parameter alpha and beta.

sigmas = zeros(NS, NMC);

for i = 1:NS
sigmas(i, iloop) = acghhmmsample('gamma', sg_alpha(i), sg_beta(i), sg_bounds(i));
end

Determine means for the four states by sampling from truncated normal distribution between the
lower and upper bounds of the means. Note: The fourth state lower bound will be determined by the
third state.

4-64
 Analyzing Array-Based CGH Data Using Bayesian Hidden Markov Modeling

mus = zeros(NS, NMC);

for i = 1:NS
if i == 4
mu_low_bounds(4) = mus(3,iloop) + 3*sigmas(3,iloop);
end
mus(i, iloop) = acghhmmsample('normal', mus_hyper(i), taus_hyper(i),...
mu_low_bounds(i), mu_up_bounds(i));
end

Assume independent Dirichlet priors for the rows of the stochastic 4x4 transition probability matrix
[1]. Generate the stochastic prior transition matrix A from the Dirichlet distributions.

a = ones(NS, NS);
A = acghhmmsample('dirichlet', a, NS);

The transition matrix has a unique stationary distribution. The stationary distribution PI is an
eigenvector of the transition matrix associated with the eigenvalue 1.

PI =@(x, n) (ones(1,n)/(eye(n) -x + ones(n)))';

Generate the prior stationary distribution PI.

Pi = PI(A, NS);

Generate the initial emission matrix B

B = zeros(NS, N);
for i = 1:NS
B(i,:) = normpdf(Y, mus(i,iloop), sigmas(i,iloop));
end

Decode initial hidden states of the clones using a stochastic forward-backward algorithm [4].

states(:, iloop) = acghhmmfb(Pi, A, B);

Generating Posterior Samples

For each MCMC iteration, the four blocks of parameters are generated as follows [1]: Update block
B1 using a Metropolis-Hastings step to generate the transition matrix, update block B2 the copy
number states using a stochastic forward propagate backward sampling algorithm, update block B3
by computing the mus, and update block B4 to generate sigmas.

for iloop = 2:NMC

% Compute the number of transitions from state i to state j
for i =1:NS
for j = 1:NS
st_counts(i, j) = sum((states(1:N-1, iloop-1) == i) .* (states(2:N, iloop-1) == j));
end
end

% Updating block B1
% Generate the transition matrix from the Dirichlet distributions
C = acghhmmsample('dirichlet', st_counts + 1, NS);

% Compute the state probabilities under stationary distribution of a

% given transition matrix C.
PiC = PI(C, NS);

4-65
4 Microarray Analysis

% Compute the accepting probability using a Metropolis-Hastings step

beta = min([1, exp(log(PiC(states(1, iloop-1))) - log(Pi(states(1, iloop-1))))]);
if rand < beta
A = C;
Pi = PiC;
end

% Updating block B2
% Generate copy number states using Forward propagate, backward sampling [4].
states(:, iloop) = acghhmmfb(Pi, A, B);

% Updating blocks B3 and B4

for i = 1:NS
idx_s = states(:, iloop) == i;
num_states = sum(idx_s);

% If state i is not observed, then draw from its prior parameters

if num_states == 0
mus(i, iloop) = acghhmmsample('normal', mus_hyper(i),...
taus_hyper(i), mu_low_bounds(i), mu_up_bounds(i));
sigmas(i, iloop)= acghhmmsample('gamma', sg_alpha(i),...
sg_beta(i), sg_bounds(i));
else
Y_avg = mean(Y(idx_s));
theta_prec = 1/taus_hyper(i)^2 + num_states/sigmas(i,iloop-1)^2;
weight_means = (mus_hyper(i)/(taus_hyper(i)^2) +...
Y_avg * num_states/(sigmas(i, iloop-1)^2))/theta_prec;
% Compute mus - B3
mus(i, iloop) = acghhmmsample('normal', weight_means, ...
1/sqrt(theta_prec), mu_low_bounds(i), mu_up_bounds(i));
% Compute sigmas - B4
Y_v = sum((Y(idx_s) - mus(i, iloop)).^2);
sigmas(i, iloop) = acghhmmsample('gamma', sg_alpha(i)+num_states/2,...
sg_beta(i)+Y_v/2, sg_bounds(i));
end
% Update the emission matrix with new mus and sigmas.
B(i,:) = normpdf(Y, mus(i,iloop),sigmas(i,iloop));
end
end

Plot the posterior mean mu distributions of the four states.

figure;
for j = 1:NS
subplot(2,2,j)
ksdensity(mus(j,:));
title(sprintf('State %d', j))
end
sgtitle('Distribution of Mu of States');
hold off;

4-66
 Analyzing Array-Based CGH Data Using Bayesian Hidden Markov Modeling

Plot the posterior sigma distributions of the four states.

figure;
for j = 1:NS
subplot(2,2,j)
ksdensity(sigmas(j,:));
title(sprintf('State %d', j))
end
sgtitle('Distribution of Sigma of States');
hold off;

4-67
4 Microarray Analysis

Posterior Inference

Draw a state label for each clone from the MCMC sampling and compute the posterior probabilities
of each state.

clone_states = zeros(1, N);

state_prob = zeros(NS, N);
state_count = zeros(NS, N);

for i = 1:N % for each clone

state = states(i, :);
for j=1:NS
state_count(j, i) = sum(state == j);
end

selState = find(state_count(:,i) == max(state_count(:,i)));

if length(selState) > 1
if i ~= 1
clone_states(i) = clone_states(i-1);
else
clone_states(i) = min(selState);
end
else
clone_states(i) = selState;
end
state_prob(:, i) = state_count(:,i)/NMC;

4-68
 Analyzing Array-Based CGH Data Using Bayesian Hidden Markov Modeling

end
clone_states = clone_states';

Plot the state label for each clone on chromosome 12 of sample PA.C.Dan.G.
figure;
leg = zeros(1,4);
for i = 1:N
if clone_states(i) == 1
leg(1) = plot(i,Y(i),'v', 'MarkerFaceColor', [1 0.2 0.2],...
'MarkerEdgeColor', 'none');
elseif clone_states(i) == 2
leg(2) = plot(i,Y(i),'o', 'Color', [0.4 0.4 0.4]);
elseif clone_states(i) == 3
leg(3) = plot(i,Y(i),'^', 'MarkerFaceColor', [0.2 1 0.2],...
'MarkerEdgeColor', 'none');
elseif clone_states(i) == 4
leg(4) = plot(i, Y(i), '^', 'MarkerFaceColor', [0.2 0.2 1],...
'MarkerEdgeColor', 'none');
end
hold on;
end
ylim(gca, ylims)
legend(leg, 'State 1', 'State 2','State 3','State 4')
xlabel('Index')
ylabel('Log2(ratio)')
title('State Label')
hold off

4-69
4 Microarray Analysis

Classifying Array CGH Profiles

For each MCMC draw, the generated states can be classified as focal aberrations, transition points,
amplifications, outliers and whole chromosomal changes [1]. In this example, you will find the high-
level amplifications, transition points and outliers on chromosome 12 of sample PA.C.Dan.G.

A clone with state = 4 is considered a high-level amplification [1]. Find high-level amplifications.
high_lvl_amp_idx = find(clone_states == 4);

A transition point is associated with large-scale regions of gains and losses and is declared when the
width of the altered region exceeds 5 mega base pair [1]. Find transition points.
region_lim = 5e6;
focalabr_idx=[1;find(diff(clone_states)~=0);N];
istranspoint = false(length(focalabr_idx), 1);
for i = 1:length(focalabr_idx)-1
region_x = X(focalabr_idx(i+1)) - X(focalabr_idx(i));
istranspoint(i+1) = region_x > region_lim;
end
trans_idx = focalabr_idx(istranspoint);
% Remove adjacent trans_idx that have the same states.
hasadjacentstate = [diff(clone_states(trans_idx))==0; true];
trans_idx = trans_idx(~hasadjacentstate)
focalabr_idx = focalabr_idx(~istranspoint);
focalabr_idx = focalabr_idx(2:end-1);

trans_idx =

107
135
323

An outlier for gains is a focal aberration satisfying its z-score greater than 2, while an outlier for
losses has a z-score less than -2 [1].

Find outliers for losses

outlier_loss_idx = focalabr_idx(clone_states(focalabr_idx) == 1)
if ~isempty(outlier_loss_idx)
[F,Xi] = ksdensity(mus(1,:));
[dummy, idx] = max(F);
mu_1 = Xi(idx);

[F,Xi] = ksdensity(sigmas(1,:));
[dummy, idx] = max(F);
sigma_1 = Xi(idx);
outlier_loss_idx = outlier_loss_idx((Y(outlier_loss_idx) - mu_1)/sigma_1 < -2)
end

outlier_loss_idx =

0x1 empty double column vector

Find outliers for gains

4-70
 Analyzing Array-Based CGH Data Using Bayesian Hidden Markov Modeling

outlier_gain_idx = focalabr_idx(clone_states(focalabr_idx) == 3);

if ~isempty(outlier_gain_idx)
[F,Xi] = ksdensity(mus(3,:));
[dummy, idx] = max(F);
mu_1 = Xi(idx);

[F,Xi] = ksdensity(sigmas(3,:));
[dummy, idx] = max(F);
sigma_1 = Xi(idx);
outlier_gain_idx = outlier_gain_idx((Y( outlier_gain_idx) - mu_1)/sigma_1 > 2)
end

outlier_gain_idx =

0x1 empty double column vector

Add the classified labels to the intensity ratio plot of chromosome 12 of sample PA.C.Dan.G. Plot the
segment means from the CBS procedure for comparison.

figure;
hl1 = plot(X, Y, '.', 'color', [0.4 0.4 0.4]);
hold on;
if ~isempty(high_lvl_amp_idx)
hl2 = line(X(high_lvl_amp_idx), Y(high_lvl_amp_idx),...
'LineStyle', 'none',...
'Marker', '^',...
'MarkerFaceColor', [0.2 0.2 1],...
'MarkerEdgeColor', 'none');
end

if ~isempty(trans_idx)
for i = 1:numel(trans_idx)
hl3 = line(ones(1,2)*X(trans_idx(i)), [-3.5, 3.5],...
'LineStyle', '--',...
'Color', [1 0.6 0.2]);
end
end

if ~isempty(outlier_gain_idx)
line(X(outlier_gain_idx), Y(outlier_gain_idx),...
'LineStyle', 'none',...
'Marker', 'v',...
'MarkerFaceColor', [1 0 0],...
'MarkerEdgeColor', 'none');
end

if ~isempty(outlier_loss_idx)
hl4 = line(X(outlier_loss_idx), Y(outlier_loss_idx),...
'LineStyle', 'none',...
'Marker', 'v',...
'MarkerFaceColor', [1 0 0],...
'MarkerEdgeColor', 'none');
end

% Plot segment means from the CBS procedure.

for i = 1:numel(PS.SegmentData.Start)

4-71
4 Microarray Analysis

hl5 = line([PS.SegmentData.Start(i) PS.SegmentData.End(i)],...

[PS.SegmentData.Mean(i) PS.SegmentData.Mean(i)],...
'Color', [1 0 0],...
'LineWidth', 1.5);
end
ylim(gca, ylims)
ylabel('Log2(Ratio)')
title(sprintf('%s - Chromosome %d', sample, chromID))

% Adding chromosome 12 ideogram and legends to the plot.

chromosomeplot('hs_cytoBand.txt', chromID, 'addtoplot', gca)
legend([hl1, hl2, hl3,hl5], 'IntensityRatio', 'Amplification',...
'TransitionPoint', 'CBS SegmentMean')

The Bayesian HMM algorithm found 3 transition points indicated by the broken vertical lines in the
plot. The Bayesian HMM algorithm identified two high-level amplified regions marked by blue up-
triangles in the plot. The two high-level amplified regions correspond to the two minimal common
regions (MCRs)[2] on chromosome 12, associated with copy number gains as explained by Aguirre et
al.,(2004). The Bayesian HMM declared the first set of high intensity rations as a single region of
high-level amplification. In comparison, the CBS procedure failed to detect the second MCR and
segmented the first MCR into two regions. No outlier was detected in this example.

References

[1] Guha, S., Li, Y. and Neuberg, D., "Bayesian hidden Markov modeling of array CGH data", Journal
of the American Statistical Association, 103(482):485-497, 2008.

4-72
 Analyzing Array-Based CGH Data Using Bayesian Hidden Markov Modeling

[2] Aguirre, A.J., et al., "High-resolution characterization of the pancreatic adenocarcinoma genome",
PNAS, 101(24):9067-72, 2004.

[3] Olshen, A.B., et al., "Circular binary segmentation for the analysis of array-based DNA copy
number data", Biostatistics, 5(4):557-7, 2004.

[4] Shah, S.P., et al., "Integrating copy number polymorphisms into array CGH analysis using a robust
HMM", Bioinformatics, 22(14):e431-e439, 2006

4-73
4 Microarray Analysis

Visualizing Microarray Data

This example shows various ways to explore and visualize raw microarray data. The example uses
microarray data from a study of gene expression in mouse brains [1].

Exploring the Microarray Data Set

Brown, V.M et.al. [1] used microarrays to explore the gene expression patterns in the brain of a
mouse in which a pharmacological model of Parkinson's disease (PD) was induced using
methamphetamine. The raw data for this experiment is available from the Gene Expression Omnibus
website using the accession number GSE30 [1].

The file mouse_h3pd.gpr contains the data for one of the microarrays used in the study, specifically
from a sample collected from voxel H3 of the brain in a Parkinson's Disease (PD) model mouse. The
file uses the GenePix® GPR file format. The voxel sample was labeled with Cy3 (green) and the
control (RNA from a total, not voxelated, normal mouse brain) was labeled with Cy5.

GPR formatted files provide a large amount of information about the array including the mean,
median and standard deviation of the foreground and background intensities of each spot at the
635nm wavelength (the red, Cy5 channel) and the 532nm wavelength (the green, Cy3 channel).

The command gprread reads the data from the file into a structure.
pd = gprread('mouse_h3pd.gpr')

pd =

struct with fields:

Header: [1x1 struct]

Data: [9504x38 double]
Blocks: [9504x1 double]
Columns: [9504x1 double]
Rows: [9504x1 double]
Names: {9504x1 cell}
IDs: {9504x1 cell}
ColumnNames: {38x1 cell}
Indices: [132x72 double]
Shape: [1x1 struct]

You can access the fields of a structure using dot notation. For example, access the first ten column
names.
pd.ColumnNames(1:10)

ans =

10x1 cell array

{'X' }
{'Y' }
{'Dia.' }
{'F635 Median' }

4-74
 Visualizing Microarray Data

{'F635 Mean' }
{'F635 SD' }
{'B635 Median' }
{'B635 Mean' }
{'B635 SD' }
{'% > B635+1SD'}

You can also access the first ten gene names.

pd.Names(1:10)

ans =

10x1 cell array

{'AA467053'}
{'AA388323'}
{'AA387625'}
{'AA474342'}
{'Myo1b' }
{'AA473123'}
{'AA387579'}
{'AA387314'}
{'AA467571'}
{0x0 char }

Spatial Images of Microarray Data

The maimage command can take the microarray data structure and create a pseudocolor image of
the data arranged in the same order as the spots on the array, i.e., a spatial plot of the microarray.
The "F635 Median" field shows the median pixel values for the foreground of the red (Cy5) channel.

figure
maimage(pd,'F635 Median','title',{'Parkinson''s Model','Foreground Median Pixels','Red Channel'})

4-75
4 Microarray Analysis

The "F532 Median" field corresponds to the foreground of the green (Cy3) channel.

figure
maimage(pd,'F532 Median','title',{'Parkinson''s Model','Foreground Median Pixels','Green Channel'

4-76
 Visualizing Microarray Data

The "B635 Median" field shows the median values for the background of the red channel. Notice the
very high background levels down the right side of the array.

figure
maimage(pd,'B635 Median','title',{'Parkinson''s Model','Background Median Pixels','Red Channel'})

4-77
4 Microarray Analysis

The "B532 Median" shows the median values for the background of the green channel.

figure
maimage(pd,'B532 Median','title',{'Parkinson''s Model','Background Median Pixels','Green Channel'

4-78
 Visualizing Microarray Data

You can now consider the data obtained for the same brain voxel in an untreated control mouse. In
this case, the voxel sample was labeled with Cy3, and the control (RNA from a total, not voxelated
brain) was labeled with Cy5.
wt = gprread('mouse_h3wt.gpr')

wt =

struct with fields:

Header: [1x1 struct]

Data: [9504x38 double]
Blocks: [9504x1 double]
Columns: [9504x1 double]
Rows: [9504x1 double]
Names: {9504x1 cell}
IDs: {9504x1 cell}
ColumnNames: {38x1 cell}
Indices: [132x72 double]
Shape: [1x1 struct]

Use maimage to show pseudocolor images of the foreground and background corresponding to the
untreated mouse. The subplot command can be used to combine the plots.
figure
subplot(2,2,1);

4-79
4 Microarray Analysis

maimage(wt,'F635 Median','title',{'Foreground','(Red)'})
subplot(2,2,2);
maimage(wt,'F532 Median','title',{'Foreground','(Green)'})
subplot(2,2,3);
maimage(wt,'B635 Median','title',{'Background','(Red)'})
subplot(2,2,4);
maimage(wt,'B532 Median','title',{'Background','(Green)'})

annotation('textbox','String','Wild Type Median Pixel Values', ...

'Position', [0.3 0.05 0.9 0.01],'EdgeColor','none','FontSize',12);

If you look at the scale for the background images, you will notice that the background levels are
much higher than those for the PD mouse and there appears to be something non random affecting
the background of the Cy3 channel of this slide. Changing the colormap can sometimes provide more
insight into what is going on in pseudocolor plots. For more control over the color, try the
colormapeditor function. You can also right-click on the colorbar to bring up various options for
modifying the colormap of the plot including interactive colormap shifting.

colormap hot

4-80
 Visualizing Microarray Data

The maimage command is a simple way to quickly create pseudocolor images of microarray data.
However, sometimes it is convenient to create customizable plots using the imagesc command, as
shown below.

Use magetfield to extract data for the B532 median field and the Indices field to index into the
Data. You can bound the intensities of the background plot to give more contrast in the image.

b532Data = magetfield(wt,'B532 Median');

maskedData = b532Data;
maskedData(b532Data<500) = 500;
maskedData(b532Data>2000) = 2000;

figure
subplot(1,2,1);
imagesc(b532Data(wt.Indices))
axis image
colorbar
title('B532, WT')

subplot(1,2,2);
imagesc(maskedData(wt.Indices))
axis image
colorbar
title('Enhanced B532, WT')

4-81
4 Microarray Analysis

Statistics of the Microarrays

The maboxplot function can be used to look at the distribution of data in each of the blocks.

figure
subplot(2,1,1)
maboxplot(pd,'F532 Median','title','Parkinson''s Disease Model Mouse')
subplot(2,1,2)
maboxplot(pd,'B532 Median','title','Parkinson''s Disease Model Mouse')

figure
subplot(2,1,1)
maboxplot(wt,'F532 Median','title','Untreated Mouse')
subplot(2,1,2)
maboxplot(wt,'B532 Median','title','Untreated Mouse')

4-82
Visualizing Microarray Data

4-83
4 Microarray Analysis

From the box plots you can clearly see the spatial effects in the background intensities. Blocks
number 1,3,5 and 7 are on the left side of the arrays, and blocks number 2,4,6 and 8 are on the right
side.

There are two columns in the microarray data structure labeled "F635 Median - B635" and "F532
Median - B532". These columns are the differences between the median foreground and the median
background for the 635 nm channel and 532 nm channel respectively. These give a measure of the
actual expression levels. The spatial effect is less noticeable in these plots.

figure
subplot(2,1,1)
maboxplot(pd,'F635 Median - B635','title','Parkinson''s Disease Model Mouse ')
subplot(2,1,2)
maboxplot(pd,'F532 Median - B532','title','Parkinson''s Disease Model Mouse')

figure
subplot(2,1,1)
maboxplot(wt,'F635 Median - B635','title','Untreated Mouse')
subplot(2,1,2)
maboxplot(wt,'F532 Median - B532','title','Untreated Mouse')

4-84
Visualizing Microarray Data

4-85
4 Microarray Analysis

Scatter Plots of Microarray Data

Rather than work with the data in the larger structure, it is often easier to extract the data into
separate variables.

cy5Data = magetfield(pd,'F635 Median - B635');

cy3Data = magetfield(pd,'F532 Median - B532');

A simple way to compare the two channels is with a loglog plot. The function maloglog is used to
do this. Points that are above the diagonal in this plot correspond to genes that have higher
expression levels in the H3 voxel than in the brain as a whole.

figure
maloglog(cy5Data,cy3Data)
title('Loglog Scatter Plot of PD Model');
xlabel('F635 Median - B635 (Control)');
ylabel('F532 Median - B532 (Voxel H3)');

Warning: Zero values are ignored.

Warning: Negative values are ignored.

4-86
 Visualizing Microarray Data

Notice how the loglog function gives some warnings about negative and zero elements. This is
because some of the values in the 'F635 Median - B635' and 'F532 Median - B532' columns are zero
or less than zero. Spots where this happened might be bad spots or spots that failed to hybridize.
Similarly, spots with positive, but very small, differences between foreground and background are
also considered bad spots. These warnings can be disabled using the warning command.

warnState = warning; % Save the current warning state

warning('off','bioinfo:maloglog:ZeroValues');
warning('off','bioinfo:maloglog:NegativeValues');

figure
maloglog(cy5Data,cy3Data)
title('Loglog Scatter Plot of PD Model');
xlabel('F635 Median - B635 (Control)');
ylabel('F532 Median - B532 (Voxel H3)');

warning(warnState); % Reset the warning state

4-87
4 Microarray Analysis

An alternative to simply ignoring or disabling the warnings is to remove the bad spots from the data
set. This can be done by finding points where either the red or green channel have values less than or
equal to a threshold value, for example 10.

threshold = 10;
badPoints = (cy5Data <= threshold) | (cy3Data <= threshold);

You can then remove these points and redraw the loglog plot.

cy5Data(badPoints) = []; cy3Data(badPoints) = [];

figure
maloglog(cy5Data,cy3Data)
title('Refined Loglog Scatter Plot of PD Model');
xlabel('F635 Median - B635 (Control)');
ylabel('F532 Median - B532 (Voxel H3)');

4-88
 Visualizing Microarray Data

The distribution plot can be annotated by labeling the various points with the corresponding genes.

figure
maloglog(cy5Data,cy3Data,'labels',pd.Names(~badPoints),'factorlines',2)
title('Loglog Scatter Plot of PD Model');
xlabel('F635 Median - B635 (Control)');
ylabel('F532 Median - B532 (Voxel H3)');

4-89
4 Microarray Analysis

Try using the mouse to click on some of the outlier points. You will see the gene name associated with
the point. Most of the outliers are below the y = x line. In fact most of the points are below this line.
Ideally the points should be evenly distributed on either side of this line. In order for this to happen,
the points need to be normalized. You can use the manorm function to perform global mean
normalization.

normcy5 = manorm(cy5Data);
normcy3 = manorm(cy3Data);

If you plot the normalized data you will see that the points are more evenly distributed about the y =
x line.

figure
maloglog(normcy5,normcy3,'labels',pd.Names(~badPoints),'factorlines',2)
title('Normalized Loglog Scatter Plot of PD Model');
xlabel('F635 Median - B635 (Control)');
ylabel('F532 Median - B532 (Voxel H3)');

4-90
 Visualizing Microarray Data

You will recall that the background of the chips was not uniform. You can use print-tip (block)
normalization to normalize each block separately. The function manorm will perform block
normalization automatically if block information is available in the microarray data structure.

bn_cy5Data = manorm(pd,'F635 Median - B635');

bn_cy3Data = manorm(pd,'F532 Median - B532');

Instead of removing negative or points below the threshold, you can set them to NaN. This does not
change the size or shape of the data, but NaN points will not be displayed on plots.

bn_cy5Data(bn_cy5Data <= 0) = NaN;

bn_cy3Data(bn_cy3Data <= 0) = NaN;

figure
maloglog(bn_cy5Data,bn_cy3Data,'labels',pd.Names,'factorlines',2)
title('Refined, Normalized Loglog Scatter Plot of PD Model');
xlabel('F635 Median - B635 (Control)');
ylabel('F532 Median - B532 (Voxel H3)');

4-91
4 Microarray Analysis

The function mairplot is used to create an Intensity vs. Ratio plot for the normalized data. If the
name-value pair 'PlotOnly' is set to false, you can explore the data interactively, such as select points
to see the names of the associated genes, normalize the data, highlight gene names in the up-
regulated or down-regulated lists, or change the values of the factor lines.

mairplot(normcy5,normcy3,'labels',pd.Names(~badPoints),'PlotOnly',true,...
'title','Intensity vs. Ratio of PD Model');

4-92
 Visualizing Microarray Data

You can use the Normalize option to mairplot to perform Lowess normalization on the data.

mairplot(normcy5,normcy3,'labels',pd.Names(~badPoints),'PlotOnly',true,...
'Normalize',true,'title', 'Intensity vs. Ratio of PD Model (Normalized)');

4-93
4 Microarray Analysis

GenePix is a registered trademark of Axon Instruments, Inc.

References

[1] Brown, V.M., et al., "Multiplex three dimensional brain gene expression mapping in a mouse model
of Parkinson's disease", Genome Research, 12(6):868-84, 2002.

4-94
 Gene Expression Profile Analysis

Gene Expression Profile Analysis

This example shows a number of ways to look for patterns in gene expression profiles.

Exploring the Data Set

This example uses data from the microarray study of gene expression in yeast published by DeRisi, et
al. 1997 [1]. The authors used DNA microarrays to study temporal gene expression of almost all
genes in Saccharomyces cerevisiae during the metabolic shift from fermentation to respiration.
Expression levels were measured at seven time points during the diauxic shift. The full data set can
be downloaded from the Gene Expression Omnibus website, https://www.ncbi.nlm.nih.gov/geo/query/
acc.cgi?acc=GSE28.

The MAT-file yeastdata.mat contains the expression values (log2 of ratio of CH2DN_MEAN and
CH1DN_MEAN) from the seven time steps in the experiment, the names of the genes, and an array of
the times at which the expression levels were measured.

load yeastdata.mat

To get an idea of the size of the data you can use numel(genes) to show how many genes are
included in the data set.

numel(genes)

ans =

6400

You can access the genes names associated with the experiment by indexing the variable genes, a
cell array representing the gene names. For example, the 15th element in genes is YAL054C. This
indicates that the 15th row of the variable yeastvalues contains expression levels for YAL054C.

genes{15}

ans =

'YAL054C'

A simple plot can be used to show the expression profile for this ORF.

plot(times, yeastvalues(15,:))
xlabel('Time (Hours)');
ylabel('Log2 Relative Expression Level');

4-95
4 Microarray Analysis

You can also plot the actual expression ratios, rather than the log2-transformed values.

plot(times, 2.^yeastvalues(15,:))
xlabel('Time (Hours)');
ylabel('Relative Expression Level');

4-96
 Gene Expression Profile Analysis

The gene associated with this ORF, ACS1, appears to be strongly up-regulated during the diauxic
shift. You can compare the expression of this gene to the expression of other genes by plotting
multiple lines on the same figure.

hold on
plot(times, 2.^yeastvalues(16:26,:)')
xlabel('Time (Hours)');
ylabel('Relative Expression Level');
title('Profile Expression Levels');

4-97
4 Microarray Analysis

Filtering the Genes

Typically, a gene expression dataset includes information corresponding to genes that do not show
any interesting changes during the experiment. To make it easier to find the interesting genes, you
can reduce the size of the data set to some subset that contains only the most significant genes.

If you look through the gene list, you will see several spots marked as 'EMPTY'. These are empty
spots on the array, and while they might have data associated with them, for the purposes of this
example, you can consider these points to be noise. These points can be found using the strcmp
function and removed from the data set with indexing commands.
emptySpots = strcmp('EMPTY',genes);
yeastvalues(emptySpots,:) = [];
genes(emptySpots) = [];
numel(genes)

ans =

6314

There are also see several places in the dataset where the expression level is marked as NaN. This
indicates that no data was collected for this spot at the particular time step. One approach to dealing
with these missing values would be to impute them using the mean or median of data for the
particular gene over time. This example uses a less rigorous approach of simply throwing away the
data for any genes where one or more expression level was not measured. The function isnan is used

4-98
 Gene Expression Profile Analysis

to identify the genes with missing data and indexing commands are used to remove the genes with
missing data.
nanIndices = any(isnan(yeastvalues),2);
yeastvalues(nanIndices,:) = [];
genes(nanIndices) = [];
numel(genes)

ans =

6276

If you were to plot the expression profiles of all the remaining profiles, you would see that most
profiles are flat and not significantly different from the others. This flat data is obviously of use as it
indicates that the genes associated with these profiles are not significantly affected by the diauxic
shift; however, in this example, you are interested in the genes with large changes in expression
accompanying the diauxic shift. You can use filtering functions in the Bioinformatics Toolbox™ to
remove genes with various types of profiles that do not provide useful information about genes
affected by the metabolic change.

You can use the genevarfilter function to filter out genes with small variance over time. The
function returns a logical array (i.e., a mask) of the same size as the variable genes with ones
corresponding to rows of yeastvalues with variance greater than the 10th percentile and zeros
corresponding to those below the threshold. You can use the mask to index into the values and
remove the filtered genes.
mask = genevarfilter(yeastvalues);

yeastvalues = yeastvalues(mask,:);
genes = genes(mask);
numel(genes)

ans =

5648

The function genelowvalfilter removes genes that have very low absolute expression values.
Note that these filter functions can also automatically calculate the filtered data and names, so it is
not necessary to index the original data using the mask.
[mask,yeastvalues,genes] = genelowvalfilter(yeastvalues,genes,'absval',log2(3));
numel(genes)

ans =

822

Finally, you can use the function geneentropyfilter to remove genes whose profiles have low
entropy, for example entropy levels in the 15th percentile of the data.
[mask,yeastvalues,genes] = geneentropyfilter(yeastvalues,genes,'prctile',15);
numel(genes)

4-99
4 Microarray Analysis

ans =

614

Cluster Analysis

Now that you have a manageable list of genes, you can look for relationships between the profiles
using some different clustering techniques from the Statistics and Machine Learning Toolbox™. For
hierarchical clustering, the function pdist calculates the pairwise distances between profiles and
linkage creates the hierarchical cluster tree.

corrDist = pdist(yeastvalues,'corr');
clusterTree = linkage(corrDist,'average');

The cluster function calculates the clusters based on either a cutoff distance or a maximum number
of clusters. In this case, the maxclust option is used to identify 16 distinct clusters.

clusters = cluster(clusterTree,'maxclust',16);

The profiles of the genes in these clusters can be plotted together using a simple loop and the
subplot command.

figure
for c = 1:16
subplot(4,4,c);
plot(times,yeastvalues((clusters == c),:)');
axis tight
end
sgtitle('Hierarchical Clustering of Profiles');

4-100
 Gene Expression Profile Analysis

The Statistics and Machine Learning Toolbox also has a K-means clustering function. Again, sixteen
clusters are found, but because the algorithm is different these will not necessarily be the same
clusters as those found by hierarchical clustering.

Initialize the state of the random number generator to ensure that the figures generated by these
command match the figures in the HTML version of this example.

rng('default');

[cidx, ctrs] = kmeans(yeastvalues,16,'dist','corr','rep',5,'disp','final');

figure
for c = 1:16
subplot(4,4,c);
plot(times,yeastvalues((cidx == c),:)');
axis tight
end
sgtitle('K-Means Clustering of Profiles');

Replicate 1, 21 iterations, total sum of distances = 23.4699.

Replicate 2, 22 iterations, total sum of distances = 23.5615.
Replicate 3, 10 iterations, total sum of distances = 24.823.
Replicate 4, 28 iterations, total sum of distances = 23.4501.
Replicate 5, 19 iterations, total sum of distances = 23.5109.
Best total sum of distances = 23.4501

4-101
4 Microarray Analysis

Instead of plotting all the profiles, you can plot just the centroids.

figure
for c = 1:16
subplot(4,4,c);
plot(times,ctrs(c,:)');
axis tight
axis off
end
sgtitle('K-Means Clustering of Profiles');

4-102
 Gene Expression Profile Analysis

You can use the clustergram function to create a heat map of the expression levels and a
dendrogram from the output of the hierarchical clustering.

cgObj = clustergram(yeastvalues(:,2:end),'RowLabels',genes,'ColumnLabels',times(2:end));

4-103
4 Microarray Analysis

Principal Component Analysis

Principal-component analysis(PCA) is a useful technique that can be used to reduce the

dimensionality of large data sets, such as those from microarrays. PCA can also be used to find
signals in noisy data. The function mapcaplot calculates the principal components of a data set and
create scatter plots of the results. You can interactively select data points from one of the plots, and
these points are automatically highlighted in the other plot. This lets you visualize multiple
dimensions simultaneously.

h = mapcaplot(yeastvalues,genes);

4-104
 Gene Expression Profile Analysis

Notice that the scatter plot of the scores of the first two principal components shows that there are
two distinct regions. This is not unexpected as the filtering process removed many of the genes with
low variance or low information. These genes would have appeared in the middle of the scatter plot.

If you want to look at the values of the principal components, the pca function in the Statistics and
Machine Learning Toolbox is used to calculate the principal components of a data set.

4-105
4 Microarray Analysis

[pc, zscores, pcvars] = pca(yeastvalues);

The first output, pc, is a matrix of the principal components of the yeastvalues data. The first
column of the matrix is the first principal component, the second column is the second principal
component, and so on. The second output, zscores, consists of the principal component scores, i.e.,
a representation of yeastvalues in the principal component space. The third output, pcvars, contains
the principal component variances, which give a measure of how much of the variance of the data is
accounted for by each of the principal components.

It is clear that the first principal component accounts for a majority of the variance in the model. You
can compute the exact percentage of the variance accounted for by each component as shown below.

pcvars./sum(pcvars) * 100

ans =

79.8316
9.5858
4.0781
2.6486
2.1723
0.9747
0.7089

This means that almost 90% of the variance is accounted for by the first two principal components.
You can use the cumsum command to see the cumulative sum of the variances.

cumsum(pcvars./sum(pcvars) * 100)

ans =

79.8316
89.4174
93.4955
96.1441
98.3164
99.2911
100.0000

If you want to have more control over the plotting of the principal components, you can use the
scatter function.

figure
scatter(zscores(:,1),zscores(:,2));
xlabel('First Principal Component');
ylabel('Second Principal Component');
title('Principal Component Scatter Plot');

4-106
 Gene Expression Profile Analysis

An alternative way to create a scatter plot is with the function gscatter from the Statistics and
Machine Learning Toolbox. gscatter creates a grouped scatter plot where points from each group
have a different color or marker. You can use clusterdata, or any other clustering function, to
group the points.

figure
pcclusters = clusterdata(zscores(:,1:2),'maxclust',8,'linkage','av');
gscatter(zscores(:,1),zscores(:,2),pcclusters,hsv(8))
xlabel('First Principal Component');
ylabel('Second Principal Component');
title('Principal Component Scatter Plot with Colored Clusters');

4-107
4 Microarray Analysis

Self-Organizing Maps

If you have the Deep Learning Toolbox™, you can use a self-organizing map (SOM) to cluster the
data.

% Check to see if the Deep Learning Toolbox is installed

if ~exist(which('selforgmap'),'file')
disp('The Self-Organizing Maps section of this example requires the Deep Learning Toolbox.')
return
end

The selforgmap function creates a new SOM network object. This example will generate a SOM
using the first two principal components.

P = zscores(:,1:2)';
net = selforgmap([4 4]);

Train the network using the default parameters.

net = train(net,P);

4-108
 Gene Expression Profile Analysis

Use plotsom to display the network over a scatter plot of the data. Note that the SOM algorithm
uses random starting points so the results will vary from run to run.

figure
plot(P(1,:),P(2,:),'.g','markersize',20)
hold on
plotsom(net.iw{1,1},net.layers{1}.distances)
hold off

4-109
4 Microarray Analysis

You can assign clusters using the SOM by finding the nearest node to each point in the data set.

distances = dist(P',net.IW{1}');
[d,cndx] = min(distances,[],2); % cndx contains the cluster index

figure
gscatter(P(1,:),P(2,:),cndx,hsv(numel(unique(cndx)))); legend off;
hold on
plotsom(net.iw{1,1},net.layers{1}.distances);
hold off

4-110
 Gene Expression Profile Analysis

Close all figures.

close('all');
delete(cgObj);
delete(h);

References

[1] DeRisi, J.L., Iyer, V.R. and Brown, P.O., "Exploring the metabolic and genetic control of gene
expression on a genomic scale", Science, 278(5338):680-6, 1997.

4-111
4 Microarray Analysis

Working with Affymetrix Data

This example shows how to use the functions in the Bioinformatics Toolbox™ for working with
Affymetrix® GeneChip® data.

About Affymetrix Data Files

The function affyread can read four types of Affymetrix data files. These are DAT files, which
contain raw image data, CEL files which contain information about the intensity values of the
individual probes, CHP files which contain information about probe sets, and EXP files, which contain
information about experimental conditions and protocols. affyread can also read CDF and GIN
library files. The CDF file contains information about which probes belong to which probe set and the
GIN file contains information about the probe sets such as the gene name with which the probe set is
associated. To learn more about the actual files, you can download sample data files from the
Affymetrix Support Site. Most of the data sets are stored in DTT archives. To extract the DAT, CEL
and CHP files you will need to install the Data Transfer Tool.

Downloading the E. coli Antisense Data Set

For this example, you will need some sample data files (DAT, CEL, CHP) from the E. coli Antisense
Genome Array. Download these from Demo_Data_E-coli-antisense.zip Extract the data files from the
DTT archive using the Data Transfer Tool. Set the variable exampleDataDir to the name of the path
and directory to which you extracted the sample data files.

exampleDataDir = 'C:\Examples\affydemo\data';

Downloading E. coli Antisense Library Files

In addition to the data files, you will also need Ecoli_ASv2.CDF and Ecoli_ASv2.GIN, the library files
for the E. coli Antisense Genome Array. You may already have these files if you have any Affymetrix
GeneChip software installed on your machine. If not, get the library files by downloading and
unzipping the E. coli Antisense Genome Array zip file

Note that you will have to register in order to access the library files.

You only have to unzip the files, you do not have to run the Setup.exe file in the archive.

Set the variable libDir to the name of the path and directory to which you extracted the library
files.

libDir = 'C:\Examples\affydemo\libfiles';

Image Files (DAT Files)

The raw image data from the chip scanner is saved in the DAT file. If you use affyread to read a
DAT file you will see that it creates a MATLAB® structure.

datStruct = affyread(fullfile(exampleDataDir,'Ecoli-antisense-121502.dat'))

datStruct =

struct with fields:

Name: 'Ecoli-antisense-121502.dat'

4-112
 Working with Affymetrix Data

DataPath: 'I:\qe\test_data\Bioinformatics_Toolbox\v000\demoData\affydemo\data'
LibPath: 'I:\qe\test_data\Bioinformatics_Toolbox\v000\demoData\affydemo\data'
FullPathName: 'I:\qe\test_data\Bioinformatics_Toolbox\v000\demoData\affydemo\data\Ecoli-an
ChipType: 'Ecoli_ASv2'
NumPixelsPerRow: 4733
NumRows: 4733
MinData: 0
MaxData: 46108
PixelSize: 3
CellMargin: 2
ScanSpeed: 17
ScanDate: '13-Aug-0001 11:31:58'
ScannerID: ''
UpperLeftX: 231
UpperLeftY: 235
UpperRightX: 4492
UpperRightY: 253
LowerLeftX: 220
LowerLeftY: 4501
LowerRightX: 4482
LowerRightY: 4519
ServerName: ''
Image: [4733x4733 uint16]

You can access fields of the structure using the dot notation.

datStruct.NumRows

ans =

4733

Displaying an Image File

You can use the imagesc command to display the image.

datFigure = figure;
imagesc(datStruct.Image);

4-113
4 Microarray Analysis

You can change the colormap from the default jet to another using the colormap command.

colormap pink

4-114
 Working with Affymetrix Data

You can zoom in on a particular area by using the Zoom In tool with the mouse, or by using the axis
command. Notice that this stretches the y-axis.

axis([1900 2800 160 650])

4-115
4 Microarray Analysis

You can use the axis image command to set the correct aspect ratio.

axis image
axis([1900 2800 160 650])

4-116
 Working with Affymetrix Data

Probe Results Files (CEL Files)

The information about each probe on the chip is extracted from the image data by the Affymetrix
image analysis software. The information is stored in the CEL file. affyread reads a CEL file into a
structure. Notice that many of the fields are the same as those in the DAT structure.
celStruct = affyread(fullfile(exampleDataDir,'Ecoli-antisense-121502.CEL'))

celStruct =

struct with fields:

Name: 'Ecoli-antisense-121502.CEL'
DataPath: 'I:\qe\test_data\Bioinformatics_Toolbox\v000\demoData\affydemo\data'
LibPath: 'I:\qe\test_data\Bioinformatics_Toolbox\v000\demoData\affydemo\data'
FullPathName: 'I:\qe\test_data\Bioinformatics_Toolbox\v000\demoData\affydemo\data\Ecoli-a
ChipType: 'Ecoli_ASv2'
Date: '01-Feb-2013 11:55:24'
FileVersion: 3
Algorithm: 'Percentile'
AlgParams: 'Percentile:75;CellMargin:2;OutlierHigh:1.500;OutlierLow:1.004'
NumAlgParams: 4
CellMargin: 2
Rows: 544
Cols: 544
NumMasked: 0
NumOutliers: 115

4-117
4 Microarray Analysis

NumProbes: 295936
UpperLeftX: 231
UpperLeftY: 235
UpperRightX: 4492
UpperRightY: 253
LowerLeftX: 220
LowerLeftY: 4501
LowerRightX: 4482
LowerRightY: 4519
ProbeColumnNames: {8x1 cell}
Probes: [295936x8 single]

The CEL file contains information about where each probe is on the chip and also the intensity values
for the probe. You can use the maimage function to display the chip.

celFigure = figure;
maimage(celStruct)

Again, you can zoom in on a specific region.

axis([200 340 0 70])

4-118
 Working with Affymetrix Data

If you compare the image created from the CEL file and the image created from the DAT file, you will
notice that the CEL image is lower resolution. This is because there is only one pixel per probe in this
image, whereas the DAT file image has many pixels per probe.

The structures created by affyread can be very large. It is a good idea to clear them from memory
once they are no longer needed.

clear datStruct
close(datFigure); close(celFigure);

The Probes field of the CEL structure contains information about the individual probes. There are
eight values per probe. These are stored in the ProbeColumnNames field of the structure.

celStruct.ProbeColumnNames

ans =

8x1 cell array

{'PosX' }
{'PosY' }
{'Intensity'}
{'StdDev' }
{'Pixels' }
{'Outlier' }
{'Masked' }

4-119
4 Microarray Analysis

{'ProbeType'}

So if you look at one row of the Probes field of the CEL structure you will see eight values
corresponding to the X position, Y position, intensity, and so forth.

celStruct.Probes(1:10,:)

ans =

10x8 single matrix

1.0e+04 *

Columns 1 through 7

0 0 0.0082 0.0030 0.0036 0 0

0.0001 0 1.4202 0.3160 0.0036 0 0
0.0002 0 0.0080 0.0014 0.0030 0 0
0.0003 0 1.4760 0.2265 0.0036 0 0
0.0004 0 0.0050 0.0014 0.0036 0 0
0.0005 0 0.0073 0.0015 0.0036 0 0
0.0006 0 1.3595 0.2367 0.0036 0 0
0.0007 0 0.0087 0.0018 0.0036 0 0
0.0008 0 1.3284 0.2926 0.0036 0 0
0.0009 0 0.0104 0.0018 0.0030 0 0

Column 8

0.0001
0.0001
0.0001
0.0001
0.0001
0.0001
0.0001
0.0001
0.0001
0.0001

Results Files (CHP Files)

The CHP file contains the results of the experiment. These include the average signal measures for
each probe set as determined by the Affymetrix software and information about which probe sets are
called as present, absent or marginal and the p-values for these calls.

chpStruct = affyread(fullfile(exampleDataDir,'Ecoli-antisense-121502.CHP'),libDir)

chpStruct =

struct with fields:

Name: 'Ecoli-antisense-121502.CHP'
DataPath: 'I:\qe\test_data\Bioinformatics_Toolbox\v000\demoData\affydemo\data'
LibPath: 'I:\qe\test_data\Bioinformatics_Toolbox\v000\demoData\affydemo\libfiles'

4-120
 Working with Affymetrix Data

FullPathName: 'I:\qe\test_data\Bioinformatics_Toolbox\v000\demoData\affydemo\data\Ecoli-an
ChipType: 'Ecoli_ASv2'
AssayType: 'Expression'
Date: '01-Feb-2013 11:55:24'
CellFile: 'c:\documents and settings\bkolou\desktop\demo_data_e-coli-antisense\Ecoli-a
Algorithm: 'ExpressionStat'
AlgVersion: '5.0'
NumAlgParams: 13
AlgParams: 'SFGene=All SF=5.578290 NF=1.000000 TGT=500 Perturbation=1.1 Gamma2L=0.006 G
NumChipSummary: 3
ChipSummary: 'RawQ=1.62 Noise=Avg:1.33,Stdev:0.20,Max:1.7,Min:1.0 Background=Avg:42.81,St
BackgroundZones: [1x1 struct]
Rows: 544
Cols: 544
NumProbeSets: 7312
NumQCProbeSets: 0
ProbeSets: [7312x1 struct]

The ProbeSets field contains information about the probe sets. This includes some library
information, such as the ID and the type of probe set, and also results information such as the
calculated signal value and the Present/Absent/Marginal call information. The call is given in the
Detection field of the ProbeSets structure. The 'argG_b3172_at' probe set is called as being
'Present'.

chpStruct.ProbeSets(5213)

ans =

struct with fields:

Name: 'argG_b3172_at'
ProbeSetType: 'Expression'
CompDataExists: 0
NumPairs: 15
NumPairsUsed: 15
Signal: 127.6070
Detection: 'Present'
DetectionPValue: 0.0134
CommonPairs: []
SignalLogRatio: []
SignalLogRatioLow: []
SignalLogRatioHigh: []
Change: []
ChangePValue: []

However, the 'IG_2069_3319273_3319712_rev_at' probe set is called 'Absent'.

chpStruct.ProbeSets(5216)

ans =

struct with fields:

Name: 'IG_2069_3319273_3319712_rev_at'

4-121
4 Microarray Analysis

ProbeSetType: 'Expression'
CompDataExists: 0
NumPairs: 15
NumPairsUsed: 15
Signal: 35.0037
Detection: 'Absent'
DetectionPValue: 0.2661
CommonPairs: []
SignalLogRatio: []
SignalLogRatioLow: []
SignalLogRatioHigh: []
Change: []
ChangePValue: []

And the 'yhbX_b3173_at' probe set is called 'Marginal'.

chpStruct.ProbeSets(5215)

ans =

struct with fields:

Name: 'yhbX_b3173_at'
ProbeSetType: 'Expression'
CompDataExists: 0
NumPairs: 15
NumPairsUsed: 15
Signal: 147.7237
Detection: 'Marginal'
DetectionPValue: 0.0559
CommonPairs: []
SignalLogRatio: []
SignalLogRatioLow: []
SignalLogRatioHigh: []
Change: []
ChangePValue: []

You can calculate how many probe sets are called as being 'Present',

numPresent = sum(strcmp('Present',{chpStruct.ProbeSets.Detection}))

numPresent =

4605

'Absent',

numAbsent = sum(strcmp('Absent',{chpStruct.ProbeSets.Detection}))

numAbsent =

2524

4-122
 Working with Affymetrix Data

and 'Marginal'.

numMarginal = sum(strcmp('Marginal',{chpStruct.ProbeSets.Detection}))

numMarginal =

183

maboxplot will display a box plot of the log2 signal values for all probe sets.

maboxplot(chpStruct,'Signal','title',chpStruct.Name)

Library Files (CDF Files)

The CHP file gives summary information about probe sets but if you want more detailed information
about how the individual probes in a probe set behave you need to connect the probe information in
the CEL file to the corresponding probe sets. This information is stored in the CDF library file
associated with a chip type. The CDF files are typically stored in a central library directory.

cdfStruct = affyread('Ecoli_ASv2.cdf',libDir)

cdfStruct =

struct with fields:

4-123
4 Microarray Analysis

Name: 'Ecoli_ASv2.cdf'
ChipType: 'Ecoli_ASv2'
LibPath: 'I:\qe\test_data\Bioinformatics_Toolbox\v000\demoData\affydemo\libfiles'
FullPathName: 'I:\qe\test_data\Bioinformatics_Toolbox\v000\demoData\affydemo\libfiles\
Date: '04-Feb-2013 11:14:01'
Rows: 544
Cols: 544
NumProbeSets: 7312
NumQCProbeSets: 13
ProbeSetColumnNames: {6x1 cell}
ProbeSets: [7325x1 struct]

Most of the information in the file is about the probe sets. In this example there are 7312 regular
probe sets and 13 QC probe sets. The ProbeSets field of the structure is a 7325x1 array of
structures.

cdfStruct.ProbeSets

ans =

7325x1 struct array with fields:

Name
ProbeSetType
CompDataExists
NumPairs
NumQCProbes
QCType
GroupNames
ProbePairs

A probe set record contains information about the name, type and number of probe pairs in the probe
set.

probeSetIndex = 5213;
cdfStruct.ProbeSets(probeSetIndex)

ans =

struct with fields:

Name: 'argG_b3172_at'
ProbeSetType: 'Expression'
CompDataExists: 0
NumPairs: 15
NumQCProbes: 0
QCType: 0
GroupNames: {'argG_b3172_at'}
ProbePairs: [15x6 int32]

The information about where the probes for a probe set are on the chip is stored in the ProbePairs
field. This is a matrix with one row for each probe pair and six columns. The information in the
columns corresponds to the ProbeSetColumnNames of the CDF structure.

4-124
 Working with Affymetrix Data

cdfStruct.ProbeSetColumnNames
cdfStruct.ProbeSets(probeSetIndex).ProbePairs

ans =

6x1 cell array

{'GroupNumber'}
{'Direction' }
{'PMPosX' }
{'PMPosY' }
{'MMPosX' }
{'MMPosY' }

ans =

15x6 int32 matrix

1 2 430 177 430 178

1 2 431 177 431 178
1 2 432 177 432 178
1 2 433 177 433 178
1 2 434 177 434 178
1 2 435 177 435 178
1 2 436 177 436 178
1 2 437 177 437 178
1 2 438 177 438 178
1 2 439 177 439 178
1 2 440 177 440 178
1 2 441 177 441 178
1 2 442 177 442 178
1 2 443 177 443 178
1 2 444 177 444 178

The first column shows the probe group number. The second column shows the probe direction. The
group number is always 1 for expression arrays. Direction 1 corresponds to 'sense' and 2 corresponds
to 'anti-sense'. The remaining columns give the X and Y coordinates of the PM and MM probes on the
chip. You can use these coordinates to find the index of a probe in the celStruct.
PMX = cdfStruct.ProbeSets(probeSetIndex).ProbePairs(1,3);
PMY = cdfStruct.ProbeSets(probeSetIndex).ProbePairs(1,4);
theProbe = find((celStruct.Probes(:,1) == PMX) & ...
(celStruct.Probes(:,2) == PMY))

theProbe =

96719

You can then extract all the information about this probe from the CEL structure.
celStruct.Probes(theProbe,:)

ans =

4-125
4 Microarray Analysis

1x8 single row vector

Columns 1 through 7

430.0000 177.0000 169.0000 35.4000 25.0000 0 0

Column 8

1.0000

If you want to do this lookup for all probes, you can use the function probelibraryinfo. This
creates a matrix with one row per probe and three columns. The first column is the index of the probe
set to which the probe belongs. The second column contains the probe pair index and the third
column indicates if the probe is a perfect match (1) or mismatch (-1) probe. Notice that index of the
probe pair index is 1 based.
probeinfo = probelibraryinfo(celStruct,cdfStruct);

probeinfo(theProbe,:)

ans =

5213 1 1

The function probesetvalues does the reverse of this lookup and creates a matrix of information
from the CEL and CDF structures containing all the information about a given probe set. This matrix
has 20 columns corresponding to ProbeSetNumber, ProbePairNumber, UseProbePair,
Background, PMPosX, PMPosY, PMIntensity, PMStdDev, PMPixels, PMOutlier, PMMasked,
MMPosX, MMPosY, MMIntensity, MMStdDev, MMPixels, MMOutlier, MMMasked, Group, and
Direction.
probeName = cdfStruct.ProbeSets(probeSetIndex).Name;
psvals = probesetvalues(celStruct,cdfStruct,probeName);
sprintf( ['%4d %2d %d %d PM: %3d %3d %5.1f %5.1f %2d %d %d',...
' MM: %3d %3d %5.1f %5.1f %2d %d %d %d %d\n'],psvals')

ans =

'5212 0 0 4.543512e+01 PM: 430 177 169.0 35.4 25 0 0 MM: 430 178 163.5 24.1 30 0 0 1 2
5212 1 0 4.545356e+01 PM: 431 177 127.3 21.8 30 0 0 MM: 431 178 100.3 14.6 36 0 0 1 2
5212 2 0 4.547230e+01 PM: 432 177 127.0 23.7 30 0 0 MM: 432 178 175.0 28.6 36 0 0 1 2
5212 3 0 4.549129e+01 PM: 433 177 133.3 25.9 36 0 0 MM: 433 178 94.0 22.7 30 0 0 1 2
5212 4 0 4.551051e+01 PM: 434 177 212.3 43.3 36 0 0 MM: 434 178 171.8 36.5 30 0 0 1 2
5212 5 0 4.552995e+01 PM: 435 177 149.5 27.5 36 0 0 MM: 435 178 154.0 30.3 30 0 0 1 2
5212 6 0 4.554958e+01 PM: 436 177 50.3 11.2 30 0 0 MM: 436 178 46.0 9.8 25 0 0 1 2
5212 7 0 4.556938e+01 PM: 437 177 152.5 37.7 36 0 0 MM: 437 178 107.0 21.0 36 0 0 1 2
5212 8 0 4.558934e+01 PM: 438 177 164.5 31.2 36 0 0 MM: 438 178 97.3 21.9 36 0 0 1 2
5212 9 0 4.560939e+01 PM: 439 177 126.0 23.4 36 0 0 MM: 439 178 121.3 25.3 36 0 0 1 2
5212 10 0 4.562955e+01 PM: 440 177 54.0 11.2 36 0 0 MM: 440 178 54.0 12.9 36 0 0 1 2
5212 11 0 4.564975e+01 PM: 441 177 83.3 17.4 36 0 0 MM: 441 178 62.3 12.5 36 0 0 1 2
5212 12 0 4.566998e+01 PM: 442 177 95.5 17.1 30 0 0 MM: 442 178 84.0 18.6 30 0 0 1 2
5212 13 0 4.569022e+01 PM: 443 177 110.0 19.6 36 0 0 MM: 443 178 92.5 22.0 36 0 0 1 2
5212 14 0 4.571042e+01 PM: 444 177 251.0 46.0 36 0 0 MM: 444 178 111.8 20.7 36 0 0 1 2

4-126
 Working with Affymetrix Data

'

You can extract the intensity values from the matrix and look at some of the statistics of the data.

pmIntensity = psvals(:,7);
mmIntensity = psvals(:,14);
boxplot([pmIntensity,mmIntensity],'labels',{'Perfect Match','Mismatch'})
title(sprintf('Boxplot of raw intensity values for probe set %s',...
probeName),'interpreter','none')
% Use interpreter none to prevent the TeX interpreter treating the _ as
% subscript.

Plotting the Probe Set Values

Now that you have the intensity values for the probes, you can plot the values for the perfect match
and mismatch probes.

figure
plot(pmIntensity,'b'); hold on
plot(mmIntensity,'r'); hold off
title(sprintf('Probe intensity values for probe set %s',...
probeName),'interpreter','none')

4-127
4 Microarray Analysis

Alternatively, you can use the function probesetplot to create this plot directly from the CEL and
CDF structures. The showstats option adds the mean, and lines for +/- one standard deviation for
both the perfect match and the mismatch probes to the plot.

probesetplot(celStruct,cdfStruct,probeName,'showstats',true);

4-128
 Working with Affymetrix Data

Gene Names and Probe Set IDs

The Affymetrix probe set IDs are not particularly descriptive. The mapping between the probe set IDs
and the gene IDs is stored in the GIN library file. This is a text file so you can open it in an editor and
browse through the file, or you can use affyread to read the information into a structure.

ginStruct = affyread('Ecoli_ASv2.GIN',libDir)

ginStruct =

struct with fields:

Name: 'Ecoli_ASv2'
Version: 2
ProbeSetName: {7312x1 cell}
ID: {7312x1 cell}
Description: {7312x1 cell}
SourceNames: {2x1 cell}
SourceURL: {2x1 cell}
SourceID: [7312x1 double]

You can search through the structure for a particular probe set. Alternatively, you can use the
function probesetlookup to find information about the gene for a probe set.

info = probesetlookup(cdfStruct,probeName)

4-129
4 Microarray Analysis

info =

struct with fields:

Identifier: '3315278'
ProbeSetName: 'argG_b3172_at'
CDFIndex: 5213
GINIndex: 3074
Description: '/start=3316278 /end=3317621 /direction=+ /description=argininosuccinate synthe
Source: 'NCBI EColi Genome'
SourceURL: 'http://www.ncbi.nlm.nih.gov/cgi-bin/Entrez/altvik?gi=115&db=g&from=3315278'

Getting Sequence Information About a Probe Set

The function probesetlink will link out to the NetAffx™ Web site to show the actual sequences
used for the probes. Note that you will need to be a registered user of NetAffx to access this
information.

probesetlink(cdfStruct,probeName);

Affymetrix, GeneChip, and NetAffx are registered trademarks of Affymetrix, Inc.

4-130
 Preprocessing Affymetrix Microarray Data at the Probe Level

Preprocessing Affymetrix Microarray Data at the Probe Level

This example shows how to use MATLAB® and Bioinformatics Toolbox™ for preprocessing
Affymetrix® oligonucleotide microarray probe-level data with two preprocessing techniques, Robust
Multi-array Average (RMA) and GC Robust Multi-array Average (GCRMA).

Introduction

With Affymetrix oligonucleotide microarray platforms, gene expression is measured using probe sets
consisting of 11 to 20 perfect match (PM) probes (25 nucleotides in length) complementary to target
mRNA sequences. Each probe set also has the same number of mismatch (MM) probes, in which the
13th nucleotide has been changed to its complement. The PM probes are designed for gene specific
hybridization. The control MM probe measurements are thought to comprise most of the background
non-specific binding, such as cross-hybridization. A PM probe and its corresponding MM probe are
referred to as a probe pair.

The measured probe intensities and locations from a hybridized microarray are stored in a CEL file.
For each Affymetrix microarray platform, the information relating probe pairs to probe set IDs, and to
locations on the array is stored in a CDF library file. The probe sequence information is stored in a
sequence file (FASTA or tab-separated format).

In general, preprocessing Affymetrix probe-level expression data consists of three steps: background
adjustment, normalization, and summarization at the probe set level as a measure of the expression
level of corresponding mRNA. Many methods exist for the statistical procedures of these three steps.
Two popular techniques, RMA (Irizarry et al., 2003) and GCRMA (Wu et al., 2004), are used in this
example.

Note: This example shows the RMA and GCRMA preprocessing procedures to compute expression
values from input CEL files in step-by-step detail, using several functions. You can also complete the
same RMA or GCRMA techniques in one function call by using the Bioinformatics Toolbox affyrma
or affygcrma functions, respectively.

A publicly available dataset containing Affymetrix microarray measurements of 42 tumor tissues of

the embryonal central nervous system (CNS, Pomeroy et al., 2002) is used for this example. You will
import and access the probe level data of multiple arrays, and then perform expression level
measurements with RMA and GCRMA preprocessing methods.

Importing Data

The CNS experiment was conducted using the Affymetrix HuGeneFL GeneChip® array, and the data
were stored in CEL files. Information related to each probe is contained in the Affymetrix Hu6800
CDF library file.

If you dont already have the Hu6800 CDF library file, download the HuGeneFL Genome Array library
zip file. Extract the Hu6800.CDF file into a directory, such as C:\Examples\affypreprocessdemo
\libfiles. Note: You will have to register in order to access the library files, but you do not have to
run the setup.exe file in the archive.

The CNS dataset (CEL files) is available here. To complete this example, download the CEL files of
the CNS dataset into a directory, such as C:\Examples\affypreprocessdemo\data. Unzip the
CEL file archives. Note: This dataset contains more CEL files than are needed for this example.

CNS_DataA_Sample_CEL.txt, a file provided with Bioinformatics Toolbox, contains a list of the 42 CEL
filenames used for this example, and the samples (10 medulloblastomas, 10 rhabdoid, 10 malignant

4-131
4 Microarray Analysis

glioma, 8 supratentorial PNETS, and 4 normal human cerebella) to which they belong. Load this data
into two MATLAB variables.

fid = fopen('CNS_DataA_Sample_CEL.txt','r');
ftext = textscan(fid,'%q%q');
fclose(fid);
samples = ftext{1};
cels = ftext{2};

Set the variables celPath and libPath to the paths of the CEL files and library directories.

celPath = 'C:\Examples\affypreprocessdemo\data';
libPath = 'C:\Examples\affypreprocessdemo\libfiles';

Rename the cel files so that each file name starts with the MG number that follows the underscore "_"
in the original file name. For instance, GSM1688666_MG1999060202AA.CEL is renamed to
MG1999060202AA.CEL. You do not need to run this code if the file names are already in the required
format.

A = dir(fullfile(celPath,'*.cel'));
fileNames = string({A.name});
for iFile = 1:numel(A)
newName = fullfile(celPath,extractAfter(fileNames(iFile),"_"));
movefile(fullfile(celPath,fileNames(iFile)),newName);
end

The function celintensityread can read multiple CEL files and access a CDF library file. It returns
a MATLAB structure containing the probe information and probe intensities. The matrices of PM and
MM intensities from multiple CEL files are stored in the PMIntensities and MMIntensities
fields. In each probe intensity matrix, the column indices correspond to the order in which the CEL
files were read, and each row corresponds to a probe. Create a MATLAB structure of PM and MM
probe intensities by loading data from the CEL files from the directory where the CEL files are
stored, and pass in the path to where you stored the CDF library file. (Note: celintensityread will
report the progress to the MATLAB command window. You can turn the progress report off by setting
the input parameter VERBOSE to false.)

probeData = celintensityread(cels, 'Hu6800.CDF',...

'celpath', celPath, 'cdfpath', libPath, 'pmonly', false)

Reading CDF file: Hu6800.CDF

Reading file 1 of 42: MG2000040501AA
Reading file 2 of 42: MG2000040502AA
Reading file 3 of 42: MG2000040504AA
Reading file 4 of 42: MG2000040505AA
Reading file 5 of 42: MG2000040508AA
Reading file 6 of 42: MG2000040509AA
Reading file 7 of 42: MG2000040510AA
Reading file 8 of 42: MG2000040511AA
Reading file 9 of 42: MG2000040512AA
Reading file 10 of 42: MG2000040513AA
Reading file 11 of 42: MG2000051201AA
Reading file 12 of 42: MG2000051202AA
Reading file 13 of 42: MG2000051204AA
Reading file 14 of 42: MG2000051205AA
Reading file 15 of 42: MG2000051209AA
Reading file 16 of 42: MG2000071102AA
Reading file 17 of 42: MG2000051207AA

4-132
 Preprocessing Affymetrix Microarray Data at the Probe Level

Reading file 18 of 42: MG2000051208AA

Reading file 19 of 42: MG2000051211AA
Reading file 20 of 42: MG2000051213AA
Reading file 21 of 42: MG2000061902AA
Reading file 22 of 42: MG2000061903AA
Reading file 23 of 42: MG2000061904AA
Reading file 24 of 42: MG2000061905AA
Reading file 25 of 42: MG2000061906AA
Reading file 26 of 42: MG2000070709AA
Reading file 27 of 42: MG2000070710AA
Reading file 28 of 42: MG2000070711AA
Reading file 29 of 42: MG2000070712AA
Reading file 30 of 42: MG2000070713AA
Reading file 31 of 42: MG1999112206AA
Reading file 32 of 42: MG2000033109AA
Reading file 33 of 42: MG2000033106AA
Reading file 34 of 42: MG2000033107AA
Reading file 35 of 42: MG1999112202AA
Reading file 36 of 42: MG1999112204AA
Reading file 37 of 42: MG2000011801AA
Reading file 38 of 42: MG2000031503AA
Reading file 39 of 42: MG2000032015AA
Reading file 40 of 42: MG2000030308AA
Reading file 41 of 42: MG2000011803AA
Reading file 42 of 42: MG2000011807AA

probeData =

struct with fields:

CDFName: 'Hu6800.CDF'
CELNames: {1×42 cell}
NumChips: 42
NumProbeSets: 7129
NumProbes: 140983
ProbeSetIDs: {7129×1 cell}
ProbeIndices: [140983×1 uint8]
GroupNumbers: [140983×1 uint8]
PMIntensities: [140983×42 single]
MMIntensities: [140983×42 single]

Determine the number of CEL files loaded.

nSamples = probeData.NumChips

nSamples =

42

Determine the number of probe sets on a HuGeneFL array.

nProbeSets = probeData.NumProbeSets

nProbeSets =

4-133
4 Microarray Analysis

7129

Determine the number of probes on a HuGeneFL array.

nProbes = probeData.NumProbes

nProbes =

140983

To perform GCRMA preprocessing, the probe sequence information of the HuGeneFL array is also
required. The Affymetrix support site provides probe sequence information for most of the available
arrays, either as FASTA formatted or tab-delimited files. This example assumes you have the
HuGeneFL_probe_tab file in the library files directory. Use the function affyprobeseqread to parse
the sequence file and return the probe sequences in an nProbes x 25 matrix of integers that
represents the PM probe sequence bases, with rows corresponding to the probes on the chip and
columns corresponding to the base positions of the 25-mer.

S = affyprobeseqread('HuGeneFL_probe_tab', 'Hu6800.CDF',...
'seqpath', libPath, 'cdfpath', libPath, 'seqonly', true)

S =

struct with fields:

SequenceMatrix: [140983×25 uint8]

Preprocessing Probe-Level Expression Data

The RMA procedure uses only PM probe intensities for background adjustment (Irizarry et al., 2003),
while GCRMA adjusts background using probe sequence information and MM control probe
intensities to estimate non-specific binding (Wu et al., 2004). Both RMA and GCRMA are preceded by
quantile normalization (Bolstad et al., 2003) and median polish summarization (Irizarry et al., 2003)
of PM intensities.

Using the RMA Procedure

The RMA background adjustment method corrects PM probe intensities chip by chip. The PM probe
intensities are modeled as the sum of a normal noise component and an exponential signal
component. Use rmabackadj to background adjust the PM intensities in the CNS data. You can
inspect the intensity distribution histogram and the estimated background adjustment of a specific
chip by setting the input parameter SHOWPLOT to the column index of the chip.

pms_bg = rmabackadj(probeData.PMIntensities, 'showplot', 1);

4-134
 Preprocessing Affymetrix Microarray Data at the Probe Level

Several nonlinear normalization methods have been successfully applied to Affymetrix microarray
data. The RMA procedure normalizes the probe-level data with a quantile normalization method. Use
quantilenorm to normalize the background adjusted PM intensities in the CNS data. Note: If you
are interested in a rank-invariant set normalization method, use the affyinvarsetnorm function
instead.

pms_bgnorm = quantilenorm(pms_bg);

A median polish procedure is applied to the PM intensities in summarization. To calculate the

expression values, use rmasummary to summarize probe intensities of each probe set across multiple
chips. The expression values are the probe set intensity summaries on a log-2 scale.

cns_rma_exp = rmasummary(probeData.ProbeIndices, pms_bgnorm);

Using the GCRMA Procedure

The GCRMA procedure adjusts for optical noise and non-specific binding (NSB) taking into account
the effect of the stronger bonding of G/C pairs (Naef et al., 2003, Wu et al., 2004). GCRMA uses probe
sequence information to estimate probe affinities for computing non-specific binding. The probe
affinity is modeled as a sum of the position-dependent base effects. Usually, the probe affinities are
estimated from the MM intensities of an NSB experiment. If NSB data is not available, the probe
affinities can still be estimated from sequence information and MM probe intensities normalized by
the probe set median intensity (Naef et al., 2003).

For the CNS dataset, use the data from the microarray hybridized with the normal cerebella sample
(Brain_Ncer_1) to compute the probe affinities for the HuGeneFL array. Use affyprobeaffinities

4-135
4 Microarray Analysis

to estimate the probe affinities of an Affymetrix microarray. Use the SHOWPLOT input parameter to
inspect a plot showing the effects of base A, C, G, and T at the 25 positions.

figure
idx = find(strcmpi('Brain_Ncer_1', samples));
[pmAlpha, mmAlpha] = affyprobeaffinities(S.SequenceMatrix,...
probeData.MMIntensities(:, idx), 'showplot', true);

Note: There are 496 probes on a HuGeneFL array that do not have sequence information; the
affinities for these probes were NaN.

With the probe affinities available, the amount of NSB can be estimated by fitting a LOWESS curve
through MM probe intensities vs. MM probe affinities. The function gcrmabackadj performs optical
and NSB corrections. The input parameter SHOWPLOT shows a plot of the optical noise adjusted MM
intensities against its affinities, and the smooth fit of a specified chip. You can compute the
background intensities with one of two estimation methods, Maximum Likelihood Estimate (MLE) and
Empirical-Bayes (EB), which computes the posterior mean of specific binding given prior observed
intensities. Here you will background adjust four arrays using both estimation methods. (Note:
gcrmabackadj will report the progress to the MATLAB command window. You can turn the progress
report off by setting the input parameter VERBOSE to false.)

Background adjust the first four chips using GCRMA-MLE method, and inspect the plot of intensity
vs. affinity for data from the third array.

pms_MLE_bg = gcrmabackadj(probeData.PMIntensities(:,1:4),...
probeData.MMIntensities(:, 1:4),...

4-136
 Preprocessing Affymetrix Microarray Data at the Probe Level

pmAlpha, mmAlpha, 'showplot', 3);

% Adjust YLIM for better view

ylim([4 16]);

Adjusting background for chip # 1 of 4 using MLE method.

Adjusting background for chip # 2 of 4 using MLE method.
Adjusting background for chip # 3 of 4 using MLE method.
Adjusting background for chip # 4 of 4 using MLE method.

Background adjust the first four chips using the GCRMA-EB method. Processing with this method is
more computationally intensive and will take longer.

pms_EB_bg = gcrmabackadj(probeData.PMIntensities(:,1:4),...
probeData.MMIntensities(:, 1:4),...
pmAlpha, mmAlpha, 'method', 'EB');

Adjusting background for chip # 1 of 4 using EB method.

Adjusting background for chip # 2 of 4 using EB method.
Adjusting background for chip # 3 of 4 using EB method.
Adjusting background for chip # 4 of 4 using EB method.

You can continue the preprocessing with the quatilenorm and rmasummary functions, or use the
gcrma function to do everything. The gcrma function performs background adjustment and returns
expression measures of background adjusted PM probe intensities using the same normalization and
summarization methods as RMA. You can also pass in the sequence matrix instead of affinities. The
function will automatically compute the affinities in this case. (Note: gcrma will report the progress

4-137
4 Microarray Analysis

to the MATLAB command window. You can turn the progress report off by setting the input parameter
VERBOSE to false.)

cns_mle_exp = gcrma(probeData.PMIntensities, probeData.MMIntensities,...

probeData.ProbeIndices, pmAlpha, mmAlpha);

Adjusting background for chip # 1 of 42 using MLE method.

Adjusting background for chip # 2 of 42 using MLE method.
Adjusting background for chip # 3 of 42 using MLE method.
Adjusting background for chip # 4 of 42 using MLE method.
Adjusting background for chip # 5 of 42 using MLE method.
Adjusting background for chip # 6 of 42 using MLE method.
Adjusting background for chip # 7 of 42 using MLE method.
Adjusting background for chip # 8 of 42 using MLE method.
Adjusting background for chip # 9 of 42 using MLE method.
Adjusting background for chip # 10 of 42 using MLE method.
Adjusting background for chip # 11 of 42 using MLE method.
Adjusting background for chip # 12 of 42 using MLE method.
Adjusting background for chip # 13 of 42 using MLE method.
Adjusting background for chip # 14 of 42 using MLE method.
Adjusting background for chip # 15 of 42 using MLE method.
Adjusting background for chip # 16 of 42 using MLE method.
Adjusting background for chip # 17 of 42 using MLE method.
Adjusting background for chip # 18 of 42 using MLE method.
Adjusting background for chip # 19 of 42 using MLE method.
Adjusting background for chip # 20 of 42 using MLE method.
Adjusting background for chip # 21 of 42 using MLE method.
Adjusting background for chip # 22 of 42 using MLE method.
Adjusting background for chip # 23 of 42 using MLE method.
Adjusting background for chip # 24 of 42 using MLE method.
Adjusting background for chip # 25 of 42 using MLE method.
Adjusting background for chip # 26 of 42 using MLE method.
Adjusting background for chip # 27 of 42 using MLE method.
Adjusting background for chip # 28 of 42 using MLE method.
Adjusting background for chip # 29 of 42 using MLE method.
Adjusting background for chip # 30 of 42 using MLE method.
Adjusting background for chip # 31 of 42 using MLE method.
Adjusting background for chip # 32 of 42 using MLE method.
Adjusting background for chip # 33 of 42 using MLE method.
Adjusting background for chip # 34 of 42 using MLE method.
Adjusting background for chip # 35 of 42 using MLE method.
Adjusting background for chip # 36 of 42 using MLE method.
Adjusting background for chip # 37 of 42 using MLE method.
Adjusting background for chip # 38 of 42 using MLE method.
Adjusting background for chip # 39 of 42 using MLE method.
Adjusting background for chip # 40 of 42 using MLE method.
Adjusting background for chip # 41 of 42 using MLE method.
Adjusting background for chip # 42 of 42 using MLE method.
Normalizing.
Calculating expression.

Inspecting the Background Adjustment Results

Use boxplots to inspect the PM intensity distributions of the first four chips with three background
adjustment procedures.

figure
subplot(4,1,1)

4-138
 Preprocessing Affymetrix Microarray Data at the Probe Level

maboxplot(log2(probeData.PMIntensities(:, 1:4)), samples(1:4),...

'title','Raw Intensities', 'orientation', 'horizontal')
subplot(4,1,2)
maboxplot(log2(pms_bg(:,1:4)), samples(1:4),...
'title','After RMA Background Adjustment','orient','horizontal')
subplot(4,1,3)
maboxplot(log2(pms_MLE_bg), samples(1:4),...
'title','After GCRMA-MLE Background Adjustment','orient','horizontal')
subplot(4,1,4)
maboxplot(log2(pms_EB_bg), samples(1:4),...
'title','After GCRMA-EB Background Adjustment','orient','horizontal')

Use boxplots to inspect the background corrected and normalized PM intensity distributions of the
first four chips with three background adjustment procedures.
pms_MLE_bgnorm = quantilenorm(pms_MLE_bg);
pms_EB_bgnorm = quantilenorm(pms_EB_bg);

figure
subplot(3,1,1)
maboxplot(log2(pms_bgnorm(:, 1:4)), samples(1:4),...
'title','Normalized RMA Background Adjusted Intensity',...
'orientation', 'horizontal')
subplot(3,1,2)
maboxplot(log2(pms_MLE_bgnorm), samples(1:4),...
'title','Normalized GCRMA-MLE Background Adjusted Intensity',...
'orientation', 'horizontal')
subplot(3,1,3)

4-139
4 Microarray Analysis

maboxplot(log2(pms_EB_bgnorm), samples(1:4),...
'title','Normalized GCRMA-EB Background Adjusted Intensity',...
'orientation', 'horizontal')

Final Remarks

You can perform importing of data from CEL files and all three preprocessing steps of the RMA and
GCRMA techniques shown in this example by using the affyrma and affygcrma functions
respectively.

Affymetrix and GeneChip are registered trademarks of Affymetrix, Inc.

References

[1] Pomeroy, S.L., et al., "Prediction of central nervous system embryonal tumour outcome based on
gene expression", Nature, 415(6870):436-42, 2002.

[2] Irizarry, R.A., et al., "Exploration, normalization, and summaries of high density oligonucleotide
array probe level data", Biostatistics, 4(2):249-64, 2003.

[3] Wu, Z., et al., "A model based background adjustment for oligonucleotide expression arrays",
Journal of the American Statistical Association, 99(468):909-17, 2004.

[4] Bolstad, B.M., et al., "A comparison of normalization methods for high density oligonucleotide
array data based on variance and bias", Bioinformatics, 19(2):185-93, 2003.

4-140
 Preprocessing Affymetrix Microarray Data at the Probe Level

[5] Naef, F., and Magnasco, M.O. "Solving the riddle of the bright mismatches: labeling and effective
binding in oligonucleotide arrays", Physical Review, E, Statistical, Nonlinear and Soft Matter Physics,
68(1Pt1):011906, 2003.

4-141
4 Microarray Analysis

Analyzing Affymetrix SNP Arrays for DNA Copy Number

Variants

This example shows how to study DNA copy number variants by preprocessing and analyzing data
from the Affymetrix® GeneChip® Human Mapping 100k array.

Introduction

A copy number variant (CNV) is defined as a chromosomal segment that is 1kb or larger in length,
whose copy number varies in comparison to a reference genome. CNV is one of the hallmarks of
genetic instability common to most human cancers. When studying cancers, an important goal is to
quickly and precisely identify copy number amplifications and deletions, and to assess their
frequencies at the genome level. Recently, single nucleotide polymorphism (SNP) arrays have been
used to detect and quantify genome-wide copy number alterations with high resolution. SNP array
approaches also provide genotype information. For example, they can reveal loss of heterozygosity
(LOH), which can provide supporting evidence for the presence of a deletion.

The Affymetrix GeneChip Mapping Array Set is a popular platform for high-throughput SNP
genotyping and CNV detection. In this example, we use a publicly available data set from the
Affymetrix 100K SNP array that interrogates over 100,000 SNP sites. You will import and preprocess
the probe level data, estimate the raw signal ratios of the samples compared to references, and then
infer copy numbers at each SNP locus after segmentation.

Data

Zhao et al. studied genome-wide copy number alterations of human lung carcinoma cell lines and
primary tumors [1]. The samples were hybridized to Affymetrix 100K SNP arrays, each containing
115,593 mapped SNP loci. For this example, you will analyze data from 24 small cell lung carcinoma
(SCLC) samples, of which 19 were primary tumor samples and 5 were cell line samples.

For each sample, SNPs were genotyped with two different arrays, Early Access 50KXba and Early
Access 50KHind, in parallel. In brief, two aliquots of DNA samples were first digested with an XbaI or
HindIII restriction enzyme, respectively. The digested DNA was ligated to an adaptor before
subsequent polymerase chain reaction (PCR) amplification. Four PCR reactions were set up for each
XbaI or HindIII adaptor-ligated DNA sample. The PCR products from the four reactions were pooled,
concentrated, and fragmented to a size range of 250 to 2,000 bp. Fragmented PCR products were
then labeled, denatured, and hybridized to the arrays.

For this example, you will work with data from the EA 50KXba array. To analyze the data from EA
50KHind array just repeat the steps. The SNP array data are stored in CEL files with each CEL file
containing data from one array.

Note: High density SNP microarray data analysis requires extended amounts of memory from the
operating system; if you receive "Out of memory" errors when running this example, try increasing
the virtual memory (or swap space) of your operating system or try setting the 3GB switch (32-bit
Windows® XP only). For details, see “Resolve “Out of Memory” Errors”.

This example uses the 50KXba and 50KHind SNP array data sets (not included in the toolbox) from
the Meyerson Laboratory at the Dana-Farber Cancer Institute. You may use any other dataset to
perform similar analyses.

The CDF library files used for these two arrays are CentXbaAv2.cdf and CentHindAv2.cdf. You
can obtain these files from Affymetrix Web Site.

4-142
 Analyzing Affymetrix SNP Arrays for DNA Copy Number Variants

Set the variable Xba_celPath with the path to the location you stored the Xba array CEL files, and
the variable libPath with the path to the location of the CDF library file for the EA 50KXba SNP
array. (These files are not distributed with Bioinformatics Toolbox™).
Xba_celPath = 'C:\Examples\affysnpcnvdemo\Xba_array';
libPath = 'C:\Examples\affysnpcnvdemo\LibFiles';

SCLC_Sample_CEL.txt, a file provided with the Bioinformatics Toolbox™ software, contains a list of
the 24 CEL file names used for this example, and the samples (5 SCLC cell lines and 19 primary
tumors) to which they belong. Load this data into two MATLAB® variables.
fid = fopen('SCLC_Sample_CEL.txt','r');
ftext = textscan(fid, '%q%q');
fclose(fid);
samples = ftext{1};
cels = ftext{2};
nSample = numel(samples)

nSample =

24

Accessing SNP Array Probe-Level Data

The Affymetrix 50KXba SNP array has a density up to 50K SNP sites. Each SNP on the array is
represented by a collection of probe quartets. A probe quartet consists of a set of probe pairs for both
alleles (A and B) and for both forward and reverse strands (antisense and sense) for the SNP. Each
probe pair consists a perfect match (PM) probe and a mismatch (MM) probe. The Bioinformatics
Toolbox software provides functions to access the probe-level data.

The function affyread reads the CEL files and the CDF library files for Affymetrix SNP arrays.

Read the sixth CEL file of the EA 50KXba data into a MATLAB structure.
s_cel = affyread(fullfile(Xba_celPath, [cels{6} '.CEL']))

s_cel =

struct with fields:

Name: 'S0168T.CEL'
DataPath: 'C:\Examples\affysnpcnvdemo\Xba_array'
LibPath: 'C:\Examples\affysnpcnvdemo\Xba_array'
FullPathName: 'C:\Examples\affysnpcnvdemo\Xba_array\S0168T.CEL'
ChipType: 'CentXbaAv2'
Date: '01-Feb-2013 11:54:13'
FileVersion: 3
Algorithm: 'Percentile'
AlgParams: 'Percentile:75;CellMargin:2;OutlierHigh:1.500;OutlierLow:1.004;AlgVersion:6
NumAlgParams: 16
CellMargin: 2
Rows: 1600
Cols: 1600
NumMasked: 0
NumOutliers: 12478

4-143
4 Microarray Analysis

NumProbes: 2560000
UpperLeftX: 222
UpperLeftY: 236
UpperRightX: 8410
UpperRightY: 219
LowerLeftX: 252
LowerLeftY: 8426
LowerRightX: 8440
LowerRightY: 8410
ProbeColumnNames: {8×1 cell}
Probes: [2560000×8 single]

Read the CDF library file for the EA 50KXba array into a MATLAB structure.

s_cdf = affyread(fullfile(libPath, 'CentXbaAv2.cdf'))

s_cdf =

struct with fields:

Name: 'CentXbaAv2.cdf'
ChipType: 'CentXbaAv2'
LibPath: 'C:\Examples\affysnpcnvdemo\LibFiles'
FullPathName: 'C:\Examples\affysnpcnvdemo\LibFiles\CentXbaAv2.cdf'
Date: '01-Feb-2013 11:54:12'
Rows: 1600
Cols: 1600
NumProbeSets: 63434
NumQCProbeSets: 9
ProbeSetColumnNames: {6×1 cell}
ProbeSets: [63443×1 struct]

You can inspect the overall quality of the array by viewing the probe-level intensity data using the
function maimage.

maimage(s_cel)

4-144
 Analyzing Affymetrix SNP Arrays for DNA Copy Number Variants

The affysnpquartets function creates a table of probe quartets for a SNP. On Affymetrix 100K SNP
arrays, a probe quartet contains 20 probe pairs. For example, to get detailed information on probe set
number 6540, you can type the following commands:

ps_id = 6540;
ps_qt = affysnpquartets(s_cel, s_cdf, ps_id)

ps_qt =

struct with fields:

ProbeSet: '2685329'
AlleleA: 'A'
AlleleB: 'G'
Quartet: [1×6 struct]

You can also view the heat map of the intensities of the PM and MM probe pairs of a SNP probe
quartet using the probesetplot function. Click the Insert Colorbar button to show the color scale
of the heat map.

probesetplot(s_cel, s_cdf, ps_id, 'imageonly', true);

4-145
4 Microarray Analysis

In this view, the 20 probe pairs are ordered from left to right. The first two rows (10 probe pairs)
correspond to allele A, and the last two rows (10 probe pairs) corresponds to allele B. For each allele,
the left 5 probe pairs correspond to the sense strand (-), while the right 5 probe pairs correspond to
the antisense (+) strand.

Importing and Converting the Data Set

You will use the celintensityread function to read all 24 CEL files. The celintensityread
function returns a structure containing the matrices of PM and MM (optional) intensities for the
probes and their group numbers. In each probe intensity matrix, the column indices correspond to
the order in which the CEL files were read, and each row corresponds to a probe. For copy number
(CN) analysis, only PM intensities are needed.

Import the probe intensity data of all EA 50KXba arrays into a MATLAB structure.

XbaData = celintensityread(cels, 'CentXbaAv2.cdf',...

'celpath', Xba_celPath, 'cdfpath', libPath)

Reading CDF file: CentXbaAv2.cdf

Reading file 1 of 24: H524
Reading file 2 of 24: H526
Reading file 3 of 24: H1184
Reading file 4 of 24: H1607
Reading file 5 of 24: H1963
Reading file 6 of 24: S0168T
Reading file 7 of 24: S0169T
Reading file 8 of 24: S0170T

4-146
 Analyzing Affymetrix SNP Arrays for DNA Copy Number Variants

Reading file 9 of 24: S0171T

Reading file 10 of 24: S0172T
Reading file 11 of 24: S0173T
Reading file 12 of 24: S0177T
Reading file 13 of 24: S0185T
Reading file 14 of 24: S0187T
Reading file 15 of 24: S0188T
Reading file 16 of 24: S0189T
Reading file 17 of 24: S0190T
Reading file 18 of 24: S0191T
Reading file 19 of 24: S0192T
Reading file 20 of 24: S0193T
Reading file 21 of 24: S0194T
Reading file 22 of 24: S0196T
Reading file 23 of 24: S0198T
Reading file 24 of 24: S0199T

XbaData =

struct with fields:

CDFName: 'CentXbaAv2.cdf'
CELNames: {1×24 cell}
NumChips: 24
NumProbeSets: 63434
NumProbes: 1268480
ProbeSetIDs: {63434×1 cell}
ProbeIndices: [1268480×1 uint8]
GroupNumbers: [1268480×1 uint8]
PMIntensities: [1268480×24 single]

Affymetrix Early Access arrays are the same as the current commercial Mapping 100K arrays with
the exception of some the probes being masked out. The data obtained from Affymetrix EA 100K SNP
arrays can be converted to Mapping 100K arrays by filtering out the rejected SNP probe IDs on Early
Access array and converting the SNP IDs to Mapping 100K SNP IDs. The SNP IDs for EA 50KXba and
50KHind arrays and their corresponding SNP IDs on Mapping 50KXba and 50KHind arrays are
provided in two MAT files shipped with the Bioinformatics Toolbox software, Mapping50K_Xba_V_EA
and Mapping50K_Hind_V_EA, respectively.
load Mapping50K_Xba_V_EA

The helper function affysnpemconvert converts the data to Mapping 50KXba data.
XbaData = affysnpemconvert(XbaData, EA50K_Xba_SNPID, Mapping50K_Xba_SNPID)

XbaData =

struct with fields:

CDFName: 'CentXbaAv2.cdf'
CELNames: {1×24 cell}
NumChips: 24
NumProbeSets: 58960
NumProbes: 1179200
ProbeSetIDs: {58960×1 cell}
ProbeIndices: [1179200×1 uint8]

4-147
4 Microarray Analysis

GroupNumbers: [1179200×1 uint8]

PMIntensities: [1179200×24 single]

Probe Intensity Normalization

You can view the density plots of the log-transformed PM intensity distribution across the 24 samples
before preprocessing.
f=zeros(nSample, 100);
xi = zeros(nSample, 100);
for i = 1:nSample
[f(i,:),xi(i,:)] = ksdensity(log2(XbaData.PMIntensities(:,i)));
end

figure;
plot(xi', f')
xlabel('log2(PM)')
ylabel('Density')
title('Density Plot')
hold on

Quantile normalization is particularly effective in normalizing non-linearities in data introduced by

experimental biases. Perform quantile normalization using the quantilenorm function.
XbaData.PMIntensities = quantilenorm(XbaData.PMIntensities);

Plot the resulting quantile distribution using a dashed red curve.

4-148
 Analyzing Affymetrix SNP Arrays for DNA Copy Number Variants

[f,xi] = ksdensity(log2(XbaData.PMIntensities(:,1)));
plot(xi', f', '--r', 'Linewidth', 3)
hold off

Note: You can also use the RMA or GCRMA procedures for background correction. The RMA
procedure estimates the background by a mixture model where the background signals are assumed
to be normally distributed and the true signals are exponentially distributed, while the GCRMA
process consists of optical background correction and probe-sequence based background adjustment.
For more information on how to use the RMA and GCRMA procedures, see “Preprocessing Affymetrix
Microarray Data at the Probe Level” on page 4-131.

Probe-Level Summarization

By using the GroupNumbers field data from the structure XbaData, you can extract the intensities
for allele A and allele B for each probe. Use the function affysnpintensitysplit to split the
probe intensities matrix PMIntensities into two single-precision matrices, PMAIntensities and
PMBIntensities, for allele A and allele B probes respectively. The number of probes in each matrix
is the maximum number of probes for each allele.
XbaData = affysnpintensitysplit(XbaData)

XbaData =

struct with fields:

CDFName: 'CentXbaAv2.cdf'

4-149
4 Microarray Analysis

CELNames: {1×24 cell}

NumChips: 24
NumProbeSets: 58960
NumProbes: 589600
ProbeSetIDs: {58960×1 cell}
ProbeIndices: [589600×1 uint8]
PMAIntensities: [589600×24 single]
PMBIntensities: [589600×24 single]

For total copy number analysis, a simplification is to ignore the allele A and allele B sequences and
their strand information and, instead, combine the PM intensities for allele A and allele B of each
probe pair.
PM_Xba = XbaData.PMAIntensities + XbaData.PMBIntensities;

For a particular SNP, we now have K (K=5 for Affymetrix Mapping 100K arrays) added signals, each
signal being a measure of the same thing - the total CN. However, each of the K signals has slightly
different sequences, so their hybridization efficiency might differ. You can use RMA summarization
methods to sum up allele probe intensities for each SNP probe set.
PM_Xba = rmasummary(XbaData.ProbeIndices, PM_Xba);

Getting SNP Probe Information

Affymetrix provides CSV-formatted annotation files for their SNP arrays. You can download the
annotation files for Mapping 100K arrays from https://www.thermofisher.com/us/en/home/life-science/
microarray-analysis/microarray-data-analysis/genechip-array-annotation-files.html.

For this example, download and unzip the annotation file for the Mapping, 50KXba array
Mapping50K_Xba240.na29.annot.csv. The SNP probe information of the Mapping 50KXba array,
can be read from this annotation file. Set the variable annoPath with the path to the location where
you saved the annotation file.
annoPath = 'C:\Examples\affysnpcnvdemo\AnnotFiles';

The function affysnpannotread reads the annotation file and returns a structure containing SNP
chromosome information, chromosomal positions, sequences and PCR fragment length information
ordered by probe set IDs from the second input variable.
annoFile = fullfile(annoPath, 'Mapping50K_Xba240.na29.annot.csv');
annot_Xba = affysnpannotread(annoFile, XbaData.ProbeSetIDs)

annot_Xba =

struct with fields:

ProbeSetIDs: {58960×1 cell}

Chromosome: [58960×1 int8]
ChromPosition: [58960×1 double]
Cytoband: {58960×1 cell}
Sequence: {58960×1 cell}
AlleleA: {58960×1 cell}
AlleleB: {58960×1 cell}
Accession: {58960×1 cell}
FragmentLength: [58960×1 double]

4-150
 Analyzing Affymetrix SNP Arrays for DNA Copy Number Variants

Raw CN Estimation

The relative copy number at a SNP between two samples is estimated based on the log2 ratio of the
normalized signals. The averaged normalized signals from normal samples are used as the global
reference. The preprocessed reference mean log-transformed signals for the Mapping 50KXBa array
and the 50KHind array are provided in the MAT-files, SCLC_normal_Xba and SCLC_normal_Hind
respectively.

load SCLC_Normal_Xba

Estimate the log2 ratio of normalized signals.

log2Ratio_Xba = bsxfun(@minus, PM_Xba, mean_normal_PM_Xba);

Filtering and Ordering

SNPs probes with missing chromosome number, genomic position or fragment length in the
annotation file don't have enough information for further CN analysis. Also for CN analysis, Y
chromosomes are usually ignored. Filter out these SNP probes.

fidx = annot_Xba.Chromosome == -1 | annot_Xba.Chromosome == 24 |...

annot_Xba.ChromPosition == -1 | annot_Xba.FragmentLength == 0;
log2Ratio_Xba(fidx, :) = [];
chromosome_Xba = annot_Xba.Chromosome(~fidx);
genomepos_Xba = annot_Xba.ChromPosition(~fidx);
probesetids_Xba = XbaData.ProbeSetIDs(~fidx);
fragmentlen_Xba = annot_Xba.FragmentLength(~fidx);
accession_Xba = annot_Xba.Accession(~fidx);

Order CN estimation by chromosomes numbers:

[chr_sort, sidx] = sort(chromosome_Xba);

gpos_sort = genomepos_Xba(sidx);
log2Ratio_sort = log2Ratio_Xba(sidx, :);
probesetids_sort = probesetids_Xba(sidx);
fragmentlen_sort = fragmentlen_Xba(sidx);
accession_sort = accession_Xba(sidx);

Order CN estimation by chromosomal genomic positions:

u_chr = unique(chr_sort);
gpsidx = zeros(length(gpos_sort),1);
for i = 1:length(u_chr)
uidx = find(chr_sort == u_chr(i));
gp_s = gpos_sort(uidx);
[gp_ss, ssidx] = sort(gp_s);
s_res = uidx(ssidx);
gpsidx(uidx) = s_res;
end

gpos_ssort = gpos_sort(gpsidx);
log2Ratio_ssort = log2Ratio_sort(gpsidx, :);
probesetids_ssort = probesetids_sort(gpsidx);
fragmentlen_ssort = fragmentlen_sort(gpsidx);
accession_ssort = accession_sort(gpsidx);

4-151
4 Microarray Analysis

PCR Fragment Length Normalization

In the analysis, systematic effects from the PCR process should be taken into account. For example,
longer fragments usually result in less PCR amplification, which leads to less material to hybridize
and weaker signals. You can see this by plotting the raw CNs with fragment-length effect.

figure;
plot(fragmentlen_ssort, log2Ratio_ssort(:, 1), '.')
hold on
plot([0 2200], [0 0], '-.g')
xlim([0 2200])
ylim([-5 5])
xlabel('Fragment Length')
ylabel('log2(Ratio)')
title('Pre PCR fragment length normalization')

Nannya et al., 2005 introduced a robust linear model to estimate and remove this effect. For this
example, use the malowess function for PCR fragment length normalization for sample 1. Then
display the smooth fit curve.

smoothfit = malowess(fragmentlen_ssort,log2Ratio_ssort(:,1));
hold on
plot(fragmentlen_ssort, smoothfit, 'r+')
hold off

4-152
 Analyzing Affymetrix SNP Arrays for DNA Copy Number Variants

log2Ratio_norm = log2Ratio_ssort(:,1) - smoothfit;

Plot the PCR fragment length normalized raw CN estimation:

figure;
plot(fragmentlen_ssort, log2Ratio_norm, '.');
hold on
plot([0 2200], [0 0], '-.g')
xlim([0 2200])
ylim([-5 5])
xlabel('Fragment Length')
ylabel('log2(Ratio)')
title('Post PCR fragment length normalization')
hold off

4-153
4 Microarray Analysis

You can normalize PCR fragment length effect for all the samples using the malowess function.

Again, you can repeat the previous steps for the 50KHind array data.

CN Genome Profile

Load a MAT-file containing the preprocessed and normalized CN data from both the 50KXba arrays
and 50KHind arrays.

load SCLC_CN_Data

You can now plot the whole-genome profile of total CNs. For example, plot the whole-genome profile
for sample 1 (CL_H524) using a helper function plotcngenomeprofile.

plotcngenomeprofile(SCLC_CN.GenomicPosition,SCLC_CN.Log2Ratio(:, 1),...
SCLC_CN.Chromosome, 1:23, SCLC_CN.Sample{1})

4-154
 Analyzing Affymetrix SNP Arrays for DNA Copy Number Variants

You can also plot each chromosome CN profile in a subplot. For example, plot each chromosome CN
profile for sample 12 (PT_0177T):

plotcngenomeprofile(SCLC_CN.GenomicPosition,SCLC_CN.Log2Ratio(:, 12),...
SCLC_CN.Chromosome, 1:23, SCLC_CN.Sample{12}, 'S')

4-155
4 Microarray Analysis

8q Amplification in SCLS Samples

In the Zhao et al., 2005 study, a high-level amplification was observed in the q12.2-q12.3 region on
chromosome 8 for SCLS samples. You can perform CBS segmentation on chromosome 8 for sample
PT_S0177T.

sampleid = find(strcmpi(samples, 'PT_S0177T'));

ps = cghcbs(SCLC_CN, 'sampleind', sampleid, 'chromosome', 8, 'showplot', 8)

Analyzing: PT_S0177T. Current chromosome 8

ps =

struct with fields:

Sample: 'PT_S0177T'
SegmentData: [1×1 struct]

4-156
 Analyzing Affymetrix SNP Arrays for DNA Copy Number Variants

Add the ideogram for chromosome 8 to the plot:

chromosomeplot('hs_cytoBand.txt', 8, 'addtoplot', gca)

4-157
4 Microarray Analysis

Infer copy number changes:

segment_cn = ceil((2.^ps.SegmentData.Mean)*2);
cnv = segment_cn(segment_cn ~= 2);
startbp = ps.SegmentData.Start(segment_cn ~= 2)
endbp = ps.SegmentData.End(segment_cn ~= 2)
startMB = startbp/10^6;
endMB = endbp/10^6;

startbp =

62089326
62182830
128769526

endbp =

62182830
62729651
129006828

You can also get cytoband information for the CNVs. The function cytobandread returns cytoband
information in a structure.
ct = cytobandread('hs_cytoBand.txt')

4-158
 Analyzing Affymetrix SNP Arrays for DNA Copy Number Variants

ct =

struct with fields:

ChromLabels: {862×1 cell}

BandStartBPs: [862×1 int32]
BandEndBPs: [862×1 int32]
BandLabels: {862×1 cell}
GieStains: {862×1 cell}

Find cytoband labels for CNVs:

cn_cytobands = cell(length(cnv),1);
for i = 1:length(cnv)
istart = find(ct.BandStartBPs <= startbp(i) & ct.BandEndBPs >= startbp(i) & strcmp(ct.ChromLa
iend = find(ct.BandStartBPs <= endbp(i) & ct.BandEndBPs >= endbp(i) & strcmpi(ct.ChromLabels,
if strcmpi(ct.BandLabels{istart}, ct.BandLabels{iend})
cn_cytobands{i} = ['8' ct.BandLabels{istart}];
else
cn_cytobands{i} = ['8' ct.BandLabels{istart} '-' '8' ct.BandLabels{iend}];
end
end

Create a report displaying the start positions, end positions and size of the CNVs.

report = sprintf('Cytobands \tStart(Mb)\tEnd(Mb)\t\tSize(Mb)\tCN\n');

for i = 1:length(cnv)
report = sprintf('%s%-15s\t%3.2f\t\t%3.2f\t\t%3.2f\t\t%d\n',...
report, cn_cytobands{i},startMB(i),endMB(i),endMB(i)-startMB(i),cnv(i));
end
disp(report)

Cytobands Start(Mb) End(Mb) Size(Mb) CN

8q12.2 62.09 62.18 0.09 4
8q12.2-8q12.3 62.18 62.73 0.55 7
8q24.21 128.77 129.01 0.24 7

Among the three regions of amplification, the 8q12-13 region has been confirmed by interphase FISH
analysis (Zhao et al., 2005).

CN Gain/Loss Summary Plot

You can also visualize the fraction of samples with copy number amplifications of at least three copies
(red), and copy number losses to less than 1.5 copies (blue), across all SNPs for all SCLS samples.
The function cghfreqplot displays frequency of copy number alterations across multiple samples.
To better visualize the data, plot only the SNPs with gain or loss frequency over 25%.

gainThrd = log2(3/2);
lossThrd = log2(1.5/2);
cghfreqplot(SCLC_CN, 'Color', [1 0 0; 0 0 1],...
'Threshold', [gainThrd, lossThrd], 'cutoff', 0.25)
title('SCLC Summary Plot')

4-159
4 Microarray Analysis

4-160
 Analyzing Affymetrix SNP Arrays for DNA Copy Number Variants

References

[1] Zhao, X., et al., "Homozygous deletions and chromosome amplifications in human lung carcinomas
revealed by single nucleotide polymorphism array analysis", Cancer Research, 65(13):5561-70, 2005.

[2] Nannya, Y., et al., "A robust algorithm for copy number detection using high-density
oligonucleotide single nucleotide polymorphism genotyping arrays", Cancer Research, 65(14):6071-8,
2005.

4-161
4 Microarray Analysis

Working with GEO Series Data

This example shows how to retrieve gene expression data series (GSE) from the NCBI Gene
Expression Omnibus (GEO) and perform basic analysis on the expression profiles.

Introduction

The NCBI Gene Expression Omnibus (GEO) is the largest public repository of high-throughput
microarray experimental data. GEO data have four entity types including GEO Platform (GPL), GEO
Sample (GSM), GEO Series (GSE) and curated GEO DataSet (GDS).

A Platform record describes the list of elements on the array in the experiment (e.g., cDNAs,
oligonucleotide probesets). Each Platform record is assigned a unique and stable GEO accession
number (GPLxxx).

A Sample record describes the conditions under which an individual Sample was handled, the
manipulations it underwent, and the abundance measurement of each element derived from it. Each
Sample record is assigned a unique and stable GEO accession number (GSMxxx).

A Series record defines a group of related Samples and provides a focal point and description of the
whole study. Series records may also contain tables describing extracted data, summary conclusions,
or analyses. Each Series record is assigned a unique GEO accession number (GSExxx).

A DataSet record (GDSxxx) represents a curated collection of biologically and statistically

comparable GEO Samples. GEO DataSets (GDSxxx) are curated sets of GEO Sample data.

More information about GEO can be found in GEO Overview. Bioinformatics Toolbox™ provides
functions that can retrieve and parse GEO format data files. GSE, GSM, GSD and GPL data can be
retrieved by using the getgeodata function. The getgeodata function can also save the retrieved
data in a text file. GEO Series records are available in SOFT format files and in tab-delimited text
format files. The function geoseriesread reads the GEO Series text format file. The geosoftread
function reads the usually quite large SOFT format files.

In this example, you will retrieve the GSE5847 data set from GEO database, and perform statistical
testing on the data. GEO Series GSE5847 contains experimental data from a gene expression study of
tumor stroma and epithelium cells from 15 inflammatory breast cancer (IBC) cases and 35 non-
inflammatory breast cancer cases (Boersma et al. 2008).

Retrieving GEO Series Data

The function getgeodata returns a structure containing data retrieved from the GEO database. You
can also save the returned data in its original format to your local file system for use in subsequent
MATLAB® sessions. Note that data in public repositories is frequently curated and updated;
therefore the results of this example might be slightly different when you use up-to-date datasets.

gseData = getgeodata('GSE5847', 'ToFile', 'GSE5847.txt')

Use the geoseriesread function to parse the previously downloaded GSE text format file.

gseData = geoseriesread('GSE5847.txt')

gseData =

4-162
 Working with GEO Series Data

struct with fields:

Header: [1x1 struct]

Data: [22283x95 bioma.data.DataMatrix]

The structure returned contains a Header field that stores the metadata of the Series data, and a
Data field that stores the Series matrix data.

Exploring GSE Data

The GSE5847 matrix data in the Data field are stored as a DataMatrix object. A DataMatrix object is
a data structure like MATLAB 2-D array, but with additional metadata of row names and column
names. The properties of a DataMatrix can be accessed like other MATLAB objects.
get(gseData.Data)

Name: ''
RowNames: {22283x1 cell}
ColNames: {1x95 cell}
NRows: 22283
NCols: 95
NDims: 2
ElementClass: 'double'

The row names are the identifiers of the probe sets on the array; the column names are the GEO
Sample accession numbers.
gseData.Data(1:5, 1:5)

ans =

GSM136326 GSM136327 GSM136328 GSM136329 GSM136330

1007_s_at 10.45 9.3995 9.4248 9.4729 9.2788
1053_at 5.7195 4.8493 4.7321 4.7289 5.3264
117_at 5.9387 6.0833 6.448 6.1769 6.5446
121_at 8.0231 7.8947 8.345 8.1632 8.2338
1255_g_at 3.9548 3.9632 3.9641 4.0878 3.9989

The Series metadata are stored in the Header field. The structure contains Series information in the
Header.Series field, and sample information in the Header.Sample field.
gseData.Header

ans =

struct with fields:

Series: [1x1 struct]

Samples: [1x1 struct]

The Series field contains the title of the experiment and the microarray GEO Platform ID.
gseData.Header.Series

4-163
4 Microarray Analysis

ans =

struct with fields:

title: 'Tumor and stroma from breast by LCM'

geo_accession: 'GSE5847'
status: 'Public on Sep 30 2007'
submission_date: 'Sep 15 2006'
last_update_date: 'Nov 14 2012'
pubmed_id: '17999412'
summary: 'Tumor epithelium and surrounding stromal cells were isolated usi
overall_design: 'We applied LCM to obtain samples enriched in tumor epithelium an
type: 'Expression profiling by array'
contributor: 'Stefan,,Ambs...'
sample_id: 'GSM136326 GSM136327 GSM136328 GSM136329 GSM136330 GSM136331 GSM1
contact_name: 'Stefan,,Ambs'
contact_laboratory: 'LHC'
contact_institute: 'NCI'
contact_address: '37 Convent Dr Bldg 37 Room 3050'
contact_city: 'Bethesda'
contact_state: 'MD'
contact_zip0x2Fpostal_code: '20892'
contact_country: 'USA'
supplementary_file: 'ftp://ftp.ncbi.nlm.nih.gov/pub/geo/DATA/supplementary/series/GSE
platform_id: 'GPL96'
platform_taxid: '9606'
sample_taxid: '9606'
relation: 'BioProject: http://www.ncbi.nlm.nih.gov/bioproject/97251'

gseData.Header.Samples

ans =

struct with fields:

title: {1x95 cell}

geo_accession: {1x95 cell}
status: {1x95 cell}
submission_date: {1x95 cell}
last_update_date: {1x95 cell}
type: {1x95 cell}
channel_count: {1x95 cell}
source_name_ch1: {1x95 cell}
organism_ch1: {1x95 cell}
characteristics_ch1: {2x95 cell}
molecule_ch1: {1x95 cell}
extract_protocol_ch1: {1x95 cell}
label_ch1: {1x95 cell}
label_protocol_ch1: {1x95 cell}
taxid_ch1: {1x95 cell}
hyb_protocol: {1x95 cell}
scan_protocol: {1x95 cell}
description: {1x95 cell}
data_processing: {1x95 cell}
platform_id: {1x95 cell}

4-164
 Working with GEO Series Data

contact_name: {1x95 cell}

contact_laboratory: {1x95 cell}
contact_institute: {1x95 cell}
contact_address: {1x95 cell}
contact_city: {1x95 cell}
contact_state: {1x95 cell}
contact_zip0x2Fpostal_code: {1x95 cell}
contact_country: {1x95 cell}
supplementary_file: {1x95 cell}
data_row_count: {1x95 cell}

The data_processing field contains the information of the preprocessing methods, in this case the
Robust Multi-array Average (RMA) procedure.

gseData.Header.Samples.data_processing(1)

ans =

1x1 cell array

{'RMA'}

The source_name_ch1 field contains the sample source:

sampleSources = unique(gseData.Header.Samples.source_name_ch1);

sampleSources{:}

ans =

'human breast cancer stroma'

ans =

'human breast cancer tumor epithelium'

The field Header.Samples.characteristics_ch1 contains the characteristics of the samples.

gseData.Header.Samples.characteristics_ch1(:,1)

ans =

2x1 cell array

{'IBC' }
{'Deceased'}

Determine the IBC and non-IBC labels for the samples from the
Header.Samples.characteristics_ch1 field as group labels.

sampleGrp = gseData.Header.Samples.characteristics_ch1(1,:);

4-165
4 Microarray Analysis

Retrieving GEO Platform (GPL) Data

The Series metadata told us the array platform id: GPL96, which is an Affymetrix® GeneChip®
Human Genome U133 array set HG-U133A. Retrieve the GPL96 SOFT format file from GEO using the
getgeodata function. For example, assume you used the getgeodata function to retrieve the
GPL96 Platform file and saved it to a file, such as GPL96.txt. Use the geosoftread function to
parse this SOFT format file.

gplData = geosoftread('GPL96.txt')

gplData =

struct with fields:

Scope: 'PLATFORM'
Accession: 'GPL96'
Header: [1x1 struct]
ColumnDescriptions: {16x1 cell}
ColumnNames: {16x1 cell}
Data: {22283x16 cell}

The ColumnNames field of the gplData structure contains names of the columns for the data:

gplData.ColumnNames

ans =

16x1 cell array

{'ID' }
{'GB_ACC' }
{'SPOT_ID' }
{'Species Scientific Name' }
{'Annotation Date' }
{'Sequence Type' }
{'Sequence Source' }
{'Target Description' }
{'Representative Public ID' }
{'Gene Title' }
{'Gene Symbol' }
{'ENTREZ_GENE_ID' }
{'RefSeq Transcript ID' }
{'Gene Ontology Biological Process'}
{'Gene Ontology Cellular Component'}
{'Gene Ontology Molecular Function'}

You can get the probe set ids and gene symbols for the probe sets of platform GPL69.

gplProbesetIDs = gplData.Data(:, strcmp(gplData.ColumnNames, 'ID'));

geneSymbols = gplData.Data(:, strcmp(gplData.ColumnNames, 'Gene Symbol'));

Use gene symbols to label the genes in the DataMatrix object gseData.Data. Be aware that the
probe set IDs from the platform file may not be in the same order as in gseData.Data. In this
example they are in the same order.

4-166
 Working with GEO Series Data

Change the row names of the expression data to gene symbols.

gseData.Data = rownames(gseData.Data, ':', geneSymbols);

Display the first five rows and five columns of the expression data with row names as gene symbols.

gseData.Data(1:5, 1:5)

ans =

GSM136326 GSM136327 GSM136328 GSM136329 GSM136330

DDR1 10.45 9.3995 9.4248 9.4729 9.2788
RFC2 5.7195 4.8493 4.7321 4.7289 5.3264
HSPA6 5.9387 6.0833 6.448 6.1769 6.5446
PAX8 8.0231 7.8947 8.345 8.1632 8.2338
GUCA1A 3.9548 3.9632 3.9641 4.0878 3.9989

Analyzing the Data

Bioinformatics Toolbox and Statistics and Machine Learning Toolbox™ offer a wide spectrum of
analysis and visualization tools for microarray data analysis. However, because it is not our main goal
to explain the analysis methods in this example, you will apply only a few of the functions to the gene
expression profile from stromal cells. For more elaborate examples about feature selection tools
available, see “Select Features for Classifying High-Dimensional Data”.

The experiment was performed on IBC and non-IBC samples derived from stromal cells and epithelial
cells. In this example, you will work with the gene expression profile from stromal cells. Determine
the sample indices for the stromal cell type from the
gseData.Header.Samples.source_name_ch1 field.

stromaIdx = strcmpi(sampleSources{1}, gseData.Header.Samples.source_name_ch1);

Determine number of samples from stromal cells.

nStroma = sum(stromaIdx)

nStroma =

47

stromaData = gseData.Data(:, stromaIdx);

stromaGrp = sampleGrp(stromaIdx);

Determine the number of IBC and non-IBC stromal cell samples.

nStromaIBC = sum(strcmp('IBC', stromaGrp))

nStromaIBC =

13

nStromaNonIBC = sum(strcmp('non-IBC', stromaGrp))

4-167
4 Microarray Analysis

nStromaNonIBC =

34

You can also label the columns in stromaData with the group labels:

stromaData = colnames(stromaData, ':', stromaGrp);

Display the histogram of the normalized gene expression measurements of a specified gene. The x-
axes represent the normalized expression level. For example, inspect the distribution of the gene
expression values of these genes.

fID = 331:339;

zValues = zscore(stromaData.(':')(':'), 0, 2);

bw = 0.25;
edges = -10:bw:10;
bins = edges(1:end-1) + diff(edges)/2;

histStroma = histc(zValues(fID, :)', edges) ./ (stromaData.NCols*bw);

figure;
for i = 1:length(fID)
subplot(3,3,i);
bar(edges, histStroma(:,i), 'histc')
xlim([-3 3])
if i <= length(fID)-3
ax = gca;
ax.XTickLabel = [];
end
title(sprintf('gene%d - %s', fID(i), stromaData.RowNames{fID(i)}))
end
sgtitle('Gene Expression Value Distributions')

4-168
 Working with GEO Series Data

The gene expression profile was accessed using the Affymetrix GeneChip more than 22,000 features
on a small number of samples (~100). Among the 47 tumor stromal samples, there are 13 IBC and 34
non-IBC. But not all the genes are differentially expressed between IBC and non-IBC tumors.
Statistical tests are needed to identify a gene expression signature that distinguish IBC from non-IBC
stromal samples.

Use genevarfilter to filter out genes with a small variance across samples.

[mask, stromaData] = genevarfilter(stromaData);

stromaData.NRows

ans =

20055

Apply a t-statistic on each gene and compare p-values for each gene to find significantly differentially
expressed genes between IBC and non-IBC groups by permuting the samples (1,000 times for this
example).

rng default
[pvalues, tscores] = mattest(stromaData(:, 'IBC'), stromaData(:, 'non-IBC'),...
'Showhist', true', 'showplot', true, 'permute', 1000);

4-169
4 Microarray Analysis

4-170
 Working with GEO Series Data

Select the genes at a specified p-value.

sum(pvalues < 0.001)

ans =

52

There are about 50 genes selected directly at p-values < 0.001.

Sort and list the top 20 genes:

testResults = [pvalues, tscores];

testResults = sortrows(testResults);
testResults(1:20, :)

ans =

p-values t-scores
INPP5E 2.3318e-05 5.0389
ARFRP1 /// IGLJ3 2.7575e-05 4.9753
USP46 3.4336e-05 -4.9054
GOLGB1 4.7706e-05 -4.7928
TTC3 0.00010695 -4.5053
THUMPD1 0.00013164 -4.4317

4-171
4 Microarray Analysis

0.00016042 4.3656
MAGED2 0.00017042 -4.3444
DNAJB9 0.0001782 -4.3266
KIF1C 0.00022122 4.2504
0.00022237 -4.2482
DZIP3 0.00022414 -4.2454
COPB1 0.00023199 -4.2332
PSD3 0.00024649 -4.2138
PLEKHA4 0.00026505 4.186
DNAJB9 0.0002767 -4.1708
CNPY2 0.0002801 -4.1672
USP9X 0.00028442 -4.1619
SEC22B 0.00030146 -4.1392
GFER 0.00030506 -4.1352

References

[1] Boersma, B.J., Reimers, M., Yi, M., Ludwig, J.A., et al. "A stromal gene signature associated with
inflammatory breast cancer", International Journal of Cancer, 122(6):1324-32, 2008.

4-172
 Identifying Biomolecular Subgroups Using Attractor Metagenes

Identifying Biomolecular Subgroups Using Attractor

Metagenes

This example shows workflows for the analysis of gene expression data with the attractor metagene
algorithm. Gene expression data is available for many model organisms and disease conditions. This
example shows how to use the metafeatures function to explore biomolecular phenotypes in breast
cancer.

The Cancer Genome Atlas Data

The Cancer Genome Atlas (TCGA) includes several kinds of data across multiple cancer indications.
TCGA includes measurements of gene expression, protein expression, clinical outcomes, and more. In
this example, you explore breast cancer gene expression.

Researchers collected tumor samples, and used Agilent G4502A microarrays to measure their gene
expression. In this example you use the Level-3 expression data, which has been post-processed from
the original measurements into the expression calls. Data was retrieved May 20, 2014.

Load the data into MATLAB®. The MAT-file TCGA_Breast_Gene_Expression.mat contains gene
expression data of 17814 genes for 590 different patients. The expression data is stored in the
variable geneExpression. The gene names are stored in the variable geneNames.
load TCGA_Breast_Gene_Expression

To see for the organization of the data, check number of genes and samples in this data set.
size(geneExpression)

ans = 1×2

17814 590

geneNames is a cell array of the gene names. You can access the entries using MATLAB cell array
indexing:
geneNames{655}

ans =
'EGFR'

This cell array indicates that the 655th row of the variable geneExpression contains expression
measurements for the gene expression of Epidermal Growth Factor Receptor (EGFR).

Attractor Metagene Algorithm

The attractor metagene algorithm was developed as part of the DREAM 8 challenge to develop
prognostic biomarkers for breast cancer survival. The attractor metagene approach discovers and
quantifies underlying biomolecular events. These events reduce the dimensionality of the gene
expression data, and also allow for subtype classification and investigation of regulatory machinery
[1].

A metagene is defined as any weighted sum of gene expression. Suppose you have a collection of co-
expressed genes. You can create a metagene by averaging the expression levels of the genes in the
collection.

4-173
4 Microarray Analysis

There is the potential to refine our understanding of the gene expression captured in this metagene.
Suppose you create a set of weights that quantify the similarity between the genes in our collection
and the metagene. Genes that are more similar to the metagene receive larger weights, while genes
that are less similar receive smaller weights. Using these new weights, you can form a new metagene
that is a weighted average of gene expression. The new metagene better captures a biomolecular
event that governs some element of gene regulation in the expression data.

This procedure forms the core of the attractor metagene algorithm. Form a metagene using some
current estimate of the weights, then update the weights based on a measure of similarity. Attractor
metagenes are defined as the attracting fixed points of this iterative process.

The algorithm exists within the broad family of unsupervised machine learning algorithms. Related
algorithms include principal component analysis, various clustering algorithms (especially fuzzy c-
means), non-negative matrix factorization, and others. The main advantage of the metagene approach
is that the results of the algorithm tend to be more clearly linked with a phenotype defined by gene
expression.

Concretely, in the ith iteration of the algorithm. You have a vector of weights, Wi , of size 1-by-number
of genes. The estimate of the metagene during the i th iteration is:

Mi = Wi * G

G is the number of genes by number of samples gene expression matrix. To update the weights:

W j, i + 1 = J(Mi, G j)

W j, i + 1 is the j th element of Wi + 1, G j is the j th row of G, and J is a similarity metric. In the

metagene attractor algorithm, J is defined as:

α
J(Mi, G j) = MI(Mi, G j)

if the correlation between Mi and G j is greater than 0. MI is the mutual information between Mi and
G j. The function metafeatures uses the B-spline estimator of mutual information described in [3].

If, instead, the correlation between Mi and G j is less than or equal to 0, then:

J(Mi, G j) = 0

The weights are all greater than or equal to zero. Because mutual information is scale invariant, you
can normalize the weights in whatever way you choose. Here, they are normalized so their sum is 1.

The algorithm is initialized by either random or user-selected weights. It proceeds until the change in
Mi between iterations is small, or a prespecified number of iterations is exhausted.

Cleaning the Data

The data has several NaN values. To check how many, sum over an indicator returned by isnan.

sum(sum(isnan(geneExpression)))

ans = 1695

Out of the approximately 10 million entries of geneExpression, there are 1695 missing entries.
Before proceeding you will need to deal with these missing entries.

4-174
 Identifying Biomolecular Subgroups Using Attractor Metagenes

There are several ways to impute these missing values. You can use a simple method called K nearest
neighbor imputation supplied by the Bioinformatics Toolbox (TM). K-nearest neighbor imputation
works by replacing missing data with the corresponding value from a weighted average of the k
nearest columns to the column with the missing data.

Use k = 3, and replace the current value of geneExpression with one that has no NaN values.
geneExpression = knnimpute(geneExpression,3);

The variable geneExpression has no NaN values.

sum(sum(isnan(geneExpression)))

ans = 0

For more information about knnimpute, see the Bioinformatics Toolbox documentation.
doc knnimpute

Identifying Biomolecular Events Using the Attractor Metagene Algorithm

The function metafeatures uses the attractor metagene algorithm to identify motifs of gene
regulation.

Setup an options structure. In this case, set the display to provide the information about the
algorithm at each iteration.
opts = struct('Display','iter');

metafeatures also allows for specifying start values. You can seed the starting weights to
emphasize genes that you are interested in. There are three common drivers of breast cancer, ERBB2
(also called HER2), estrogen, and progestrone.

Set the weight for each of these genes to 1 in three different rows of startValues. Each row
corresponds to initial values for a different replicate. strcmp compares the genes of interest and the
list of genes in the data set. find returns the index in the list of the gene.
erbb = find(strcmp('ERBB2',geneNames));
estrogen = find(strcmp('ESR1',geneNames));
progestrone = find(strcmp('PGR',geneNames));

startValues = zeros(size(geneExpression,1),3);
startValues(erbb,1) = 1;
startValues(estrogen,2) = 1;
startValues(progestrone,3) = 1;

Call metafeatures with the imputed data set. The second argument, geneNames is the list of all the
genes in the data set. Supplying the gene names is not required. However, the gene names can allow
exploration of the highly ranked genes that are returned by the algorithm to get insights into the
biomolecular event described by the metagene.
[meta, weights, genes_sorted] = metafeatures(geneExpression,geneNames,'start',startValues,'option

Caching self information ...

... done. Took 108.8018 seconds.
Caching entropy and binning information...
... done. Took 51.8503 seconds.
non-zero

4-175
4 Microarray Analysis

Found iter diff weights

1 1 1.26e+01 8924
1 2 7.29e+00 8885
1 3 4.22e+00 8796
1 4 2.54e+00 8761
1 5 1.63e+00 8745
1 6 1.14e+00 8720
1 7 8.59e-01 8706
1 8 7.18e-01 8682
1 9 7.04e-01 8687
1 10 6.44e-01 8680
1 11 5.53e-01 8676
1 12 4.56e-01 8664
1 13 3.67e-01 8654
1 14 2.91e-01 8649
1 15 2.30e-01 8642
1 16 1.83e-01 8636
1 17 1.46e-01 8634
1 18 1.17e-01 8631
1 19 9.45e-02 8632
1 20 7.65e-02 8634
1 21 6.22e-02 8633
1 22 5.06e-02 8631
1 23 4.13e-02 8635
1 24 3.38e-02 8639
1 25 2.76e-02 8636
1 26 2.26e-02 8633
1 27 1.85e-02 8633
1 28 1.51e-02 8635
1 29 1.24e-02 8635
1 30 1.02e-02 8634
1 31 8.35e-03 8633
1 32 6.85e-03 8633
1 33 5.57e-03 8633
1 34 4.59e-03 8631
1 35 3.78e-03 8631
1 36 3.07e-03 8632
1 37 2.53e-03 8632
1 38 2.06e-03 8632
1 39 1.70e-03 8632
1 40 1.40e-03 8632
1 41 1.15e-03 8632
1 42 9.24e-04 8632
1 43 7.70e-04 8632
1 44 6.21e-04 8632
1 45 5.20e-04 8632
1 46 4.43e-04 8632
1 47 3.49e-04 8632
1 48 2.97e-04 8632
1 49 2.36e-04 8632
1 50 1.93e-04 8632
1 51 1.56e-04 8632
1 52 1.42e-04 8632
1 53 8.98e-05 8632
1 54 9.72e-05 8632
1 55 5.37e-05 8632
1 56 7.47e-05 8632
1 57 5.17e-05 8632

4-176
 Identifying Biomolecular Subgroups Using Attractor Metagenes

1 58 4.81e-05 8632
1 59 2.85e-05 8632
1 60 1.97e-05 8632
1 61 3.05e-05 8632
1 62 1.41e-05 8632
1 63 1.02e-05 8632
1 64 7.89e-06 8632
1 65 9.34e-06 8632
1 66 2.07e-05 8632
1 67 1.52e-05 8632
1 68 2.26e-05 8632
1 69 1.55e-05 8632
1 70 2.24e-05 8632
1 71 1.75e-05 8632
1 72 2.01e-05 8632
1 73 6.47e-06 8632
1 74 1.62e-05 8632
1 75 2.23e-05 8632
1 76 1.93e-05 8632
1 77 1.71e-05 8632
1 78 6.94e-06 8632
1 79 3.21e-06 8632
1 80 1.58e-05 8632
1 81 2.02e-05 8632
1 82 1.99e-05 8632
1 83 2.12e-05 8632
1 84 1.79e-05 8632
1 85 1.60e-05 8632
1 86 1.78e-05 8632
1 87 1.87e-05 8632
1 88 1.66e-05 8632
1 89 5.98e-06 8632
1 90 1.26e-05 8632
1 91 2.14e-05 8632
1 92 1.82e-05 8632
1 93 6.97e-06 8632
1 94 1.04e-05 8632
1 95 2.13e-05 8632
1 96 6.39e-06 8632
1 97 1.75e-05 8632
1 98 2.37e-05 8632
1 99 2.01e-05 8632
1 100 1.98e-05 8632

Warning: 'Maximum iterations exceeded, terminating early.'

2 1 1.93e+01 9893
2 2 6.04e+00 9885
2 3 3.80e+00 9883
2 4 2.53e+00 9886
2 5 1.73e+00 9881
2 6 1.13e+00 9873
2 7 7.19e-01 9869
2 8 4.63e-01 9866
2 9 3.08e-01 9870
2 10 2.13e-01 9874
2 11 1.54e-01 9872
2 12 1.15e-01 9874

4-177
4 Microarray Analysis

2 13 8.72e-02 9874
2 14 6.68e-02 9874
2 15 5.14e-02 9874
2 16 3.97e-02 9875
2 17 3.07e-02 9875
2 18 2.37e-02 9873
2 19 1.84e-02 9871
2 20 1.42e-02 9871
2 21 1.10e-02 9871
2 22 8.54e-03 9872
2 23 6.62e-03 9872
2 24 5.05e-03 9872
2 25 4.01e-03 9872
2 26 3.09e-03 9872
2 27 2.38e-03 9872
2 28 1.85e-03 9872
2 29 1.43e-03 9872
2 30 1.09e-03 9872
2 31 8.46e-04 9872
2 32 6.73e-04 9872
2 33 5.10e-04 9872
2 34 3.81e-04 9872
2 35 2.98e-04 9872
2 36 2.46e-04 9872
2 37 1.51e-04 9872
2 38 1.63e-04 9872
2 39 1.15e-04 9872
2 40 7.11e-05 9872
2 41 1.18e-04 9872
2 42 7.28e-05 9872
2 43 1.89e-05 9872
2 44 4.24e-05 9872
2 45 1.60e-05 9872
2 46 6.75e-06 9872
2 47 4.81e-05 9872
2 48 2.47e-05 9872
2 49 1.04e-05 9872
2 50 7.46e-06 9872
2 51 9.31e-06 9872
2 52 5.25e-06 9872
2 53 3.89e-05 9872
2 54 9.38e-06 9872
2 55 3.33e-05 9872
2 56 1.48e-05 9872
2 57 2.45e-05 9872
2 58 2.58e-05 9872
2 59 1.00e-05 9872
2 60 1.86e-05 9872
2 61 5.87e-05 9872
2 62 2.97e-05 9872
2 63 1.07e-05 9872
2 64 8.84e-06 9872
2 65 8.29e-06 9872
2 66 1.58e-05 9872
2 67 1.48e-05 9872
2 68 5.00e-06 9872
2 69 2.74e-05 9872
2 70 1.20e-05 9872

4-178
 Identifying Biomolecular Subgroups Using Attractor Metagenes

2 71 2.91e-05 9872
2 72 9.45e-06 9872
2 73 1.75e-05 9872
2 74 1.56e-05 9872
2 75 6.56e-06 9872
2 76 1.79e-05 9872
2 77 2.67e-05 9872
2 78 5.55e-05 9872
2 79 2.55e-05 9872
2 80 1.03e-05 9872
2 81 2.74e-05 9872
2 82 2.04e-05 9872
2 83 1.00e-05 9872
2 84 1.11e-05 9872
2 85 9.83e-06 9872
2 86 2.71e-05 9872
2 87 1.42e-05 9872
2 88 1.28e-05 9872
2 89 2.24e-05 9872
2 90 4.58e-05 9872
2 91 3.36e-05 9872
2 92 9.74e-06 9872
2 93 1.06e-05 9872
2 94 1.50e-05 9872
2 95 5.05e-05 9872
2 96 1.12e-05 9872
2 97 2.52e-05 9872
2 98 9.77e-06 9872
2 99 6.10e-06 9872
2 100 2.97e-05 9872

Warning: 'Maximum iterations exceeded, terminating early.'

3 1 3.75e+00 9963
3 2 1.08e+00 9966
3 3 4.29e-01 9959
3 4 1.87e-01 9961
3 5 8.45e-02 9958
3 6 3.88e-02 9957
3 7 1.80e-02 9956
3 8 8.36e-03 9956
3 9 3.89e-03 9956
3 10 1.78e-03 9956
3 11 8.68e-04 9956
3 12 3.96e-04 9956
3 13 1.89e-04 9956
3 14 8.92e-05 9956
3 15 4.25e-05 9956
3 16 1.16e-05 9956
3 17 1.57e-05 9956
3 18 1.67e-05 9956
3 19 1.59e-05 9956
3 20 1.07e-05 9956
3 21 9.21e-06 9956
3 22 1.59e-05 9956
3 23 6.23e-06 9956
3 24 8.68e-06 9956
3 25 1.56e-05 9956

4-179
4 Microarray Analysis

3 26 1.55e-05 9956
3 27 9.65e-06 9956
3 28 9.74e-06 9956
3 29 9.75e-06 9956
3 30 9.75e-06 9956
3 31 9.84e-06 9956
3 32 1.49e-05 9956
3 33 1.05e-05 9956
3 34 1.43e-05 9956
3 35 2.14e-05 9956
3 36 6.64e-06 9956
3 37 1.54e-06 9956
3 38 2.23e-06 9956
3 39 2.98e-06 9956
3 40 8.89e-06 9956
3 41 1.60e-05 9956
3 42 1.06e-05 9956
3 43 9.08e-06 9956
3 44 1.60e-05 9956
3 45 6.74e-06 9956
3 46 8.71e-06 9956
3 47 9.45e-06 9956
3 48 1.48e-05 9956
3 49 1.05e-05 9956
3 50 1.43e-05 9956
3 51 2.15e-05 9956
3 52 6.64e-06 9956
3 53 1.41e-06 9956
3 54 1.98e-06 9956
3 55 2.58e-06 9956
3 56 9.07e-06 9956
3 57 6.54e-06 9956
3 58 5.44e-06 9956
3 59 4.36e-06 9956
3 60 8.43e-06 9956
3 61 1.08e-05 9956
3 62 9.73e-06 9956
3 63 9.72e-06 9956
3 64 9.72e-06 9956
3 65 9.75e-06 9956
3 66 9.78e-06 9956
3 67 9.82e-06 9956
3 68 1.50e-05 9956
3 69 1.02e-05 9956
3 70 1.34e-05 9956
3 71 2.08e-05 9956
3 72 1.30e-05 9956
3 73 2.11e-05 9956
3 74 1.56e-05 9956
3 75 9.45e-06 9956
3 76 1.48e-05 9956
3 77 1.11e-05 9956
3 78 8.97e-06 9956
3 79 1.31e-05 9956
3 80 2.19e-05 9956
3 81 8.99e-06 9956
3 82 1.60e-05 9956
3 83 7.51e-06 9956

4-180
 Identifying Biomolecular Subgroups Using Attractor Metagenes

3 84 6.78e-06 9956
3 85 7.51e-06 9956
3 86 1.10e-05 9956
3 87 1.39e-05 9956
3 88 6.38e-06 9956
3 89 6.05e-06 9956
3 90 4.66e-06 9956
3 91 7.28e-06 9956
3 92 7.98e-06 9956
3 93 1.15e-05 9956
3 94 8.72e-06 9956
3 95 1.56e-05 9956
3 96 1.82e-05 9956
3 97 1.23e-05 9956
3 98 6.69e-06 9956
3 99 1.63e-06 9956
3 100 1.15e-06 9956

Warning: 'Maximum iterations exceeded, terminating early.'

The variable meta has the value of the three metagenes discovered for each sample. You can plot the
three metagenes to gain insight into the nature of gene regulation across different phenotypes of
breast cancer.
plot3(meta(1,:),meta(2,:),meta(3,:),'o')
xlabel('ERBB2 metagene')
ylabel('Estrogen metagene')
zlabel('Progestrone metagene')

4-181
4 Microarray Analysis

In the plot you can observe a few things.

In the plot, there is a group of points bunched together with low values for all three metagenes.
Based on mRNA levels, the expectation is that points are associated with tumor samples that are
triple-negative or basal type.

There is also a group of points that have high estrogen receptor metagene expression. This group
spans both high and low progestrone metagene expression. There are no points with high
progestrone metagene expression and low estrogen metagene expression. This finding is consistent
with the observation that ER-/PR+ breast cancers are extremely rare [2].

The remaining points are the ERBB2 positive cancers. They have less representation in this data set
than the hormone-driven and triple-negative cancers. There are also no firmly established
relationships between hormone receptor expression and ERBB2 status.

To develop a better understanding of the gene regulation captured by the metagenes, take a closer
look at the metagene discovered by initializing the estrogen receptor to have weight 1. You can list
the top ten genes contributing to the metagene for the 11th metagene discovered.
genes_sorted(1:10,2)

ans = 10x1 cell

{'AGR3' }
{'ESR1' }
{'CA12' }
{'AGR2' }
{'MLPH' }
{'FOXA1'}
{'THSD4'}
{'FSIP1'}
{'ANXA9'}
{'XBP1' }

This metagene captures the biomolecular event associated with the transition to estrogen-driven
breast cancer. The four, top-ranked, genes listed are:

• Anterior Gradient Homolog 3 (AGR3)

• Estrogen Receptor 1 (ESR1)
• Carbonic anhydrase 12 (CA12)
• Anterior Gradient Homolog 2 (AGR2)

Transcriptional changes in each of these genes are implicated in estrogen-driven breast cancer. The
three genes other than ESR1 are known to be coexpressed with ESR1. Identification of these genes
illustrates the power of the attractor metagene algorithm to link gene expression with phenotypes.

Similar versions of the estrogen metagene and the ERBB2 metagene are described in [1]. The
ordering of the gene contributions differs slightly between this analysis and [1] because a different
breast cancer data set was used. Variations in the weights are to be expected, but the ordering of the
genes by weights are roughly the same. Specifically, genes with the top 10 weights are mostly the
same between this version, and the version described in [1]. Similarly, there is significant overlap
between the genes with the top 100 weights.

Genes can contribute to multiple metagenes. In this sense, the attractor metagene algorithm is a
"soft" clustering technique. In this example, finding metagenes in breast cancer data, there is overlap

4-182
 Identifying Biomolecular Subgroups Using Attractor Metagenes

in the sets of genes that have larger contribution weights to the estrogen and progestrone
metagenes.

If a weight is "elevated" when it is larger than .001, then:

elevated_weights = weights>.001;

The column sum of the elevated_weights is the total number of elevated weights in each of the
three metagenes.
sum(elevated_weights)

ans = 1×3

19 96 27

Of the 96 elevated weights for the estrogen metagene, and the 27 for the progestrone metagene,
there are 22 elevated weights that are in both sets.
sum(elevated_weights(:,2) & elevated_weights(:,3))

ans = 22

However, there is no overlap between the ERBB2 metagene and the estrogen metagene:
sum(elevated_weights(:,1) & elevated_weights(:,2))

ans = 0

as well as no overlap between the ERBB2 metagene and the progestrone metagene:
sum(elevated_weights(:,1) & elevated_weights(:,3))

ans = 0

The Role of Alpha

In the similarity metric of the algorithm, the parameter alpha controls the degree of nonlinearity. As
alpha is increased, the number of metagenes tends to increase. The default alpha is 5, because this
value was good for the work in [1], but for different data sets or use cases, you must adjust alpha.

To illustrate the effects of alpha, if alpha is 1 in the breast cancer analysis, then the progesterone and
estrogen metagenes are not distinct.
[meta_alpha_1, weights_alpha_1, genes_sorted_alpha_1] = ...
metafeatures(geneExpression,geneNames,'start',startValues,'alpha',1);

Warning: 'Maximum iterations exceeded, terminating early.'

Warning: 'Maximum iterations exceeded, terminating early.'

Warning: 'Maximum iterations exceeded, terminating early.'

In this case, only two metagenes are returned, despite the fact that we ran the algorithm three times.
size(meta_alpha_1)

ans = 1×2

4-183
4 Microarray Analysis

2 590

This result is because, by default, metafeatures returns only the unique metagenes. The
initialization with the weight for ESR1 set to 1, and the initialization with the weight for PGR set to 1,
both converge to metagenes that are effectively the same.

References

[1] Cheng, Wei-Yi, Tai-Hsien Ou Yang, and Dimitris Anastassiou. "Biomolecular events in cancer
revealed by attractor metagenes." PLoS computational biology 9.2 (2013): e1002920.

[2] Hefti, Marco M., et al."Estrogen receptor negative/progesterone receptor positive breast cancer is
not a reproducible subtype." Breast Cancer Research 15.4 (2013): R68.

[3] Daub, Carsten O., et al. "Estimating mutual information using B-spline functions?an improved
similarity measure for analysing gene expression data." BMC bioinformatics 5.1 (2004): 118.

4-184
 Working with the Clustergram Function

Working with the Clustergram Function

This example shows how to work with the clustergram function.

The clustergram function creates a heat map with dendrograms to show hierarchical clustering of
data. These types of heat maps have become a standard visualization method for microarray data
since first applied by Eisen et al. [1]. This example illustrates some of the options of the
clustergram function. The example uses data from the van't Veer et al. breast cancer microarray
study [2].

Importing Data

A study by van't Veer et al. investigated whether tumor ability for metastasis is obtained later in
development or inherent in the initial gene expression signature [2]. The study analyzed tumor
samples from 117 young breast cancer patients, of whom 78 were sporadic lymph-node-negative. The
gene expression profiles of these 78 patients were searched for prognostic signatures. Of the 78
patients, 44 exhibited non-recurrences within five years of surgical treatment while 34 had
recurrences. Samples were hybridized to Agilent® two-color oligonucleotide microarrays
representing approximately 25,000 human genes. The authors selected 4,918 significant genes that
had at least a two-fold differential expression relative to the reference and a p-value for being
expressed < 0.01 in at least 3 samples. By using supervised classification, the authors identified a
poor prognosis gene expression signature of 231 genes [2].

A subset of the preprocessed gene expression data from [2] is provided in the
bc_train_filtered.mat MAT-file. Samples for 78 lymph-node-negative patients are included, each
one containing the gene expression values for the 4,918 significant genes. Gene expression values
have already been preprocessed, by normalization and background subtraction, as described in [2].

load bc_train_filtered
bcTrainData

bcTrainData =

struct with fields:

Samples: {78x1 cell}

Log10Ratio: [4918x78 single]
Accession: {4918x1 cell}

The list of 231 genes in the prognosis profile proposed by van't Veer et al. is also provided in the
bc_proggenes231.mat MAT-file. Genes are ordered according to their correlation coefficient with
the prognostic groups.

load bc_proggenes231

Extract the gene expression values for the prognosis profile.

[tf, idx] = ismember(bcProgGeneList.Accession, bcTrainData.Accession);

progValues = bcTrainData.Log10Ratio(idx, :);
progAccession = bcTrainData.Accession(idx);
progSamples = bcTrainData.Samples;

4-185
4 Microarray Analysis

For this example, you will work with the 35 most positive correlated genes and the 35 most negative
correlated genes.

progValues = progValues([1:35 197:231],:);

progAccession = progAccession([1:35 197:231]);

Clustering

You will use the clustergram function to perform hierarchical clustering and generate a heat map
and dendrogram of the data. The simplest form of clustergram clusters the rows or columns of a
data set using Euclidean distance metric and average linkage. In this example, you will cluster the
samples (columns) only.

The matrix of gene expression data, progValues, contains some missing data. These are marked as
NaN. You need to provide an imputation function name or function handle to impute values for
missing data. In this example, you will use the k-nearest neighbors imputation procedure
implemented in the function knnimpute.

cg_s = clustergram(progValues, 'RowLabels', progAccession,...

'ColumnLabels', progSamples,...
'Cluster', 'Row',...
'ImputeFun', @knnimpute)

Clustergram object with 78 columns of nodes.

4-186
 Working with the Clustergram Function

The dendrogram at the top of the heat map shows the clustering of samples. The missing data are
shown in the heat map in gray. The data has been standardized across all samples for each gene, so
that the mean is 0 and the standard deviation is 1.

Inspecting and Changing Clustering Options

You can determine and change properties of a clustergram object. For example, you can find out
which distance metric was used in the clustering.
cg_s.ColumnPDist

ans =

1x1 cell array

{'Euclidean'}

Then you can change the distance metric for the columns to correlation.
cg_s.ColumnPDist = 'correlation';

4-187
4 Microarray Analysis

By changing the distance metric from Euclidean to correlation, the tumor samples are clearly
clustered into a good prognosis group and a poor prognosis group.

To see all the properties of the clustergram, simply use the get method.

get(cg_s)

Cluster: 'ROW'
RowPDist: {'Euclidean'}
ColumnPDist: {'correlation'}
Linkage: {'Average'}
Dendrogram: {}
OptimalLeafOrder: 1
LogTrans: 0
DisplayRatio: [0.2000 0.2000]
RowGroupMarker: []
ColumnGroupMarker: []
ShowDendrogram: 'on'
Standardize: 'NONE'
Symmetric: 1
DisplayRange: 3

4-188
 Working with the Clustergram Function

Colormap: [11x3 double]

ImputeFun: {@knnimpute}
ColumnLabels: {1x78 cell}
RowLabels: {70x1 cell}
ColumnLabelsRotate: 90
RowLabelsRotate: 0
Annotate: 'off'
AnnotPrecision: 2
AnnotColor: 'w'
ColumnLabelsColor: []
RowLabelsColor: []
LabelsWithMarkers: 0

Clustering the Rows and the Columns of a Data Set

Next, you will cluster both the rows and the columns of the data to produce a heat map with two
dendrograms. In this example, the left dendrogram shows the clustering of the genes (rows), and the
top dendrogram shows the clustering of the samples (columns).

cg = clustergram(progValues, 'RowLabels', progAccession,...

'ColumnLabels', progSamples,...
'RowPdist', 'correlation',...
'ColumnPdist', 'correlation',...
'ImputeFun', @knnimpute)

Clustergram object with 70 rows of nodes and 78 columns of nodes.

4-189
4 Microarray Analysis

You can also change the dendrogram option to differentiate clusters of genes and clusters of samples
with distances 1 unit apart.

cg.Dendrogram = 1;

4-190
 Working with the Clustergram Function

Interacting with the Heat Map

You can zoom in, zoom out and pan the heat map by selecting the corresponding toolbar buttons or
menu items.

4-191
4 Microarray Analysis

Click the Data Cursor button or select Tools > Data Cursor to activate Data Cursor Mode. In
this mode, click the heat map to display a data tip showing the expression value, the gene label and
the sample label of current data point. You can click-drag the data tip to other data points in the
heatmap. To delete the data tip, right-click, then select Delete Current Datatip from the context
menu.

4-192
 Working with the Clustergram Function

Click the Insert Colorbar button to show the color scale of the heat map.

4-193
4 Microarray Analysis

Interacting with the Dendrogram

To interact with dendrogram, be sure that the Data Cursor Mode is deactivated (click the Data
Cursor button again). Move the mouse over the dendrogram. When the mouse is over a branch node
a red marker appears and the branch is highlighted.

4-194
 Working with the Clustergram Function

Click and hold the red marker to display a data tip with the group number and the number of nodes
in the group. If the space is available, it also displays the labels for the nodes. For example, mouse
over and click on a dendrogram clustering group of the samples.

4-195
4 Microarray Analysis

Right-click the red marker to display a context menu. From the context menu you can change the
dendrogram color for the select group, print the group to a separate Figure window, copy the group
to a new Clustergram window, export it as a clustergram object to the MATLAB® Workspace, or
export the clustering group information as a structure to the MATLAB® Workspace.

4-196
 Working with the Clustergram Function

For example, select group 55 from the gene clustering dendrogram, and export it to the MATLAB®
Workspace by right-clicking then selecting Export Group to Workspace. You can view the
dendrograms and heat map for this clustergram object in a new Clustergram window by using the
view method.

4-197
4 Microarray Analysis

Changing the Color Scheme and Display Range

The default color scheme is the red-green color scale that is widely used in microarray data analysis.
In this example, a different color scheme may be more useful. The colormap option allows you to
specify an alternate colormap.

cg.Colormap = redbluecmap;
cg.DisplayRange = 2;

4-198
 Working with the Clustergram Function

Adding Color Markers

The clustergram function also lets you add color markers and text labels to annotate specific
regions of rows or columns. For example, to denote specific dendrogram groups of genes and groups
of samples, create structure arrays to specify the annotations for each dimension.

Create a structure array to annotate groups 34 and 50 in the gene dendrogram.

gene_markers = struct('GroupNumber', {34, 50},...
'Annotation', {'A', 'B'},...
'Color', {'b', 'm'});

Create a structure array to annotate groups 63 and 65 of the sample dendrogram.

sample_markers = struct('GroupNumber', {63, 65},...
'Annotation', {'Recurrences', 'Non-recurrences'},...
'Color', {[1 1 0], [0.6 0.6 1]});

Add the markers to the clustergram.

cg.RowGroupMarker = gene_markers;
cg.ColumnGroupMarker = sample_markers;

4-199
4 Microarray Analysis

Clustering 5000 Significant Genes

In this example, you will perform hierarchical clustering for almost 5,000 genes of the filtered data
[2].

cg_all = clustergram(bcTrainData.Log10Ratio,...
'RowLabels', bcTrainData.Accession,...
'ColumnLabels', bcTrainData.Samples,...
'RowPdist', 'correlation',...
'ColumnPdist', 'correlation',...
'Displayrange', 0.6,...
'Standardize', 3,...
'ImputeFun', @knnimpute)

Clustergram object with 4918 rows of nodes and 78 columns of nodes.

4-200
 Working with the Clustergram Function

Tip: When working with large data sets, MATLAB® can run out of memory during the clustering
computation. You can convert double precision data to single precision using the single function.
Note that the gene expression data in bcTrainData are already single precision.

You can resize a clustergram window like any other MATLAB® Figure window by click-dragging the
edge of the window.

4-201
4 Microarray Analysis

If you want even more control over the clustering, you can use the clustering functions in the
Statistics and Machine Learning Toolbox™ directly. See the “Gene Expression Profile Analysis” on
page 4-95 example for some examples of how to do this.

References

[1] Eisen, M. B., et al., "Cluster analysis and display of genome-wide expression patterns", PNAS,
95(25):14863-8, 1998.

[2] van't Veer, L., et al., "Gene expression profiling predicts clinical outcome of breast cancer",
Nature, 415(6871):530-6, 2002.

4-202
 Working with Objects for Microarray Experiment Data

Working with Objects for Microarray Experiment Data

This example shows how to create and manipulate MATLAB® containers designed for storing data
from a microarray experiment.

Containers for Gene Expression Experiment Data

Microarray experimental data are very complex, usually consisting of data and information from a
number of different sources. Storing and managing the large and complex data sets in a coherent
manner is a challenge. Bioinformatics Toolbox™ provides a set of objects to represent the different
pieces of data from a microarray experiment.

The ExpressionSet class is a single, convenient data structure for storing and managing different
types of data from a microarray gene expression experiment.

An ExpressionSet object consists of these four components that are common to all microarray gene
expression experiments:

Experiment data: Expression values from microarray experiments. These data are stored as an
instance of the ExptData class.

Sample information: The metadata describing the samples in the experiment. The sample metadata
are stored as an instance of the MetaData class.

Array feature annotations: The annotations about the features or probes on the array used in the
experiment. The annotations can be stored as an instance of the MetaData class.

Experiment descriptions: Information to describe the experiment methods and conditions. The
information can be stored as an instance of the MIAME class.

The ExpressionSet class coordinates and validates these data components. The class provides
methods for retrieving and setting the data stored in an ExpressionSet object. An ExpressionSet
object also behaves like many other MATLAB data structures that can be subsetted and copied.

Experiment Data

In a microarray gene expression experiment, the measured expression values for each feature per
sample can be represented as a two-dimensional matrix. The matrix has F rows and S columns, where
F is the number of features on the array, and S is the number of samples on which the expression
values were measured. A DataMatrix object is a two-dimensional matrix that you can index by row
and column numbers, logical vectors, or row and column names.

Create a Datamatrix with row and column names.

dm = bioma.data.DataMatrix(rand(5,4), 'RowNames','Feature', 'ColNames', 'Sample')

dm =

Sample1 Sample2 Sample3 Sample4

Feature1 0.81472 0.09754 0.15761 0.14189
Feature2 0.90579 0.2785 0.97059 0.42176
Feature3 0.12699 0.54688 0.95717 0.91574
Feature4 0.91338 0.95751 0.48538 0.79221

4-203
4 Microarray Analysis

Feature5 0.63236 0.96489 0.80028 0.95949

The function size returns the number of rows and columns in a DataMatrix object.
size(dm)

ans =

5 4

You can index into a DataMatrix object like other MATLAB numeric arrays by using row and column
numbers. For example, you can access the elements at rows 1 and 2, column 3.
dm(1:2, 3)

ans =

Sample3
Feature1 0.15761
Feature2 0.97059

You can also index into a DataMatrix object by using its row and column names. Reassign the
elements in row 2 and 3, column 1 and 4 to different values.
dm({'Feature2', 'Feature3'}, {'Sample1', 'Sample4'}) = [2, 3; 4, 5]

dm =

Sample1 Sample2 Sample3 Sample4

Feature1 0.81472 0.09754 0.15761 0.14189
Feature2 2 0.2785 0.97059 3
Feature3 4 0.54688 0.95717 5
Feature4 0.91338 0.95751 0.48538 0.79221
Feature5 0.63236 0.96489 0.80028 0.95949

The gene expression data used in this example is a small set of data from a microarray experiment
profiling adult mouse gene expression patterns in common strains on the Affymetrix® MG-U74Av2
array [1].

Read the expression values from the tab-formatted file mouseExprsData.txt into MATLAB
Workspace as a DataMatrix object.
exprsData = bioma.data.DataMatrix('file', 'mouseExprsData.txt');
class(exprsData)

ans =

'bioma.data.DataMatrix'

Get the properties of the DataMatrix object, exprsData.

4-204
 Working with Objects for Microarray Experiment Data

get(exprsData)

Name: 'mouseExprsData'
RowNames: {500x1 cell}
ColNames: {1x26 cell}
NRows: 500
NCols: 26
NDims: 2
ElementClass: 'double'

Check the sample names.

colnames(exprsData)

ans =

1x26 cell array

Columns 1 through 8

{'A'} {'B'} {'C'} {'D'} {'E'} {'F'} {'G'} {'H'}

Columns 9 through 16

{'I'} {'J'} {'K'} {'L'} {'M'} {'N'} {'O'} {'P'}

Columns 17 through 24

{'Q'} {'R'} {'S'} {'T'} {'U'} {'V'} {'W'} {'X'}

Columns 25 through 26

{'Y'} {'Z'}

View the first 10 rows and 5 columns.

exprsData(1:10, 1:5)

ans =

A B C D E
100001_at 2.26 20.14 31.66 14.58 16.04
100002_at 158.86 236.25 206.27 388.71 388.09
100003_at 68.11 105.45 82.92 82.9 60.38
100004_at 74.32 96.68 84.87 72.26 98.38
100005_at 75.05 53.17 57.94 60.06 63.91
100006_at 80.36 42.89 77.21 77.24 40.31
100007_at 216.64 191.32 219.48 237.28 298.18
100009_r_at 3806.7 1425 2468.5 2172.7 2237.2
100010_at NaN NaN NaN 7.18 22.37
100011_at 81.72 72.27 127.61 91.01 98.13

Perform a log2 transformation of the expression values.

4-205
4 Microarray Analysis

exprsData_log2 = log2(exprsData);
exprsData_log2(1:10, 1:5)

ans =

A B C D E
100001_at 1.1763 4.332 4.9846 3.8659 4.0036
100002_at 7.3116 7.8842 7.6884 8.6026 8.6002
100003_at 6.0898 6.7204 6.3736 6.3733 5.916
100004_at 6.2157 6.5951 6.4072 6.1751 6.6203
100005_at 6.2298 5.7325 5.8565 5.9083 5.998
100006_at 6.3284 5.4226 6.2707 6.2713 5.3331
100007_at 7.7592 7.5798 7.7779 7.8904 8.22
100009_r_at 11.894 10.477 11.269 11.085 11.127
100010_at NaN NaN NaN 2.844 4.4835
100011_at 6.3526 6.1753 6.9956 6.508 6.6166

Change the Name property to be more descriptive|.

exprsData_log2 = set(exprsData_log2, 'Name', 'Log2 Based mouseExprsData');

get(exprsData_log2)

Name: 'Log2 Based mouseExprsData'

RowNames: {500x1 cell}
ColNames: {1x26 cell}
NRows: 500
NCols: 26
NDims: 2
ElementClass: 'double'

In a microarray experiment, the data set often contains one or more matrices that have the same
number of rows and columns and identical row names and column names. ExptData class is
designed to contain and coordinate one or more data matrices having identical row and column
names with the same dimension size. The data values are stored as DataMatrix objects. Each
DataMatrix object is an element of an ExptData object. The ExptData class is responsible for data
validation and coordination between these DataMatrix objects.

Store the gene expression data of natural scale and log2 base expression values separately in an
ExptData object.

mouseExptData = bioma.data.ExptData(exprsData, exprsData_log2,...

'ElementNames', {'naturalExprs', 'log2Exprs'})

mouseExptData =

Experiment Data:
500 features, 26 samples
2 elements
Element names: naturalExprs, log2Exprs

Access a DataMatrix element in mouseExptData using the element name.

exprsData2 = mouseExptData('log2Exprs');
get(exprsData2)

4-206
 Working with Objects for Microarray Experiment Data

Name: 'Log2 Based mouseExprsData'

RowNames: {500x1 cell}
ColNames: {1x26 cell}
NRows: 500
NCols: 26
NDims: 2
ElementClass: 'double'

Sample Metadata

The metadata about the samples in a microarray experiment can be represented as a table with S
rows and V columns, where S is the number of samples, and V is the number of variables. The
contents of the table are the values of each variable for each sample. For example, the file
mouseSampleData.txt contains such a table. The description of each sample variable is marked by
a # symbol.

The MetaData class is designed for storing and manipulating variable values and their metadata in a
coordinated fashion. You can read the mouseSampleData.txt file into MATLAB as a MetaData
object.
sData = bioma.data.MetaData('file', 'mouseSampleData.txt', 'vardescchar', '#')

sData =

Sample Names:
A, B, ...,Z (26 total)
Variable Names and Meta Information:
VariableDescription
Gender {' Gender of the mouse in study' }
Age {' The number of weeks since mouse birth'}
Type {' Genetic characters' }
Strain {' The mouse strain' }
Source {' The tissue source for RNA collection' }

The properties of MetaData class provide information about the samples and variables.
numSamples = sData.NSamples
numVariables = sData.NVariables

numSamples =

26

numVariables =

The variable values and the variable descriptions for the samples are stored as two dataset arrays
in a MetaData class. The MetaData class provides access methods to the variable values and the
meta information describing the variables.

Access the sample metadata using the variableValues method.

4-207
4 Microarray Analysis

sData.variableValues

ans =

Gender Age Type Strain

A {'Male'} 8 {'Wild type'} {'129S6/SvEvTac'}
B {'Male'} 8 {'Wild type'} {'129S6/SvEvTac'}
C {'Male'} 8 {'Wild type'} {'129S6/SvEvTac'}
D {'Male'} 8 {'Wild type'} {'A/J ' }
E {'Male'} 8 {'Wild type'} {'A/J ' }
F {'Male'} 8 {'Wild type'} {'C57BL/6J ' }
G {'Male'} 8 {'Wild type'} {'C57BL/6J' }
H {'Male'} 8 {'Wild type'} {'129S6/SvEvTac'}
I {'Male'} 8 {'Wild type'} {'129S6/SvEvTac'}
J {'Male'} 8 {'Wild type'} {'A/J' }
K {'Male'} 8 {'Wild type'} {'A/J' }
L {'Male'} 8 {'Wild type'} {'A/J' }
M {'Male'} 8 {'Wild type'} {'C57BL/6J' }
N {'Male'} 8 {'Wild type'} {'C57BL/6J' }
O {'Male'} 8 {'Wild type'} {'129S6/SvEvTac'}
P {'Male'} 8 {'Wild type'} {'129S6/SvEvTac'}
Q {'Male'} 8 {'Wild type'} {'A/J' }
R {'Male'} 8 {'Wild type'} {'A/J' }
S {'Male'} 8 {'Wild type'} {'C57BL/6J' }
T {'Male'} 8 {'Wild type'} {'C57BL/6J4' }
U {'Male'} 8 {'Wild type'} {'129S6/SvEvTac'}
V {'Male'} 8 {'Wild type'} {'129S6/SvEvTac'}
W {'Male'} 8 {'Wild type'} {'A/J' }
X {'Male'} 8 {'Wild type'} {'A/J' }
Y {'Male'} 8 {'Wild type'} {'C57BL/6J' }
Z {'Male'} 8 {'Wild type'} {'C57BL/6J' }

Source
A {'amygdala' }
B {'amygdala' }
C {'amygdala' }
D {'amygdala' }
E {'amygdala' }
F {'amygdala' }
G {'amygdala' }
H {'cingulate cortex'}
I {'cingulate cortex'}
J {'cingulate cortex'}
K {'cingulate cortex'}
L {'cingulate cortex'}
M {'cingulate cortex'}
N {'cingulate cortex'}
O {'hippocampus' }
P {'hippocampus' }
Q {'hippocampus' }
R {'hippocampus' }
S {'hippocampus' }
T {'hippocampus' }
U {'hypothalamus' }
V {'hypothalamus' }
W {'hypothalamus' }

4-208
 Working with Objects for Microarray Experiment Data

X {'hypothalamus' }
Y {'hypothalamus' }
Z {'hypothalamus' }

View a summary of the sample metadata.

summary(sData.variableValues)

Gender: [26x1 cell array of character vectors]

Age: [26x1 double]

min 1st quartile median 3rd quartile max

8 8 8 8 8

Type: [26x1 cell array of character vectors]

Strain: [26x1 cell array of character vectors]

Source: [26x1 cell array of character vectors]

The sampleNames and variableNames methods are convenient ways to access the names of
samples and variables. Retrieve the variable names of the sData object.

variableNames(sData)

ans =

1x5 cell array

{'Gender'} {'Age'} {'Type'} {'Strain'} {'Source'}

You can retrieve the meta information about the variables describing the samples using the
variableDesc method. In this example, it contains only the descriptions about the variables.

variableDesc(sData)

ans =

VariableDescription
Gender {' Gender of the mouse in study' }
Age {' The number of weeks since mouse birth'}
Type {' Genetic characters' }
Strain {' The mouse strain' }
Source {' The tissue source for RNA collection' }

You can subset the sample data sData object using numerical indexing.

sData(3:6, :)

ans =

4-209
4 Microarray Analysis

Sample Names:
C, D, ...,F (4 total)
Variable Names and Meta Information:
VariableDescription
Gender {' Gender of the mouse in study' }
Age {' The number of weeks since mouse birth'}
Type {' Genetic characters' }
Strain {' The mouse strain' }
Source {' The tissue source for RNA collection' }

You can display the mouse strain of specific samples by using numerical indexing.
sData.Strain([2 14])

ans =

2x1 cell array

{'129S6/SvEvTac'}
{'C57BL/6J' }

Note that the row names in sData and the column names in exprsData are the same. It is an
important relationship between the expression data and the sample data in the same experiment.
all(ismember(sampleNames(sData), colnames(exprsData)))

ans =

logical

Feature Annotation Metadata

The metadata about the features or probe set on an array can be very large and diverse. The chip
manufacturers usually provide a specific annotation file for the features of each type of array. The
metadata can be stored as a MetaData object for a specific experiment. In this example, the
annotation file for the MG-U74Av2 array can be downloaded from the Affymetrix web site. You will
need to convert the file from CSV to XLSX format using a spreadsheet software application.

Read the entire file into MATLAB as a dataset array. Alternatively, you can use the Range option in
the dataset constructor. Any blank spaces in the variable names are removed to make them valid
MATLAB variable names. A warning is displayed each time this happens.
mgU74Av2 = table2dataset(readtable('MG_U74Av2_annot.xlsx'));

Warning: Column headers from the file were modified to make them valid MATLAB
identifiers before creating variable names for the table. The original column
headers are saved in the VariableDescriptions property.
Set 'VariableNamingRule' to 'preserve' to use the original column headers as
table variable names.

Inspect the properties of this dataset array.

4-210
 Working with Objects for Microarray Experiment Data

get(mgU74Av2)

Description: ''
VarDescription: {1x43 cell}
Units: {}
DimNames: {'Row' 'Variables'}
UserData: []
ObsNames: {}
VarNames: {1x43 cell}

Determine the number of probe set IDs in the annotation file.

numel(mgU74Av2.ProbeSetID)

ans =

12488

Retrieve the names of variables describing the features on the array and view the first 20 variable
names.
fDataVariables = get(mgU74Av2, 'VarNames');
fDataVariables(1:20)'

ans =

20x1 cell array

{'ProbeSetID' }
{'GeneChipArray' }
{'SpeciesScientificName' }
{'AnnotationDate' }
{'SequenceType' }
{'SequenceSource' }
{'TranscriptID_ArrayDesign_'}
{'TargetDescription' }
{'RepresentativePublicID' }
{'ArchivalUniGeneCluster' }
{'UniGeneID' }
{'GenomeVersion' }
{'Alignments' }
{'GeneTitle' }
{'GeneSymbol' }
{'ChromosomalLocation' }
{'UnigeneClusterType' }
{'Ensembl' }
{'EntrezGene' }
{'SwissProt' }

Set the ObsNames property to the probe set IDs, so that you can access individual gene annotations
by indexing with probe set IDs.
mgU74Av2 = set(mgU74Av2,'ObsNames',mgU74Av2.ProbeSetID);
mgU74Av2('100709_at',{'GeneSymbol','ChromosomalLocation'})

4-211
4 Microarray Analysis

ans =

GeneSymbol ChromosomalLocation
100709_at {'Tpbpa'} {'chr13 B2|13 36.0 cM'}

In some cases, it is useful to extract specific annotations that are relevant to the analysis. Extract
annotations for GeneTitle, GeneSymbol, ChromosomalLocation, and Pathway relative to the
features in exprsData.
mgU74Av2 = mgU74Av2(:,{'GeneTitle',...
'GeneSymbol',...
'ChromosomalLocation',...
'Pathway'});

mgU74Av2 = mgU74Av2(rownames(exprsData),:);
get(mgU74Av2)

Description: ''
VarDescription: {1x4 cell}
Units: {}
DimNames: {'Row' 'Variables'}
UserData: []
ObsNames: {500x1 cell}
VarNames: {1x4 cell}

You can store the feature annotation dataset array as an instance of the MetaData class.
fData = bioma.data.MetaData(mgU74Av2)

fData =

Sample Names:
100001_at, 100002_at, ...,100717_at (500 total)
Variable Names and Meta Information:
VariableDescription
GeneTitle {'NA'}
GeneSymbol {'NA'}
ChromosomalLocation {'NA'}
Pathway {'NA'}

Notice that there are no descriptions for the feature variables in the fData MetaData object. You
can add descriptions about the variables in fData using the variableDesc method.
fData = variableDesc(fData, {'Gene title of a probe set',...
'Probe set gene symbol',...
'Probe set chromosomal locations',...
'The pathway the genes involved in'})

fData =

Sample Names:
100001_at, 100002_at, ...,100717_at (500 total)
Variable Names and Meta Information:

4-212
 Working with Objects for Microarray Experiment Data

VariableDescription
GeneTitle {'Gene title of a probe set' }
GeneSymbol {'Probe set gene symbol' }
ChromosomalLocation {'Probe set chromosomal locations' }
Pathway {'The pathway the genes involved in'}

Experiment Information

The MIAME class is a flexible data container designed for a collection of basic descriptions about a
microarray experiment, such as investigators, laboratories, and array designs. The MIAME class
loosely follows the Minimum Information About a Microarray Experiment (MIAME) specification [2].

Create a MIAME object by providing some basic information.

expDesc = bioma.data.MIAME('investigator', 'Jane OneName',...
'lab', 'Bioinformatics Laboratory',...
'title', 'Example Gene Expression Experiment',...
'abstract', 'An example of using microarray objects.',...
'other', {'Notes: Created from a text files.'})

expDesc =

Experiment Description:
Author name: Jane OneName
Laboratory: Bioinformatics Laboratory
Contact information:
URL:
PubMedIDs:
Abstract: A 5 word abstract is available. Use the Abstract property.
No experiment design summary available.
Other notes:
{'Notes: Created from a text files.'}

Another way to create a MIAME object is from GEO series data. The MIAME class will populate the
corresponding properties from the GEO series structure. The information associated with the gene
profile experiment in this example is available from the GEO database under the accession number
GSE3327 [1]. Retrieve the GEO Series data using the getgeodata function.
getgeodata('GSE3327', 'ToFile', 'GSE3327.txt');

Read the data into a structure.

geoSeries = geoseriesread('GSE3327.txt')

geoSeries =

struct with fields:

Header: [1x1 struct]

Data: [12488x87 bioma.data.DataMatrix]

Create a MIAME object.

exptGSE3327 = bioma.data.MIAME(geoSeries)

4-213
4 Microarray Analysis

exptGSE3327 =

Experiment Description:
Author name: Iiris,,Hovatta
David,J,Lockhart
Carrolee,,Barlow
Laboratory: The Salk Institute for Biological Studies
Contact information: Carrolee,,Barlow
URL:
PubMedIDs: 16244648
Abstract: A 14 word abstract is available. Use the Abstract property.
Experiment Design: A 8 word summary is available. Use the ExptDesign property.
Other notes:
{'ftp://ftp.ncbi.nlm.nih.gov/pub/geo/DATA/supplementary/series/GSE3327/GSE3327_RAW.tar'}

View the abstract of the experiment and its PubMed IDs.

abstract = exptGSE3327.Abstract
pubmedID = exptGSE3327.PubMedID

abstract =

'Adult mouse gene expression patterns in common strains

Keywords: mouse strain and brain region comparison'

pubmedID =

'16244648'

Creating an ExpressionSet Object

The ExpressionSet class is designed specifically for microarray gene expression experiment data.
Assemble an ExpressionSet object for the example mouse gene expression experiment from the
different data objects you just created.

exptSet = bioma.ExpressionSet(exprsData, 'SData', sData,...

'FData', fData,...
'Einfo', exptGSE3327)

exptSet =

ExpressionSet
Experiment Data: 500 features, 26 samples
Element names: Expressions
Sample Data:
Sample names: A, B, ...,Z (26 total)
Sample variable names and meta information:
Gender: Gender of the mouse in study
Age: The number of weeks since mouse birth
Type: Genetic characters
Strain: The mouse strain
Source: The tissue source for RNA collection

4-214
 Working with Objects for Microarray Experiment Data

Feature Data:
Feature names: 100001_at, 100002_at, ...,100717_at (500 total)
Feature variable names and meta information:
GeneTitle: Gene title of a probe set
GeneSymbol: Probe set gene symbol
ChromosomalLocation: Probe set chromosomal locations
Pathway: The pathway the genes involved in
Experiment Information: use 'exptInfo(obj)'

You can also create an ExpressionSet object with only the expression values in a DataMatrix or a
numeric matrix.
miniExprSet = bioma.ExpressionSet(exprsData)

miniExprSet =

ExpressionSet
Experiment Data: 500 features, 26 samples
Element names: Expressions
Sample Data: none
Feature Data: none
Experiment Information: none

Saving and Loading an ExpressionSet Object

The data objects for a microarray experiment can be saved as MAT files. Save the ExpressionSet
object exptSet to a MAT file named mouseExpressionSet.mat.
save mouseExpressionSet exptSet

Clear variables from the MATLAB Workspace.

clear dm exprs* mouseExptData ME sData

Load the MAT file mouseExpressionSet into the MATLAB Workspace.

load mouseExpressionSet

Inspect the loaded ExpressionSet object.

exptSet.elementNames

ans =

1x1 cell array

{'Expressions'}

exptSet.NSamples

ans =

26

exptSet.NFeatures

4-215
4 Microarray Analysis

ans =

500

Accessing Data Components of an ExpressionSet Object

A number of methods are available to access and update data stored in an ExpressionSet object.

You can access the columns of the sample data using dot notation.

exptSet.Strain(1:5)

ans =

5x1 cell array

{'129S6/SvEvTac'}
{'129S6/SvEvTac'}
{'129S6/SvEvTac'}
{'A/J ' }
{'A/J ' }

Retrieve the feature names using the featureNames method. In this example, the feature names are
the probe set identifiers on the array.

featureNames(exptSet, 1:5)

ans =

5x1 cell array

{'100001_at'}
{'100002_at'}
{'100003_at'}
{'100004_at'}
{'100005_at'}

The unique identifier of the samples can be accessed via the sampleNames method.

exptSet.sampleNames(1:5)

ans =

1x5 cell array

{'A'} {'B'} {'C'} {'D'} {'E'}

The sampleVarNames method lists the variable names in the sample data.

exptSet.sampleVarNames

4-216
 Working with Objects for Microarray Experiment Data

ans =

1x5 cell array

{'Gender'} {'Age'} {'Type'} {'Strain'} {'Source'}

Extract the dataset array containing sample information.

sDataset = sampleVarValues(exptSet)

sDataset =

Gender Age Type Strain

A {'Male'} 8 {'Wild type'} {'129S6/SvEvTac'}
B {'Male'} 8 {'Wild type'} {'129S6/SvEvTac'}
C {'Male'} 8 {'Wild type'} {'129S6/SvEvTac'}
D {'Male'} 8 {'Wild type'} {'A/J ' }
E {'Male'} 8 {'Wild type'} {'A/J ' }
F {'Male'} 8 {'Wild type'} {'C57BL/6J ' }
G {'Male'} 8 {'Wild type'} {'C57BL/6J' }
H {'Male'} 8 {'Wild type'} {'129S6/SvEvTac'}
I {'Male'} 8 {'Wild type'} {'129S6/SvEvTac'}
J {'Male'} 8 {'Wild type'} {'A/J' }
K {'Male'} 8 {'Wild type'} {'A/J' }
L {'Male'} 8 {'Wild type'} {'A/J' }
M {'Male'} 8 {'Wild type'} {'C57BL/6J' }
N {'Male'} 8 {'Wild type'} {'C57BL/6J' }
O {'Male'} 8 {'Wild type'} {'129S6/SvEvTac'}
P {'Male'} 8 {'Wild type'} {'129S6/SvEvTac'}
Q {'Male'} 8 {'Wild type'} {'A/J' }
R {'Male'} 8 {'Wild type'} {'A/J' }
S {'Male'} 8 {'Wild type'} {'C57BL/6J' }
T {'Male'} 8 {'Wild type'} {'C57BL/6J4' }
U {'Male'} 8 {'Wild type'} {'129S6/SvEvTac'}
V {'Male'} 8 {'Wild type'} {'129S6/SvEvTac'}
W {'Male'} 8 {'Wild type'} {'A/J' }
X {'Male'} 8 {'Wild type'} {'A/J' }
Y {'Male'} 8 {'Wild type'} {'C57BL/6J' }
Z {'Male'} 8 {'Wild type'} {'C57BL/6J' }

Source
A {'amygdala' }
B {'amygdala' }
C {'amygdala' }
D {'amygdala' }
E {'amygdala' }
F {'amygdala' }
G {'amygdala' }
H {'cingulate cortex'}
I {'cingulate cortex'}
J {'cingulate cortex'}
K {'cingulate cortex'}
L {'cingulate cortex'}
M {'cingulate cortex'}

4-217
4 Microarray Analysis

N {'cingulate cortex'}
O {'hippocampus' }
P {'hippocampus' }
Q {'hippocampus' }
R {'hippocampus' }
S {'hippocampus' }
T {'hippocampus' }
U {'hypothalamus' }
V {'hypothalamus' }
W {'hypothalamus' }
X {'hypothalamus' }
Y {'hypothalamus' }
Z {'hypothalamus' }

Retrieve the ExptData object containing expression values. There may be more than one
DataMatrix object with identical dimensions in an ExptData object. In an ExpressionSet object,
there is always a element DataMatrix object named Expressions containing the expression
matrix.

exptDS = exptData(exptSet)

exptDS =

Experiment Data:
500 features, 26 samples
1 elements
Element names: Expressions

Extract only the expression DataMatrix instance.

dMatrix = expressions(exptSet);

The returned expression DataMatrix should be identical to the exprsData DataMatrix object that
you created earlier.

get(dMatrix)

Name: 'mouseExprsData'
RowNames: {500x1 cell}
ColNames: {1x26 cell}
NRows: 500
NCols: 26
NDims: 2
ElementClass: 'double'

Get PubMed IDs for the experiment stored in exptSet.

exptSet.pubMedID

ans =

'16244648'

4-218
 Working with Objects for Microarray Experiment Data

Subsetting an ExpressionSet Object

You can subset an ExpressionSet object so that you can focus on the samples and features of
interest. The first indexing argument subsets the features and the second argument subsets the
samples.

Create a new ExpressionSet object consisting of the first five features and the samples named A, B,
and C.
mySet = exptSet(1:5, {'A', 'B', 'C'})

mySet =

ExpressionSet
Experiment Data: 5 features, 3 samples
Element names: Expressions
Sample Data:
Sample names: A, B, C
Sample variable names and meta information:
Gender: Gender of the mouse in study
Age: The number of weeks since mouse birth
Type: Genetic characters
Strain: The mouse strain
Source: The tissue source for RNA collection
Feature Data:
Feature names: 100001_at, 100002_at, ...,100005_at (5 total)
Feature variable names and meta information:
GeneTitle: Gene title of a probe set
GeneSymbol: Probe set gene symbol
ChromosomalLocation: Probe set chromosomal locations
Pathway: The pathway the genes involved in
Experiment Information: use 'exptInfo(obj)'

size(mySet)

ans =

5 3

featureNames(mySet)

ans =

5x1 cell array

{'100001_at'}
{'100002_at'}
{'100003_at'}
{'100004_at'}
{'100005_at'}

sampleNames(mySet)

ans =

4-219
4 Microarray Analysis

1x3 cell array

{'A'} {'B'} {'C'}

You can also create a subset consisting of only the samples from hippocampus tissues.
hippocampusSet = exptSet(:, nominal(exptSet.Source)== 'hippocampus')

hippocampusSet =

ExpressionSet
Experiment Data: 500 features, 6 samples
Element names: Expressions
Sample Data:
Sample names: O, P, ...,T (6 total)
Sample variable names and meta information:
Gender: Gender of the mouse in study
Age: The number of weeks since mouse birth
Type: Genetic characters
Strain: The mouse strain
Source: The tissue source for RNA collection
Feature Data:
Feature names: 100001_at, 100002_at, ...,100717_at (500 total)
Feature variable names and meta information:
GeneTitle: Gene title of a probe set
GeneSymbol: Probe set gene symbol
ChromosomalLocation: Probe set chromosomal locations
Pathway: The pathway the genes involved in
Experiment Information: use 'exptInfo(obj)'

hippocampusSet.Source

ans =

6x1 cell array

{'hippocampus'}
{'hippocampus'}
{'hippocampus'}
{'hippocampus'}
{'hippocampus'}
{'hippocampus'}

hippocampusExprs = expressions(hippocampusSet);

get(hippocampusExprs)

Name: 'mouseExprsData'
RowNames: {500x1 cell}
ColNames: {'O' 'P' 'Q' 'R' 'S' 'T'}
NRows: 500
NCols: 6
NDims: 2
ElementClass: 'double'

4-220
 Working with Objects for Microarray Experiment Data

References

[1] Hovatta, I., et al., "Glyoxalase 1 and glutathione reductase 1 regulate anxiety in mice", Nature,
438(7068):662-6, 2005.

[2] Brazma, A., et al., "Minimum information about a microarray experiment (MIAME) - toward
standards for microarray data", Nat. Genet. 29(4):365-371, 2001.

4-221
 5

Phylogenetic Analysis

• “Using the Phylogenetic Tree App” on page 5-2

• “Building a Phylogenetic Tree for the Hominidae Species” on page 5-19
• “Analyzing the Origin of the Human Immunodeficiency Virus” on page 5-25
• “Bootstrapping Phylogenetic Trees” on page 5-32
5 Phylogenetic Analysis

Using the Phylogenetic Tree App

In this section...
“Overview of the Phylogenetic Tree App” on page 5-2
“Opening the Phylogenetic Tree App” on page 5-2
“File Menu” on page 5-3
“Tools Menu” on page 5-11
“Window Menu” on page 5-17
“Help Menu” on page 5-18

Overview of the Phylogenetic Tree App

The Phylogenetic Tree app allows you to view, edit, format, and explore phylogenetic tree data. With
this app you can prune, reorder, rename branches, and explore distances. You can also open or save
Newick or ClustalW tree formatted files. The following sections give a description of menu commands
and features for creating publishable tree figures.

Opening the Phylogenetic Tree App

This section illustrates how to draw a phylogenetic tree from data in a phytree object or a
previously saved file.

The Phylogenetic Tree app can read data from Newick and ClustalW tree formatted files.

This procedure uses the phylogenetic tree data stored in the file pf00002.tree as an example. The
data was retrieved from the protein family (PFAM) Web database and saved to a file using the
accession number PF00002 and the function gethmmtree.

1 Create a phytree object. For example, to create a phytree object from tree data in the file
pf00002.tree, type

tr = phytreeread('pf00002.tree')

The MATLAB software creates a phytree object.

Phylogenetic tree object with 33 leaves (32 branches)

2 View the phylogenetic tree using the app.

phytreeviewer(tr)

Alternatively, click Phylogenetic Tree on the Apps tab.

5-2
 Using the Phylogenetic Tree App

File Menu
The File menu includes the standard commands for opening and closing a file, and it includes
commands to use phytree object data from the MATLAB Workspace. The File menu commands are
shown below.

5-3
5 Phylogenetic Analysis

New Viewer Command

Use the New Viewer command to open tree data from a file into a second Phylogenetic Tree viewer.

1 From the File menu, select New Viewer.

The Open A Phylogenetic Tree dialog box opens.

2 Choose the source for a tree.

5-4
 Using the Phylogenetic Tree App

• MATLAB Workspace — Select the Import from Workspace options, and then select a
phytree object from the list.
• File — Select the Open phylogenetic tree file option, click the Browse button, select a
directory, select a file with the extension .tree, and then click Open. The toolbox uses the
file extension .tree for Newick-formatted files, but you can use any Newick-formatted file
with any extension.

A second Phylogenetic Tree viewer opens with tree data from the selected file.

Open Command

Use the Open command to read tree data from a Newick-formatted file and display that data in the
app.

1 From the File menu, click Open.

The Select Phylogenetic Tree File dialog box opens.

2 Select a directory, select a Newick-formatted file, and then click Open. The app uses the file
extension .tree for Newick-formatted files, but you can use any Newick-formatted file with any
extension.

The app replaces the current tree data with data from the selected file.

Import from Workspace Command

Use the Import from Workspace command to read tree data from a phytree object in the MATLAB
Workspace and display the data using the app.

1 From the File menu, select Import from Workspace.

The Get Phytree Object dialog box opens.

5-5
5 Phylogenetic Analysis

2 From the list, select a phytree object in the MATLAB Workspace.

3 Click the Import button.

The app replaces the current tree data with data from the selected object.

Open Original in New Viewer

There may be times when you make changes that you would like to undo. The Phylogenetic Tree
app does not have an undo command, but you can get back to the original tree you started viewing
with the Open Original in New Viewer command.

From the File menu, select Open Original in New Viewer.

A new Phylogenetic Tree viewer opens with the original tree.

Save As Command

After you create a phytree object or prune a tree from existing data, you can save the resulting tree
in a Newick-formatted file. The sequence data used to create the phytree object is not saved with
the tree.

1 From the File menu, select Save As.

The Save Phylogenetic tree as dialog box opens.

2 In the Filename box, enter the name of a file. The toolbox uses the file extension .tree for
Newick-formatted files, but you can use any file extension.
3 Click Save.

The app saves tree data without the deleted branches, and it saves changes to branch and leaf
names. Formatting changes such as branch rotations, collapsed branches, and zoom settings are
not saved in the file.

Export to New Viewer Command

Because some of the Phylogenetic Tree viewer commands cannot be undone (for example, the Prune
command), you might want to make a copy of your tree before trying a command. At other times, you

5-6
 Using the Phylogenetic Tree App

might want to compare two views of the same tree, and copying a tree to a new tool window allows
you to make changes to both tree views independently .

1 Select File > Export to New Viewer, and then select either With Hidden Nodes or Only
Displayed.

A new Phylogenetic Tree viewer opens with a copy of the tree.

2 Use the new figure to continue your analysis.

Export to Workspace Command

The Phylogenetic Tree app can open Newick-formatted files with tree data. However, it does not
create a phytree object in the MATLAB Workspace. If you want to programmatically explore
phylogenetic trees, you need to use the Export to Workspace command.

1 Select File > Export to Workspace, and then select either With Hidden Nodes or Only
Displayed.

The Export to Workspace dialog box opens.

2 In the Workspace variable name box, enter the name for your phylogenetic tree data. For
example, enter MyTree.

3 Click OK.

The app creates a phytree object in the MATLAB Workspace.

Print to Figure Command

After you have explored the relationships between branches and leaves in your tree, you can copy the
tree to a MATLAB Figure window. Using a Figure window lets you use all the features for annotating,
changing font characteristics, and getting your figure ready for publication. Also, from the Figure
window, you can save an image of the tree as it was displayed in the Phylogenetic Tree app.

1 From the File menu, select Print to Figure, and then select either With Hidden Nodes or
Only Displayed.

The Print Phylogenetic Tree to Figure dialog box opens.

5-7
5 Phylogenetic Analysis

2 Select one of the Rendering Types.

Rendering Type Description

'square' (default)

5-8
 Using the Phylogenetic Tree App

Rendering Type Description

'angular'

'radial'

5-9
5 Phylogenetic Analysis

Rendering Type Description

'equalangle'

Tip This rendering type hides the significance of the root node
and emphasizes clusters, thereby making it useful for visually
assessing clusters and detecting outliers.
'equaldaylight'

Tip This rendering type hides the significance of the root node
and emphasizes clusters, thereby making it useful for visually
assessing clusters and detecting outliers.
3 Select the Display Labels you want on your figure. You can select from all to none of the
options.

• Branch Nodes — Display branch node names on the figure.

• Leaf Nodes — Display leaf node names on the figure.
• Terminal Nodes — Display terminal node names on the right border.
4 Click the Print button.

A new Figure window opens with the characteristics you selected.

Print Preview Command

When you print from the Phylogenetic Tree app or a MATLAB Figure window (with a tree published
from the viewer), you can specify setup options for printing a tree.

5-10
 Using the Phylogenetic Tree App

1 From the File menu, select Print Preview.

The Print Preview window opens, which you can use to select page formatting options.

2 Select the page formatting options and values you want, and then click Print.

Print Command

Use the Print command to make a copy of your phylogenetic tree after you use the Print Preview
command to select formatting options.

1 From the File menu, select Print.

The Print dialog box opens.

2 From the Name list, select a printer, and then click OK.

Tools Menu
Use the Tools menu to:

• Explore branch paths

• Rotate branches
• Find, rename, hide, and prune branches and leaves.

The Tools menu and toolbar contain most of the commands specific to trees and phylogenetic
analysis. Use these commands and modes to edit and format your tree interactively. The Tools menu
commands are:

5-11
5 Phylogenetic Analysis

Inspect Mode

Viewing a phylogenetic tree in the Phylogenetic Tree app provides a rough idea of how closely
related two sequences are. However, to see exactly how closely related two sequences are, measure
the distance of the path between them. Use the Inspect command to display and measure the path
between two sequences.

1
Select Tools > Inspect, or from the toolbar, click the Inspect Tool Mode icon .

The app is set to inspect mode.

2 Click a branch or leaf node (selected node), and then hover your cursor over another branch or
leaf node (current node).

The app highlights the path between the two nodes and displays the path length in the pop-up
window. The path length is the patristic distance calculated by the seqpdist function.

5-12
 Using the Phylogenetic Tree App

Collapse and Expand Branch Mode

Some trees have thousands of leaf and branch nodes. Displaying all the nodes can create an
unreadable tree diagram. By collapsing some branches, you can better see the relationships between
the remaining nodes.
1 Select Tools > Collapse/Expand, or from the toolbar, click the Collapse/Expand Brand Mode

icon .

The app is set to collapse/expand mode.

2 Point to a branch.

The paths, branch nodes, and leaf nodes below the selected branch appear in gray, indicating you
selected them to collapse (hide from view).

3 Click the branch node.

The app hides the display of paths, branch nodes, and leaf nodes below the selected branch.
However, it does not remove the data.

4 To expand a collapsed branch, click it or select Tools > Reset View.

Tip After collapsing nodes, you can redraw the tree by selecting Tools > Fit to Window.

Rotate Branch Mode

A phylogenetic tree is initially created by pairing the two most similar sequences and then adding the
remaining sequences in a decreasing order of similarity. You can rotate branches to emphasize the
direction of evolution.
1
Select Tools > Rotate Branch, or from the toolbar, click the Rotate Branch Mode icon .

The app is set to rotate branch mode.

2 Point to a branch node.

5-13
5 Phylogenetic Analysis

3 Click the branch node.

The branch and leaf nodes below the selected branch node rotate 180 degrees around the branch
node.
4 To undo the rotation, simply click the branch node again.

Rename Leaf or Branch Mode

The Phylogenetic Tree app takes the node names from a phytree object and creates numbered
branch names starting with Branch 1. You can edit any of the leaf or branch names.

1
Select Tools > Rename, or from the toolbar, click the Rename Leaf/Branch Mode icon .

The app is set to rename mode.

2 Click a branch or leaf node.

A text box opens with the current name of the node.

3 In the text box, edit or enter a new name.

4 To accept your changes and close the text box, click outside of the text box. To save your
changes, select File > Save As.

Prune (Delete) Leaf or Branch Mode

Your tree can contain leaves that are far outside the phylogeny, or it can have duplicate leaves that
you want to remove.

1 Select Tools > Prune, or from the toolbar, click the Prune (delete) Leaf/Branch Mode icon

The app is set to prune mode.

2 Point to a branch or leaf node.

5-14
 Using the Phylogenetic Tree App

For a leaf node, the branch line connected to the leaf appears in gray. For a branch node, the
branch lines below the node appear in gray.

Note If you delete nodes (branches or leaves), you cannot undo the changes. The Phylogenetic
Tree app does not have an Undo command.
3 Click the branch or leaf node.

The tool removes the branch from the figure and rearranges the other nodes to balance the tree
structure. It does not recalculate the phylogeny.

Tip After pruning nodes, you can redraw the tree by selecting Tools > Fit to Window.

Zoom In, Zoom Out, and Pan Commands

The Zoom and Pan commands are the standard controls for resizing and moving the screen in any
MATLAB Figure window.

1
Select Tools > Zoom In, or from the toolbar, click the Zoom In icon .

The app activates zoom in mode and changes the cursor to a magnifying glass.

2 Place the cursor over the section of the tree diagram you want to enlarge and then click.

The tree diagram doubles its size.

3
From the toolbar click the Pan icon .

5-15
5 Phylogenetic Analysis

4 Move the cursor over the tree diagram, left-click, and drag the diagram to the location you want
to view.

Tip After zooming and panning, you can reset the tree to its original view, by selecting Tools >
Reset View.

Select Submenu

Select a single branch or leaf node by clicking it. Select multiple branch or leaf nodes by Shift-
clicking the nodes, or click-dragging to draw a box around nodes.

Use the Select submenu to select specific branch and leaf nodes based on different criteria.

• Select By Distance — Displays a slider bar at the top of the window, which you slide to specify a
distance threshold. Nodes whose distance from the selected node are below this threshold appear
in red. Nodes whose distance from the selected node are above this threshold appear in blue.
• Select Common Ancestor — For all selected nodes, highlights the closest common ancestor
branch node in red.
• Select Leaves — If one or more nodes are selected, highlights the nodes that are leaf nodes in
red. If no nodes are selected, highlights all leaf nodes in red
• Propagate Selection — For all selected nodes, highlights the descendant nodes in red.
• Swap Selection — Clears all selected nodes and selects all deselected nodes.

After selecting nodes using one of the previous commands, hide and show the nodes using the
following commands:

• Collapse Selected
• Expand Selected
• Expand All

Clear all selected nodes by clicking anywhere else in the Phylogenetic Tree app.

Find Leaf or Branch Command

Phylogenetic trees can have thousands of leaves and branches, and finding a specific node can be
difficult. Use the Find Leaf/Branch command to locate a node using its name or part of its name.

1 Select Tools > Find Leaf/Branch.

The Find Leaf/Branch dialog box opens.

5-16
 Using the Phylogenetic Tree App

2 In the Regular Expression to match box, enter a name or partial name of a branch or leaf
node.
3 Click OK.

The branch or leaf nodes that match the expression appear in red.

After selecting nodes using the Find Leaf/Branch command, you can hide and show the nodes using
the following commands:

• Collapse Selected
• Expand Selected
• Expand All

Collapse Selected, Expand Selected, and Expand All Commands

When you select nodes, either manually or using the previous commands, you can then collapse them
by selecting Tools > Collapse Selected.

The data for branches and leaves that you hide using the Collapse/Expand or Collapse Selected
command are not removed from the tree. You can display selected or all hidden data using the
Expand Selected or Expand All command.

Fit to Window Command

After you hide nodes with the collapse commands, or delete nodes with the Prune command, there
can be extra space in the tree diagram. Use the Fit to Window command to redraw the tree diagram
to fill the entire Figure window.

Select Tools > Fit to Window.

Reset View Command

Use the Reset View command to remove formatting changes such as collapsed branches and zooms.

Select Tools > Reset View.

Options Submenu

Use the Options command to select the behavior for the zoom and pan modes.

• Unconstrained Zoom — Allow zooming in both horizontal and vertical directions.

• Horizontal Zoom — Restrict zooming to the horizontal direction.
• Vertical Zoom (default) — Restrict zooming to the vertical direction.
• Unconstrained Pan — Allow panning in both horizontal and vertical directions.
• Horizontal Pan — Restrict panning to the horizontal direction.
• Vertical Pan (default) — Restrict panning to the vertical direction.

Window Menu
This section illustrates how to switch to any open window.

5-17
5 Phylogenetic Analysis

The Window menu is standard on MATLAB interfaces and Figure windows. Use this menu to select
any opened window.

Help Menu
This section illustrates how to select quick links to the Bioinformatics Toolbox documentation for
phylogenetic analysis functions, tutorials, and the Phylogenetic Tree app reference.

Use the Help menu to select quick links to the Bioinformatics Toolbox documentation for
phylogenetic analysis functions, tutorials, and the phytreeviewer reference.

5-18
 Building a Phylogenetic Tree for the Hominidae Species

Building a Phylogenetic Tree for the Hominidae Species

This example shows how to construct phylogenetic trees from mtDNA sequences for the Hominidae
taxa (also known as pongidae). This family embraces the gorillas, chimpanzees, orangutans and
humans.

Introduction

The mitochondrial D-loop is one of the fastest mutating sequence regions in animal DNA, and
therefore, is often used to compare closely related organisms. The origin of modern man is a highly
debated issue that has been addressed by using mtDNA sequences. The limited genetic variability of
human mtDNA has been explained in terms of a recent common genetic ancestry, thus implying that
all modern-population mtDNAs likely originated from a single woman who lived in Africa less than
200,000 years.

Retrieving Sequence Data from GenBank®

This example uses mitochondrial D-loop sequences isolated for different hominidae species with the
following GenBank Accession numbers.

% Species Description GenBank Accession

data = {'German_Neanderthal' 'AF011222';
'Russian_Neanderthal' 'AF254446';
'European_Human' 'X90314' ;
'Mountain_Gorilla_Rwanda' 'AF089820';
'Chimp_Troglodytes' 'AF176766';
'Puti_Orangutan' 'AF451972';
'Jari_Orangutan' 'AF451964';
'Western_Lowland_Gorilla' 'AY079510';
'Eastern_Lowland_Gorilla' 'AF050738';
'Chimp_Schweinfurthii' 'AF176722';
'Chimp_Vellerosus' 'AF315498';
'Chimp_Verus' 'AF176731';
};

You can use the getgenbank function inside a for-loop to retrieve the sequences from the NCBI data
repository and load them into MATLAB®.

for ind = 1:length(data)

primates(ind).Header = data{ind,1};
primates(ind).Sequence = getgenbank(data{ind,2},'sequenceonly','true');
end

For your convenience, previously downloaded sequences are included in a MAT-file. Note that data in
public repositories is frequently curated and updated; therefore, the results of this example might be
slightly different when you use up-to-date sequences.

load('primates.mat')

Building a UPGMA Phylogenetic Tree using Distance Methods

Compute pairwise distances using the 'Jukes-Cantor' formula and the phylogenetic tree with the
'UPGMA' distance method. Since the sequences are not pre-aligned, seqpdist performs a pairwise
alignment before computing the distances.

5-19
5 Phylogenetic Analysis

distances = seqpdist(primates,'Method','Jukes-Cantor','Alpha','DNA');
UPGMAtree = seqlinkage(distances,'UPGMA',primates)

h = plot(UPGMAtree,'orient','top');
title('UPGMA Distance Tree of Primates using Jukes-Cantor model');
ylabel('Evolutionary distance')

Phylogenetic tree object with 12 leaves (11 branches)

Building a Neighbor-Joining Phylogenetic Tree using Distance Methods

Alternate tree topologies are important to consider when analyzing homologous sequences between
species. A neighbor-joining tree can be built using the seqneighjoin function. Neighbor-joining
trees use the pairwise distance calculated above to construct the tree. This method performs
clustering using the minimum evolution method.

NJtree = seqneighjoin(distances,'equivar',primates)

h = plot(NJtree,'orient','top');
title('Neighbor-Joining Distance Tree of Primates using Jukes-Cantor model');
ylabel('Evolutionary distance')

Phylogenetic tree object with 12 leaves (11 branches)

5-20
 Building a Phylogenetic Tree for the Hominidae Species

Comparing Tree Topologies

Notice that different phylogenetic reconstruction methods result in different tree topologies. The
neighbor-joining tree groups Chimp Vellerosus in a clade with the gorillas, whereas the UPGMA tree
groups it near chimps and orangutans. The getcanonical function can be used to compare these
isomorphic trees.

sametree = isequal(getcanonical(UPGMAtree), getcanonical(NJtree))

sametree =

logical

Exploring the UPGMA Phylogenetic Tree

You can explore the phylogenetic tree by considering the nodes (leaves and branches) within a given
patristic distance from the 'European Human' entry and reduce the tree to the sub-branches of
interest by pruning away non-relevant nodes.

names = get(UPGMAtree,'LeafNames')
[h_all,h_leaves] = select(UPGMAtree,'reference',3,'criteria','distance','threshold',0.3);

subtree_names = names(h_leaves)

5-21
5 Phylogenetic Analysis

leaves_to_prune = ~h_leaves;

pruned_tree = prune(UPGMAtree,leaves_to_prune)
h = plot(pruned_tree,'orient','top');
title('Pruned UPGMA Distance Tree of Primates using Jukes-Cantor model');
ylabel('Evolutionary distance')

names =

12x1 cell array

{'German_Neanderthal' }
{'Russian_Neanderthal' }
{'European_Human' }
{'Chimp_Troglodytes' }
{'Chimp_Schweinfurthii' }
{'Chimp_Verus' }
{'Chimp_Vellerosus' }
{'Puti_Orangutan' }
{'Jari_Orangutan' }
{'Mountain_Gorilla_Rwanda'}
{'Eastern_Lowland_Gorilla'}
{'Western_Lowland_Gorilla'}

subtree_names =

6x1 cell array

{'German_Neanderthal' }
{'Russian_Neanderthal' }
{'European_Human' }
{'Chimp_Troglodytes' }
{'Chimp_Schweinfurthii'}
{'Chimp_Verus' }

Phylogenetic tree object with 6 leaves (5 branches)

5-22
 Building a Phylogenetic Tree for the Hominidae Species

With view you can further explore/edit the phylogenetic tree using an interactive tool. See also
phytreeviewer.

view(UPGMAtree,h_leaves)

5-23
5 Phylogenetic Analysis

References

[1] Ovchinnikov, I.V., et al., "Molecular analysis of Neanderthal DNA from the northern Caucasus",
Nature, 404(6777):490-3, 2000.

[2] Sajantila, A., et al., "Genes and languages in Europe: an analysis of mitochondrial lineages",
Genome Research, 5(1):42-52, 1995.

[3] Krings, M., et al., "Neandertal DNA sequences and the origin of modern humans", Cell,
90(1):19-30, 1997.

[4] Jensen-Seaman, M.I. and Kidd, K.K., "Mitochondrial DNA variation and biogeography of eastern
gorillas", Molecular Ecology, 10(9):2241-7, 2001.

5-24
 Analyzing the Origin of the Human Immunodeficiency Virus

Analyzing the Origin of the Human Immunodeficiency Virus

This example shows how to construct phylogenetic trees from multiple strains of the HIV and SIV
viruses.

Introduction

Mutations accumulate in the genomes of pathogens, in this case the human/simian immunodeficiency
virus, during the spread of an infection. This information can be used to study the history of
transmission events, and also as evidence for the origins of the different viral strains.

There are two characterized strains of human AIDS viruses: type 1 (HIV-1) and type 2 (HIV-2). Both
strains represent cross-species infections. The primate reservoir of HIV-2 has been clearly identified
as the sooty mangabey (Cercocebus atys). The origin of HIV-1 is believed to be the common
chimpanzee (Pan troglodytes).

Retrieve Sequence Information from GenBank®

In this example, the variations in three longest coding regions from seventeen different isolated
strains of the Human and Simian immunodeficiency virus are used to construct a phylogenetic tree.
The sequences for these virus strains can be retrieved from GenBank® using their accession
numbers. The three coding regions of interest, the gag protein, the pol polyprotein and the envelope
polyprotein precursor, can then be extracted from the sequences using the CDS information in the
GenBank records.

% Description Accession CDS:gag/pol/env

data = {'HIV-1 (Zaire)' 'K03454' [1 2 8] ;
'HIV1-NDK (Zaire)' 'M27323' [1 2 8] ;
'HIV-2 (Senegal)' 'M15390' [1 2 8] ;
'HIV2-MCN13' 'AY509259' [1 2 8] ;
'HIV-2UC1 (IvoryCoast)' 'L07625' [1 2 8] ;
'SIVMM251 Macaque' 'M19499' [1 2 8] ;
'SIVAGM677A Green monkey' 'M58410' [1 2 7] ;
'SIVlhoest L''Hoest monkeys' 'AF075269' [1 2 7] ;
'SIVcpz Chimpanzees Cameroon' 'AF115393' [1 2 8] ;
'SIVmnd5440 Mandrillus sphinx' 'AY159322' [1 2 8] ;
'SIVAGM3 Green monkeys' 'M30931' [1 2 7] ;
'SIVMM239 Simian macaque' 'M33262' [1 2 8] ;
'CIVcpzUS Chimpanzee' 'AF103818' [1 2 8] ;
'SIVmon Cercopithecus Monkeys' 'AY340701' [1 2 8] ;
'SIVcpzTAN1 Chimpanzee' 'AF447763' [1 2 8] ;
'SIVsmSL92b Sooty Mangabey' 'AF334679' [1 2 8] ;
};

numViruses = size(data,1)

numViruses =

16

You can use the getgenbank function to copy the data from GenBank into a structure in MATLAB®.
The SearchURL field of the structure contains the address of the actual GenBank record. You can
browse this record using the web command.

5-25
5 Phylogenetic Analysis

acc_num = data{1,2};
lentivirus = getgenbank(acc_num);
web(lentivirus(1).SearchURL)

Retrieve the sequence information from the NCBI GenBank database for the rest of the accession
numbers.

for ind = 2:numViruses

lentivirus(ind) = getgenbank(data{ind,2});
end

For your convenience, previously downloaded sequences are included in a MAT-file. Note that data in
public repositories is frequently curated and updated; therefore the results of this example might be
slightly different when you use up-to-date datasets.

load('lentivirus.mat')

Extract CDS for the GAG, POL, and ENV coding regions. Then extract the nucleotide sequences using
the CDS pointers.

for ind = 1:numViruses

temp_seq = lentivirus(ind).Sequence;
temp_seq = regexprep(temp_seq,'[nry]','a');
CDSs = lentivirus(ind).CDS(data{ind,3});
gag(ind).Sequence = temp_seq(CDSs(1).indices(1):CDSs(1).indices(2));
pol(ind).Sequence = temp_seq(CDSs(2).indices(1):CDSs(2).indices(2));
env(ind).Sequence = temp_seq(CDSs(3).indices(1):CDSs(3).indices(2));
end

Phylogenetic Tree Reconstruction

The seqpdist and seqlinkage commands are used to construct a phylogenetic tree for the GAG
coding region using the 'Tajima-Nei' method to measure the distance between the sequences and the
unweighted pair group method using arithmetic averages, or 'UPGMA' method, for the hierarchical
clustering. The 'Tajima-Nei' method is only defined for nucleotides, therefore nucleotide sequences
are used rather than the translated amino acid sequences. The distance calculation may take quite a
few minutes as it is very computationally intensive.

gagd = seqpdist(gag,'method','Tajima-Nei','Alphabet','NT','indel','pair');
gagtree = seqlinkage(gagd,'UPGMA',data(:,1))
plot(gagtree,'type','angular');
title('Immunodeficiency virus (GAG protein)')

Phylogenetic tree object with 16 leaves (15 branches)

5-26
 Analyzing the Origin of the Human Immunodeficiency Virus

Next construct a phylogenetic tree for the POL polyproteins using the 'Jukes-Cantor' method to
measure distance between sequences and the weighted pair group method using arithmetic
averages, or 'WPGMA' method, for the hierarchical clustering. The 'Jukes-Cantor' method is defined
for amino-acids sequences, which, being significantly shorter than the corresponding nucleotide
sequences, means that the calculation of the pairwise distances will be significantly faster.

Convert nucleotide sequences to amino acid sequences using nt2aa.

for ind = 1:numViruses

aagag(ind).Sequence = nt2aa(gag(ind).Sequence);
aapol(ind).Sequence = nt2aa(pol(ind).Sequence);
aaenv(ind).Sequence = nt2aa(env(ind).Sequence);
end

Calculate the distance and linkage, and then generate the tree.

pold = seqpdist(aapol,'method','Jukes-Cantor','indel','pair');
poltree = seqlinkage(pold,'WPGMA',data(:,1))
plot(poltree,'type','angular');
title('Immunodeficiency virus (POL polyprotein)')

Phylogenetic tree object with 16 leaves (15 branches)

5-27
5 Phylogenetic Analysis

Construct a phylogenetic tree for the ENV polyproteins using the normalized pairwise alignment
scores as distances between sequences and the 'UPGMA', method for hierarchical clustering.

envd = seqpdist(aaenv,'method','Alignment','indel','score',...
'ScoringMatrix','Blosum62');
envtree = seqlinkage(envd,'UPGMA',data(:,1))
plot(envtree,'type','angular');
title('Immunodeficiency virus (ENV polyprotein)')

Phylogenetic tree object with 16 leaves (15 branches)

5-28
 Analyzing the Origin of the Human Immunodeficiency Virus

Build a Consensus Tree

The three trees are similar but there are some interesting differences. For example in the POL tree,
the 'SIVmnd5440 Mandrillus sphinx' sequence is placed close to the HIV-1 strains, but in the ENV
tree it is shown as being very distant to the HIV-1 sequences. Given that the three trees show slightly
different results, a consensus tree using all three regions, may give better general information about
the complete viruses. A consensus tree can be built using a weighted average of the three trees.

weights = [sum(gagd) sum(pold) sum(envd)];

weights = weights / sum(weights);
dist = gagd .* weights(1) + pold .* weights(2) + envd .* weights(3);

Note that different metrics were used in the calculation of the pairwise distances. This could bias the
consensus tree. You may wish to recalculate the distances for the three regions using the same metric
to get an unbiased tree.

tree_hiv = seqlinkage(dist,'average',data(:,1));
plot(tree_hiv,'type','angular');
title('Immunodeficiency virus (Weighted tree)')

5-29
5 Phylogenetic Analysis

Origins of the HIV Virus

The phylogenetic tree resulting from our analysis illustrates the presence of two clusters and some
other isolated strains. The most compact cluster includes all the HIV2 samples; at the top branch of
this cluster we observe the sooty mangabey which has been identified as the origin of this lentivirus
in humans. The cluster containing the HIV1 strain, however is not as compact as the HIV2 cluster.
From the tree it appears that the Chimpanzee is the source of HIV1, however, the origin of the cross-
species transmission to humans is still a matter of debate amongst HIV researchers.

% Add annotations
annotation(gcf,'textarrow',[0.29 0.31],[0.36 0.28],'Color',[1 0.5 0],...
'String',{'Possible HIV type 1 origin'},'TextColor',[1 0.5 0]);

annotation(gcf,'textarrow',[0.42 0.49],[0.45 0.50],'Color',[1 0 0],...

'String',{'HIV type 2 origin'},'TextColor',[1 0 0]);

5-30
 Analyzing the Origin of the Human Immunodeficiency Virus

References:

[1] Gao, F., et al., "Origin of HIV-1 in the chimpanzee Pan troglodytes troglodytes", Nature,
397(6718):436-41, 1999.

[2] Kestler, H.W., et al., "Comparison of simian immunodeficiency virus isolates", Nature,
331(6157):619-22, 1998.

[3] Alizon, M., et al., "Genetic variability of the AIDS virus: nucleotide sequence analysis of two
isolates from African patients", Cell, 46(1):63-74, 1986.

5-31
5 Phylogenetic Analysis

Bootstrapping Phylogenetic Trees

This example shows how to generate bootstrap replicates of DNA sequences. The data generated by
bootstrapping is used to estimate the confidence of the branches in a phylogenetic tree.

Introduction

Bootstrap, jackknife, and permutation tests are common tests used in phylogenetics to estimate the
significance of the branches of a tree. This process can be very time consuming because of the large
number of samples that have to be taken in order to have an accurate confidence estimate. The more
times the data are sampled the better the analysis. A cluster of computers can shorten the time
needed for this analysis by distributing the work to several machines and recombining the data.

Loading Sequence Data and Building the Original Tree

This example uses 12 pre-aligned sequences isolated from different hominidae species and stored in a
FASTA-formatted file. A phylogenetic tree is constructed by using the UPGMA method with pairwise
distances. More specifically, the seqpdist function computes the pairwise distances among the
considered sequences and then the function seqlinkage builds the tree and returns the data in a
phytree object. You can use the phytreeviewer function to visualize and explore the tree.

primates = fastaread('primatesaligned.fa');
num_seqs = length(primates)

num_seqs = 12

orig_primates_dist = seqpdist(primates);
orig_primates_tree = seqlinkage(orig_primates_dist,'average',primates);
phytreeviewer(orig_primates_tree);

5-32
 Bootstrapping Phylogenetic Trees

Making Bootstrap Replicates from the Data

A bootstrap replicate is a shuffled representation of the DNA sequence data. To make a bootstrap
replicate of the primates data, bases are sampled randomly from the sequences with replacement and
concatenated to make new sequences. The same number of bases as the original multiple alignment
is used in this analysis, and then gaps are removed to force a new pairwise alignment. The function
randsample samples the data with replacement. This function can also sample the data randomly
without replacement to perform jackknife analysis. For this analysis, 100 bootstrap replicates for
each sequence are created.

num_boots = 100;
seq_len = length(primates(1).Sequence);

boots = cell(num_boots,1);
for n = 1:num_boots
reorder_index = randsample(seq_len,seq_len,true);
for i = num_seqs:-1:1 %reverse order to preallocate memory
bootseq(i).Header = primates(i).Header;
bootseq(i).Sequence = strrep(primates(i).Sequence(reorder_index),'-','');
end

5-33
5 Phylogenetic Analysis

boots{n} = bootseq;
end

Computing the Distances Between Bootstraps and Phylogenetic Reconstruction

Determining the distances between DNA sequences for a large data set and building the phylogenetic
trees can be time-consuming. Distributing these calculations over several machines/cores decreases
the computation time. This example assumes that you have already started a MATLAB® pool with
additional parallel resources. For information about setting up and selecting parallel configurations,
see "Programming with User Configurations" in the Parallel Computing Toolbox™ documentation. If
you do not have the Parallel Computing Toolbox™, the following PARFOR loop executes sequentially
without any further modification.
fun = @(x) seqlinkage(x,'average',{primates.Header});
boot_trees = cell(num_boots,1);
parpool('local');

Starting parallel pool (parpool) using the 'local' profile ...

parfor (n = 1:num_boots)
dist_tmp = seqpdist(boots{n});
boot_trees{n} = fun(dist_tmp);
end
delete(gcp('nocreate'));

Counting the Branches with Similar Topology

The topology of every bootstrap tree is compared with that of the original tree. Any interior branch
that gives the same partition of species is counted. Since branches may be ordered differently among
different trees but still represent the same partition of species, it is necessary to get the canonical
form for each subtree before comparison. The first step is to get the canonical subtrees of the original
tree using the subtree and getcanonical methods from the Bioinformatics Toolbox™.
for i = num_seqs-1:-1:1 % for every branch, reverse order to preallocate
branch_pointer = i + num_seqs;
sub_tree = subtree(orig_primates_tree,branch_pointer);
orig_pointers{i} = getcanonical(sub_tree);
orig_species{i} = sort(get(sub_tree,'LeafNames'));
end

Now you can get the canonical subtrees for all the branches of every bootstrap tree.
for j = num_boots:-1:1
for i = num_seqs-1:-1:1 % for every branch
branch_ptr = i + num_seqs;
sub_tree = subtree(boot_trees{j},branch_ptr);
clusters_pointers{i,j} = getcanonical(sub_tree);
clusters_species{i,j} = sort(get(sub_tree,'LeafNames'));
end
end

For each subtree in the original tree, you can count how many times it appears within the bootstrap
subtrees. To be considered as similar, they must have the same topology and span the same species.
count = zeros(num_seqs-1,1);
for i = 1 : num_seqs -1 % for every branch
for j = 1 : num_boots * (num_seqs-1)
if isequal(orig_pointers{i},clusters_pointers{j})

5-34
 Bootstrapping Phylogenetic Trees

if isequal(orig_species{i},clusters_species{j})
count(i) = count(i) + 1;
end
end
end
end

Pc = count ./ num_boots % confidence probability (Pc)

Pc = 11×1

1.0000
1.0000
0.9900
0.9900
0.5400
0.5400
1.0000
0.4300
0.3900
0.3900
⋮

Visualizing the Confidence Values in the Original Tree

The confidence information associated with each branch node can be stored within the tree by
annotating the node names. Thus, you can create a new tree, equivalent to the original primates tree,
and annotate the branch names to include the confidence levels computed above. phytreeviewer
displays this data in datatips when the mouse is hovered over the internal nodes of the tree.

[ptrs,dist,names] = get(orig_primates_tree,'POINTERS','DISTANCES','NODENAMES');

for i = 1:num_seqs -1 % for every branch

branch_ptr = i + num_seqs;
names{branch_ptr} = [names{branch_ptr} ', confidence: ' num2str(100*Pc(i)) ' %'];
end

tr = phytree(ptrs,dist,names)

Phylogenetic tree object with 12 leaves (11 branches)

You can select the branch nodes with a confidence level greater than a given threshold, for example
0.9, and view these corresponding nodes in the Phylogenetic Tree app. You can also select these
branch nodes interactively within the app.

high_conf_branch_ptr = find(Pc > 0.9) + num_seqs;

view(tr, high_conf_branch_ptr)

5-35
5 Phylogenetic Analysis

References

[1] Felsenstein, J., "Inferring Phylogenies", Sinaur Associates, Inc., 2004.

[2] Nei, M. and Kumar, S., "Molecular Evolution and Phylogenetics", Oxford University Press. Chapter
4, 2000.

5-36
 6

Mass Spectrometry and Bioanalytics

• “Preprocessing Raw Mass Spectrometry Data” on page 6-2

• “Visualizing and Preprocessing Hyphenated Mass Spectrometry Data Sets for Metabolite and
Protein/Peptide Profiling” on page 6-19
• “Identifying Significant Features and Classifying Protein Profiles” on page 6-38
• “Differential Analysis of Complex Protein and Metabolite Mixtures using Liquid Chromatography/
Mass Spectrometry (LC/MS)” on page 6-52
• “Genetic Algorithm Search for Features in Mass Spectrometry Data” on page 6-71
• “Batch Processing of Spectra Using Sequential and Parallel Computing” on page 6-79
6 Mass Spectrometry and Bioanalytics

Preprocessing Raw Mass Spectrometry Data

This example shows how to improve the quality of raw mass spectrometry data. In particular, this
example illustrates the typical steps for preprocesssing protein surface-enhanced laser desorption/
ionization-time of flight mass spectra (SELDI-TOF).

Loading the Data

Mass spectrometry data can be stored in different formats. If the data is stored in text files with two
columns (the mass/charge (M/Z) ratios and the corresponding intensity values), you can use one of
the following MATLAB® I/O functions: importdata, dlmread, or textscan. Alternatively, if the
data is stored in JCAMP-DX formatted files, you can use the function jcampread. If the data is
contained in a spreadsheet of an Excel® workbook, you can use the function xlsread.If the data is
stored in mzXML formatted files, you can use the function mzxmlread, and finally, if the data is
stored in SPC formatted files, you can use tgspcread.

This example uses spectrograms taken from one of the low-resolution ovarian cancer NCI/FDA data
sets from the FDA-NCI Clinical Proteomics Program Databank. These spectra were generated using
the WCX2 protein-binding chip, two with manual sample handling and two with a robotic sample
dispenser/processor.
sample = importdata('mspec01.csv')

sample =

struct with fields:

data: [15154x2 double]

textdata: {'M/Z' 'Intensity'}
colheaders: {'M/Z' 'Intensity'}

The M/Z ratios are in the first column of the data field and the ion intensities are in the second.
MZ = sample.data(:,1);
Y = sample.data(:,2);

For better manipulation of the data, you can load multiple spectrograms and concatenate them into a
single matrix. Use the dlmread function to read comma separated value files. Note: This example
assumes that the M/Z ratios are the same for the four files. For data sets with different M/Z ratios,
use msresample to create a uniform M/Z vector.
files = {'mspec01.csv','mspec02.csv','mspec03.csv','mspec04.csv'};

for i = 1:4
Y(:,i) = dlmread(files{i},',',1,1); % skips the first row (header)
end

Use the plot command to inspect the loaded spectrograms.

plot(MZ,Y)
axis([0 20000 -20 105])
xlabel('Mass/Charge (M/Z)')
ylabel('Relative Intensity')
title('Four Low-Resolution Mass Spectrometry Examples')

6-2
 Preprocessing Raw Mass Spectrometry Data

Resampling the Spectra

Resampling mass spectrometry data has several advantages. It homogenizes the mass/charge (M/Z)
vector, allowing you to compare different spectra under the same reference and at the same
resolution. In high-resolution data sets, the large size of the files leads to computationally intensive
algorithms. However, high-resolution spectra can be redundant. By resampling, you can decimate the
signal into a more manageable M/Z vector, preserving the information content of the spectra. The
msresample function allows you to select a new M/Z vector and also applies an antialias filter that
prevents high-frequency noise from folding into lower frequencies.

Load a high-resolution spectrum taken from the high-resolution ovarian cancer NCI/FDA data set. For
convenience, the spectrum is included in a MAT-formatted file.
load sample_hi_res
numel(MZ_hi_res)

ans =

355760

Down-sample the spectra to 10,000 M/Z points between 2,000 and 11,000. Use the SHOWPLOT
property to create a customized plot that lets you follow and assess the quality of the preprocessing
action.
[MZH,YH] = msresample(MZ_hi_res,Y_hi_res,10000,'RANGE',[2000 11000],'SHOWPLOT',true);

6-3
6 Mass Spectrometry and Bioanalytics

Zooming into a reduced region reveals the detail of the down-sampling procedure.

axis([3875 3895 0 90])

6-4
 Preprocessing Raw Mass Spectrometry Data

Baseline Correction

Mass spectrometry data usually show a varying baseline caused by the chemical noise in the matrix
or by ion overloading. The msbackadj function estimates a low-frequency baseline, which is hidden
among high-frequency noise and signal peaks. It then subtracts the baseline from the spectrogram.

Adjust the baseline of the set of spectrograms and show only the second one and its estimated
background.

YB = msbackadj(MZ,Y,'WINDOWSIZE',500,'QUANTILE',0.20,'SHOWPLOT',2);

6-5
6 Mass Spectrometry and Bioanalytics

Spectral Alignment of Profiles

Miscalibration of the mass spectrometer leads to variations of the relationship between the observed
M/Z vector and the true time-of-flight of the ions. Therefore, systematic shifts can appear in repeated
experiments. When a known profile of peaks is expected in the spectrogram, you can use the function
msalign to standardize the M/Z values.

To align the spectrograms, provide a set of M/Z values where reference peaks are expected to appear.
You can also define a vector with relative weights that is used by the aligning algorithm to emphasize
peaks with small area.

P = [3991.4 4598 7964 9160]; % M/Z location of reference peaks

W = [60 100 60 100]; % Weight for reference peaks

Display a heat map to observe the alignment of the spectra before and after applying the alignment
algorithm.

msheatmap(MZ,YB,'MARKERS',P,'RANGE',[3000 10000])
title('Before Alignment')

6-6
 Preprocessing Raw Mass Spectrometry Data

Align the set of spectrograms to the reference peaks given.

YA = msalign(MZ,YB,P,'WEIGHTS',W);
msheatmap(MZ,YA,'MARKERS',P,'RANGE',[3000 10000])
title('After Alignment')

6-7
6 Mass Spectrometry and Bioanalytics

Normalization

In repeated experiments, it is common to find systematic differences in the total amount of desorbed
and ionized proteins. The msnorm function implements several variations of typical normalization (or
standardization) methods.

For example, one of many methods to standardize the values of the spectrograms is to rescale the
maximum intensity of every signal to a specific value, for instance 100. It is also possible to ignore
problematic regions; for example, in serum samples you might want to ignore the low-mass region
(M/Z < 1000 Da.).

YN1 = msnorm(MZ,YA,'QUANTILE',1,'LIMITS',[1000 inf],'MAX',100);

figure
plot(MZ,YN1)
axis([0 10000 -20 150])
xlabel('Mass/Charge (M/Z)')
ylabel('Relative Intensity')
title('Normalized to the Maximum Peak')

6-8
 Preprocessing Raw Mass Spectrometry Data

The msnorm function can also standardize by using the area under the curve (AUC) and then rescale
the spectrograms to have relative intensities below 100.

YN2 = msnorm(MZ,YA,'LIMITS',[1000 inf],'MAX',100);

figure
plot(MZ,YN2)
axis([0 10000 -20 150])
xlabel('Mass/Charge (M/Z)')
ylabel('Relative Intensity')
title('Normalized Using the Area Under the Curve (AUC)')

6-9
6 Mass Spectrometry and Bioanalytics

Peak Preserving Noise Reduction

Standardized spectra usually contain a mixture of noise and signal. Some applications require
denoising of the spectrograms to improve the validity and precision of the observed mass/charge
values of the peaks in the spectra. For the same reason, denoising also improves further peak
detection algorithms. However, it is important to preserve the sharpness (or high-frequency
components) of the peak as much as possible. For this, you can use Lowess smoothing (mslowess)
and polynomial filters (mssgolay).

Smooth the spectrograms with a polynomial filter of second order.

YS = mssgolay(MZ,YN2,'SPAN',35,'SHOWPLOT',3);

6-10
 Preprocessing Raw Mass Spectrometry Data

Zooming into a reduced region reveals the detail of the smoothing algorithm.

axis([8000 9000 -1 8])

6-11
6 Mass Spectrometry and Bioanalytics

Peak Finding with Wavelets Denoising

A simple approach to find putative peaks is to look at the first derivative of the smoothed signal, then
filer some of these locations to avoid small ion-intensity peaks.

P1 = mspeaks(MZ,YS,'DENOISING',false,'HEIGHTFILTER',2,'SHOWPLOT',1)

P1 =

4x1 cell array

{164x2 double}
{171x2 double}
{169x2 double}
{147x2 double}

6-12
 Preprocessing Raw Mass Spectrometry Data

The mspeaks function can also estimate the noise using wavelets denoising. This method is generally
more robust, because peak detection can be achieved directly over noisy spectra. The algorithm will
adapt to varying noise conditions of the signal, and peaks can be resolved even if low resolution or
oversegmentation exists.

P2 = mspeaks(MZ,YN2,'BASE',12,'MULTIPLIER',10,'HEIGHTFILTER',1,'SHOWPLOT',1)

P2 =

4x1 cell array

{322x2 double}
{370x2 double}
{324x2 double}
{295x2 double}

6-13
6 Mass Spectrometry and Bioanalytics

Eliminate extra peaks in the low-mass region

P3 = cellfun( @(x) x(x(:,1)>1500,:),P2,'UNIFORM',false)

P3 =

4x1 cell array

{81x2 double}
{93x2 double}
{57x2 double}
{53x2 double}

Binning: Peak Coalescing by Hierarchical Clustering

Peaks corresponding to similar compounds may still be reported with slight mass/charge differences
or drifts. Assuming that the four spectrograms correspond to comparable biological/chemical
samples, it might be useful to compare peaks from different spectra, which requires peak binning
(a.k.a. peak coalescing). The crucial task in data binning is to create a common mass/charge
reference vector (or bins). Ideally, bins should collect one peak from each signal and should avoid
collecting multiple relevant peaks from the same signal into the same bin.

This example uses hierarchical clustering to calculate a common mass/charge reference vector. The
approach is sufficient when using low-resolution spectra; however, for high-resolution spectra or for

6-14
 Preprocessing Raw Mass Spectrometry Data

data sets with many spectrograms, the function mspalign provides other scalable methods to
estimate a common mass/charge reference and perform data binning.

Put all the peaks into a single array and construct a vector with the spectrogram index for each peak.

allPeaks = cell2mat(P3);
numPeaks = cellfun(@(x) length(x),P3);
Sidx = accumarray(cumsum(numPeaks),1);
Sidx = cumsum(Sidx)-Sidx;

Create a custom distance function that penalizes clusters containing peaks from the same
spectrogram, then perform hierarchical clustering.

distfun = @(x,y) (x(:,1)-y(:,1)).^2 + (x(:,2)==y(:,2))*10^6

tree = linkage(pdist([allPeaks(:,1),Sidx],distfun));
clusters = cluster(tree,'CUTOFF',75,'CRITERION','Distance');

distfun =

function_handle with value:

@(x,y)(x(:,1)-y(:,1)).^2+(x(:,2)==y(:,2))*10^6

The common mass/charge reference vector (CMZ) is found by calculating the centroids for each
cluster.

CMZ = accumarray(clusters,prod(allPeaks,2))./accumarray(clusters,allPeaks(:,2));

Similarly, the maximum peak intensity of every cluster is also computed.

PR = accumarray(clusters,allPeaks(:,2),[],@max);

[CMZ,h] = sort(CMZ);
PR = PR(h);

figure
hold on
box on
plot([CMZ CMZ],[-10 100],'-k')
plot(MZ,YN2)
axis([7200 8500 -10 100])
xlabel('Mass/Charge (M/Z)')
ylabel('Relative Intensity')
title('Common Mass/Charge (M/Z) Locations found by Clustering')

6-15
6 Mass Spectrometry and Bioanalytics

Dynamic Programming Binning

The samplealign function allows you to use a dynamic programming algorithm to assign the
observed peaks in each spectrogram to the common mass/charge reference vector (CMZ).

When simpler binning approaches are used, such as rounding the mass/charge values or using
nearest neighbor quantization to the CMZ vector, the same peak from different spectra my be assigned
to different bins due to the small drifts that still exist. To circumvent this problem, the bin size can be
increased with the sacrifice of mass spectrometry peak resolution. By using dynamic programming
binning, you preserve the resolution while minimizing the problem of assigning similar compounds
from different spectrograms to different peak locations.

PA = nan(numel(CMZ),4);
for i = 1:4
[j,k] = samplealign([CMZ PR],P3{i},'BAND',15,'WEIGHTS',[1 .1]);
PA(j,i) = P3{i}(k,2);
end

figure
hold on
box on
plot([CMZ CMZ],[-10 100],':k')
plot(MZ,YN2)
plot(CMZ,PA,'o')
axis([7200 8500 -10 100])
xlabel('Mass/Charge (M/Z)')

6-16
 Preprocessing Raw Mass Spectrometry Data

ylabel('Relative Intensity')
title('Peaks Aligned to the Common Mass/Charge (M/Z) Reference')

Use msviewer to inspect the preprocessed spectrograms on a given range (for example, between
values 7600 and 8200).

r1 = 7600;
r2 = 8200;
range = MZ > r1 & MZ < r2;
rangeMarkers = CMZ(CMZ > r1 & CMZ < r2);

msviewer(MZ(range),YN2(range,:),'MARKERS',rangeMarkers,'GROUP',1:4)

6-17
6 Mass Spectrometry and Bioanalytics

See Also
mssgolay | msnorm | msalign | msheatmap | msbackadj | msresample | mspeaks | msviewer

Related Examples
• “Batch Processing of Spectra Using Sequential and Parallel Computing” on page 6-79
• “Visualizing and Preprocessing Hyphenated Mass Spectrometry Data Sets for Metabolite and
Protein/Peptide Profiling” on page 6-19
• “Identifying Significant Features and Classifying Protein Profiles” on page 6-38

6-18
 Visualizing and Preprocessing Hyphenated Mass Spectrometry Data Sets for Metabolite and Protein/Peptide Profiling

Visualizing and Preprocessing Hyphenated Mass Spectrometry

Data Sets for Metabolite and Protein/Peptide Profiling

This example shows how to manipulate, preprocess and visualize data from Liquid Chromatography
coupled with Mass Spectrometry (LC/MS). These large and high dimensional data sets are extensively
utilized in proteomics and metabolomics research. Visualizing complex peptide or metabolite
mixtures provides an intuitive method to evaluate the sample quality. In addition, methodical
correction and preprocessing can lead to automated high throughput analysis of samples allowing
accurate identification of significant metabolites and specific peptide features in a biological sample.

Introduction

In a typical hyphenated mass spectrometry experiment, proteins and metabolites are harvested from
cells, tissues, or body fluids, dissolved and denatured in solution, and enzymatically digested into
mixtures. These mixtures are then separated either by High Performance Liquid Chromatography
(HPLC), capillary electrophoresis (CE), or gas chromatography (GC) and coupled to a mass-
spectrometry identification method, such as Electrospray Ionization Mass Spectrometry (ESI-MS),
matrix assisted ionization (MALDI or SELDI TOF-MS), or tandem mass spectrometry (MS/MS).

Open Data Repositories and mzXML File Format

For this example, we use a test sample LC-ESI-MS data set with a seven protein mix. The data in this
example (7MIX_STD_110802_1) is from the Sashimi Data Repository. The data set is not distributed
with MATLAB®. To complete this example, you must download the data set into a local directory or
your own repository. Alternatively, you can try other data sets available in other public databases for
protein expression data such as the Peptide Atlas at the Institute of Systems Biology [1].

Most of the current mass spectrometers can translate or save the acquisition data using the mzXML
schema. This standard is an XML (eXtensible Markup Language)-based common file format developed
by the Sashimi project to address the challenges involved in representing data sets from different
manufacturers and from different experimental setups into a common and expandable schema.
mzXML files used in hyphenated mass spectrometry are usually very large. The MZXMLINFO function
allows you to obtain basic information about the dataset without reading it into memory. For example,
you can retrieve the number of scans, the range of the retention time, and the number of tandem MS
instruments (levels).

info = mzxmlinfo('7MIX_STD_110802_1.mzXML','NUMOFLEVELS',true)

info =

struct with fields:

Filename: '7MIX_STD_110802_1.mzXML'
FileModDate: '01-Feb-2013 11:54:30'
FileSize: 26789612
NumberOfScans: 7161
StartTime: 'PT0.00683333S'
EndTime: 'PT200.036S'
DataProcessingIntensityCutoff: 'N/A'
DataProcessingCentroided: 'true'
DataProcessingDeisotoped: 'N/A'
DataProcessingChargeDeconvoluted: 'N/A'

6-19
6 Mass Spectrometry and Bioanalytics

DataProcessingSpotIntegration: 'N/A'
NumberOfMSLevels: 2

The MZXMLREAD function reads the XML document into a MATLAB structure. The fields scan and
index are placed at the first level of the output structure for improved access to the spectral data.
The remainder of the mzXML document tree is parsed according to the schema specifications. This
LC/MS data set contains 7161 scans with two MS levels. For this example you will use only the first
level scans. Second level spectra are usually used for peptide/protein identification, and come at a
later stage in some types of workflow analyses. MZXMLREAD can filter the desired scans without
loading all the dataset into memory:

mzXML_struct = mzxmlread('7MIX_STD_110802_1.mzXML','LEVEL',1)

mzXML_struct =

struct with fields:

scan: [2387×1 struct]

mzXML: [1×1 struct]
index: [1×1 struct]

If you receive any errors related to memory or java heap space during the loading, try increasing your
java heap space as described here.

More detailed information pertaining the mass-spectrometer and the experimental conditions are
found in the field msRun.

mzXML_struct.mzXML.msRun

ans =

struct with fields:

scanCount: 7161
startTime: "PT0.00683333S"
endTime: "PT200.036S"
parentFile: [1×1 struct]
msInstrument: [1×1 struct]
dataProcessing: [1×1 struct]

To facilitate the handling of the data, the MZXML2PEAKS function extracts the list of peaks from each
scan into a cell array (peaks]) and their respective retention time into a column vector (time). You
can extract the spectra of certain level by setting the LEVEL input parameter.

[peaks,time] = mzxml2peaks(mzXML_struct);
numScans = numel(peaks)

numScans =

2387

6-20
Visualizing and Preprocessing Hyphenated Mass Spectrometry Data Sets for Metabolite and Protein/Peptide Profiling

The MSDOTPLOT function creates an overview display of the most intense peaks in the entire data set.
In this case, we visualize only the most intense 5% ion intensity peaks by setting the input parameter
QUANTILE to 0.95.

h = msdotplot(peaks,time,'quantile',.95);
title('5 Percent Overall Most Intense Peaks')

You can also filter the peaks individually for each scan using a percentile of the base peak intensity.
The base peak is the most intense peak found in each scan [2]. This parameter is given automatically
by most of the spectrometers. This operation requires querying into the mxXML structure to obtain
the base peak information. Note that you could also find the base peak intensity by iterating the MAX
function over the peak list.

basePeakInt = [mzXML_struct.scan.basePeakIntensity]';
peaks_fil = cell(numScans,1);
for i = 1:numScans
h = peaks{i}(:,2) > (basePeakInt(i).*0.75);
peaks_fil{i} = peaks{i}(h,:);
end

whos('basePeakInt','level_1','peaks','peaks_fil')
msdotplot(peaks_fil,time)
title('Peaks Above (0.75 x Base Peak Intensity) for Each Scan')

Name Size Bytes Class Attributes

basePeakInt 2387x1 19096 double

6-21
6 Mass Spectrometry and Bioanalytics

peaks 2387x1 14031800 cell

peaks_fil 2387x1 289568 cell

You can customize a 3-D overview of the filtered peaks using the STEM3 function. The STEM3 function
requires to put the data into three vectors, whose elements form the triplets (the retention time, the
mass/charge, and the intensity value) that represent every stem.

peaks_3D = cell(numScans,1);
for i = 1:numScans
peaks_3D{i}(:,[2 3]) = peaks_fil{i};
peaks_3D{i}(:,1) = time(i);
end
peaks_3D = cell2mat(peaks_3D);

figure
stem3(peaks_3D(:,1),peaks_3D(:,2),peaks_3D(:,3),'marker','none')
axis([0 12000 400 1500 0 1e9])
view(60,60)
xlabel('Retention Time (seconds)')
ylabel('Mass/Charge (M/Z)')
zlabel('Relative Ion Intensity')
title('Peaks Above (0.75 x Base Peak Intensity) for Each Scan')

6-22
Visualizing and Preprocessing Hyphenated Mass Spectrometry Data Sets for Metabolite and Protein/Peptide Profiling

You can plot colored stems using the PATCH function. For every triplet in peaks_3D, interleave a new
triplet with the intensity value set to zero. Then create a color vector dependent on the intensity of
the stem. A logarithmic transformation enhances the dynamic range of the colormap. For the
interleaved triplets assign a NaN, so that PATCH function does not draw lines connecting contiguous
stems.

peaks_patch = sortrows(repmat(peaks_3D,2,1));
peaks_patch(2:2:end,3) = 0;

col_vec = log(peaks_patch(:,3));
col_vec(2:2:end) = NaN;

figure
patch(peaks_patch(:,1),peaks_patch(:,2),peaks_patch(:,3),col_vec,...
'edgeColor','flat','markeredgecolor','flat','Marker','x','MarkerSize',3);
axis([0 12000 400 1500 0 1e9])
view(90,90)
xlabel('Retention Time (seconds)')
ylabel('Mass/Charge (M/Z)')
zlabel('Relative Ion Intensity')
title('Peaks Above (0.75 x Base Peak Intensity) for Each Scan')

6-23
6 Mass Spectrometry and Bioanalytics

view(40,40)

6-24
Visualizing and Preprocessing Hyphenated Mass Spectrometry Data Sets for Metabolite and Protein/Peptide Profiling

Creating Heat Maps of LC/MS Data Sets

Common techniques in the industry work with peak information (a.k.a. centroided data) instead of
raw signals. This may save memory, but some important details are not visible, especially when it is
necessary to inspect samples with complex mixtures. To further analyze this data set, we can create a
common grid in the mass/charge dimension. Since not all of the scans have enough information to
reconstruct the original signal, we use a peak preserving resampling method. By choosing the
appropriate parameters for the MSPPRESAMPLE function, you can ensure that the resolution of the
spectra is not lost, and that the maximum values of the peaks correlate to the original peak
information.

[MZ,Y] = msppresample(peaks,5000);
whos('MZ','Y')

Name Size Bytes Class Attributes

MZ 5000x1 20000 single

Y 5000x2387 47740000 single

With this matrix of ion intensities, Y, you can create a colored heat map. The MSHEATMAP function
automatically adjusts the colorbar utilized to show the statistically significant peaks with hot colors
and the noisy peaks with cold colors. The algorithm is based on clustering significant peaks and noisy
peaks by estimating a mixture of Gaussians with an Expectation-Maximization approach. Additionally,
you can use the MIDPOINT input parameter to select an arbitrary threshold to separate noisy peaks
from significant peaks, or you can interactively shift the colormap to hide or unhide peaks. When

6-25
6 Mass Spectrometry and Bioanalytics

working with heat maps, it is common to display the logarithm of the ion intensities, which enhances
the dynamic range of the colormap.

fh1 = msheatmap(MZ,time,log(Y),'resolution',.1,'range',[500 900]);

title('Original LC/MS Data Set')

You can zoom to small regions of interest to observe the data, either interactively or
programmatically using the AXIS function. We observe some regions with high relative ion intensity.
These represent peptides in the biological sample.

axis([527 539 385 496])

6-26
Visualizing and Preprocessing Hyphenated Mass Spectrometry Data Sets for Metabolite and Protein/Peptide Profiling

You can overlay the original peak information of the LC/MS data set. This lets you evaluate the
performance of the peak-preserving resampling operation. You can use the returned handle to hide/
unhide the dots.

dp1 = msdotplot(peaks,time);

6-27
6 Mass Spectrometry and Bioanalytics

Calibrating the Mass/Charge Location of Peaks to a Common Grid

The two dimensional peaks appear to be noisy or they do not show a compact shape in contiguous
spectra. This is a common problem for many mass spectrometers. Random fluctuations of the mass/
charge value extracted from peaks of replicate profiles are reported to range from 0.1% to 0.3% [3].
Such variability can be caused by several factors, e.g. poor calibration of the detector, low signal-to-
noise ratio, or problems in the peak extraction algorithms. The MSPALIGN function implements
advanced data binning algorithms that synchronize all the spectra in a data set to a common mass/
charge grid (CMZ). CMZ can be chosen arbitrarily or it can be estimated after analyzing the data
[2,4,5]. The peak matching procedure can use either a nearest neighbor rule or a dynamic
programming alignment.

[CMZ, peaks_CMZ] = mspalign(peaks);

Repeat the visualization process with the aligned peaks: perform peak preserving resampling, create
a heat map, overlay the aligned peak information, and zoom into the same region of interest as
before. When the spectrum is re-calibrated, it is possible to distinguish the isotope patterns of some
of the peptides.

[MZ_A,Y_A] = msppresample(peaks_CMZ,5000);
fh2 = msheatmap(MZ_A,time,log(Y_A),'resolution',.10,'range',[500 900]);
title('LC/MS Data Set with the Mass/Charge Calibrated to a CMZ')
dp2 = msdotplot(peaks_CMZ,time);
axis([527 539 385 496])

6-28
Visualizing and Preprocessing Hyphenated Mass Spectrometry Data Sets for Metabolite and Protein/Peptide Profiling

Calibrating the Mass/Charge Location of Peaks Locally

MSPALIGN computes a single CMZ for the whole LC/MS data set. This may not be the ideal case for
samples with more complex mixtures of peptides and/or metabolites than the data set utilized in this
example. In the case of complex mixtures, you can align each spectrum to a local set of spectra that
contain only informative peaks (high intensity) with similar retention times, otherwise the calibration
process in regions with small peaks (low intensity) can be biased by other peaks that share similar
mass/charge values but are at different retention times. To perform a finer calibration, you can
employ the SAMPLEALIGN function. This function is a generalization of the Constrained Dynamic
Time Warping (CDTW) algorithms commonly utilized in speech processing [6]. In the following for
loop, we maintain a buffer with the intensities of the previous aligned spectra (LAI). The ion
intensities of the spectra are scaled with the anonymous function SF (inside SAMPLEALIGN) to reduce
the distance between large peaks. SAMPLEALIGN reduces the overall distance of all matched points
and introduces gaps as necessary. In this case we use a finer MZ vector (FMZ), such that we preserve
the correct value of the mass/charge of the peaks as much as possible. Note: this may take some time,
as the CDTW algorithm is executed 2,387 times.

SF = @(x) 1-exp(-x./5e7); % scaling function

DF = @(R,S) sqrt((SF(R(:,2))-SF(S(:,2))).^2 + (R(:,1)-S(:,1)).^2);

FMZ = (500:0.15:900)'; % setup a finer MZ vector

LAI = zeros(size(FMZ)); % init buffer for the last alignment intensities

peaks_FMZ = cell(numScans,1);
for i = 1:numScans
% show progress

6-29
6 Mass Spectrometry and Bioanalytics

if ~rem(i,250)
fprintf(' %d...',i);
end
% align peaks in current scan to LAI
[k,j] = samplealign([FMZ,LAI],double(peaks{i}),'band',1.5,'gap',[0,2],'dist',DF);
% updating the LAI buffer
LAI = LAI*.25;
LAI(k) = LAI(k) + peaks{i}(j,2);
% save the alignment
peaks_FMZ{i} = [FMZ(k) peaks{i}(j,2)];
end

250... 500... 750... 1000... 1250... 1500... 1750... 2000... 2250...

Repeat the visualization process and zoom to the region of interest.

[MZ_B,Y_B] = msppresample(peaks_FMZ,4000);
fh3 = msheatmap(MZ_B,time,log(Y_B),'resolution',.10,'range',[500 900]);
title('LC/MS Data Set with the Mass/Charge Calibrated Locally')
dp3 = msdotplot(peaks_FMZ,time);
axis([527 539 385 496])

As a final step to improve the image, you can apply a Gaussian filter in the chromatographic direction
to smooth the whole data set.
Gpulse = exp(-.1*(-10:10).^2)./sum(exp(-.1*(-10:10).^2));
YF = convn(Y_B,Gpulse,'same');
fh4 = msheatmap(MZ_B,time,log(YF),'resolution',.10,'limits',[500 900]);

6-30
Visualizing and Preprocessing Hyphenated Mass Spectrometry Data Sets for Metabolite and Protein/Peptide Profiling

title('Final Enhanced LC/MS Data Set')

dp4 = msdotplot(peaks_FMZ,time);
axis([527 539 385 496])

You can link the axes of two heat maps, to interactively or programmatically compare regions
between two data sets. In this case we compare the original and the final enhanced LC/MS matrices.

linkaxes(findobj([fh1 fh4],'Tag','MSHeatMap'))
axis([521 538 266 617])

6-31
6 Mass Spectrometry and Bioanalytics

6-32
Visualizing and Preprocessing Hyphenated Mass Spectrometry Data Sets for Metabolite and Protein/Peptide Profiling

Extracting Spectra Using the Total Ion Chromatogram

Once the LC/MS data set is smoothed and resampled into a regular grid, it is possible to extract the
most informative spectra by looking at the local maxima of the Total Ion Chromatogram (TIC). The
TIC is straightforwardly computed by summing the rows of YF. Then, use the MSPEAKS function to
find the retention time values for extracting selected subsets of spectra.

TIC = mean(YF);
pt = mspeaks(time,TIC','multiplier',10,'overseg',100,'showplot',true);
title('Extracted peaks from the Total Ion Chromatogram (TIC)')
pt(pt(:,1)>4000,:) = []; % remove spectra above 4000 seconds
numPeaks = size(pt,1)

numPeaks =

22

6-33
6 Mass Spectrometry and Bioanalytics

Create a 3-D plot of the selected spectra.

xRows = samplealign(time,pt(:,1),'width',1); % finds the time index for every peak

xSpec = YF(:,xRows); % gets the signals to plot

figure;
hold on
box on
plot3(repmat(MZ_B,1,numPeaks),repmat(1:numPeaks,numel(MZ_B),1),xSpec)
view(20,85)

ax = gca;
ax.YTick = 1:numPeaks;
ax.YTickLabel = num2str(time(xRows));
axis([500 900 0 numPeaks 0 1e8])

xlabel('Mass/Charge (M/Z)')
ylabel('Time')
zlabel('Relative Ion Intensity')
title('Extracted Spectra Subset')

6-34
Visualizing and Preprocessing Hyphenated Mass Spectrometry Data Sets for Metabolite and Protein/Peptide Profiling

Overlay markers for the extracted spectra over the enhanced heatmap.

linkaxes(findobj(fh4,'Tag','MSHeatMap'),'off')
figure(fh4)
hold on
for i = 1:numPeaks
plot([400 1500],xRows([i i]),'m')
end
axis([500 900 100 925])
dp4.Visible = 'off';
title('Final Enhanced LC/MS Data Set with Extracted Spectra Marked')

6-35
6 Mass Spectrometry and Bioanalytics

6-36
Visualizing and Preprocessing Hyphenated Mass Spectrometry Data Sets for Metabolite and Protein/Peptide Profiling

References

[1] Desiere, F. et al., "The Peptide Atlas Project", Nucleic Acids Research, 34:D655-8, 2006.

[2] Purvine, S., Kolker, N., and Kolker, E., "Spectral Quality Assessment for High-Throughput Tandem
Mass Spectrometry Proteomics", OMICS: A Journal of Integrative Biology, 8(3):255-65, 2004.

[3] Kazmi, A.S., et al., "Alignment of high resolution mass spectra: Development of a heuristic
approach for metabolomics", Metabolomics, 2(2):75-83, 2006.

[4] Jeffries, N., "Algorithms for alignment of mass spectrometry proteomic data", Bioinformatics,
21(14):3066-3073, 2005.

[5] Yu, W., et al., "Multiple peak alignment in sequential data analysis: A scale-space based approach",
IEEE®/ACM Trans. Computational Biology and Bioinformatics, 3(3):208-219, 2006.

[6] Sakoe, H. and Chiba s., "Dynamic programming algorithm optimization for spoken word
recognition", IEEE Trans. Acoustics, Speech and Signal Processing, ASSP-26(1):43-9, 1978.

6-37
6 Mass Spectrometry and Bioanalytics

Identifying Significant Features and Classifying Protein

Profiles

This example shows how to classify mass spectrometry data and use some statistical tools to look for
potential disease markers and proteomic pattern diagnostics.

Introduction

Serum proteomic pattern diagnostics can be used to differentiate samples from patients with and
without disease. Profile patterns are generated using surface-enhanced laser desorption and
ionization (SELDI) protein mass spectrometry. This technology has the potential to improve clinical
diagnostics tests for cancer pathologies. The goal is to select a reduced set of measurements or
"features" that can be used to distinguish between cancer and control patients. These features will be
ion intensity levels at specific mass/charge values.

Preprocess Data

The ovarian cancer data set in this example is from the FDA-NCI Clinical Proteomics Program
Databank. The data set was generated using the WCX2 protein array. The data set includes 95
controls and 121 ovarian cancers. For a detailed description of this data set, see [1] and [4].

This example assumes that you already have the preprocessed data
OvarianCancerQAQCdataset.mat. However, if you do not have the data file, you can recreate by
following the steps in the example “Batch Processing of Spectra Using Sequential and Parallel
Computing” on page 6-79.

Alternatively, you can run the script msseqprocessing.m.

addpath(fullfile(matlabroot,'examples','bioinfo','main')) % Make sure the supporting files are on

type msseqprocessing

% MSSEQPROCESSING Script to create OvarianCancerQAQCdataset.mat (used in

% CANCERDETECTDEMO). Before running this file initialize the variable
% "repository" to the full path where you placed you mass-spectrometry
% files. For Example:
%
% repository = 'F:/MassSpecRepository/OvarianCD_PostQAQC/';
%
% or
%
% repository = '/home/username/MassSpecRepository/OvarianCD_PostQAQC/';
%
% The approximate time of execution is 18 minutes (Pentium 4, 4GHz). If you
% have the Parallel Computing Toolbox refer to BIODISTCOMPDEMO to see
% how you can speed this analysis up.

% Copyright 2003-2008 The MathWorks, Inc.

repositoryC = [repository 'Cancer/'];

repositoryN = [repository 'Normal/'];

filesCancer = dir([repositoryC '*.txt']);

6-38
 Identifying Significant Features and Classifying Protein Profiles

NumberCancerDatasets = numel(filesCancer);
fprintf('Found %d Cancer mass-spectrograms.\n',NumberCancerDatasets)
filesNormal = dir([repositoryN '*.txt']);
NumberNormalDatasets = numel(filesNormal);
fprintf('Found %d Control mass-spectrograms.\n',NumberNormalDatasets)

files = [ strcat('Cancer/',{filesCancer.name}) ...

strcat('Normal/',{filesNormal.name})];
N = numel(files); % total number of files

fprintf('Total %d mass-spectrograms to process...\n',N)

[MZ,Y] = msbatchprocessing(repository,files);

disp('Finished; normalizing and saving to OvarianCancerQAQCdataset.mat.')

Y = msnorm(MZ,Y,'QUANTILE',0.5,'LIMITS',[3500 11000],'MAX',50);

grp = [repmat({'Cancer'},size(filesCancer));...
repmat({'Normal'},size(filesNormal))];

save OvarianCancerQAQCdataset.mat Y MZ grp

The preprocessing steps from the script and example listed above are intended to illustrate a
representative set of possible pre-processing procedures. Using different steps or parameters may
lead to different and possibly improved results of this example.

Load Data

Once you have the preprocessed data, you can load it into MATLAB.

load OvarianCancerQAQCdataset
whos

Name Size Bytes Class Attributes

MZ 15000x1 120000 double

Y 15000x216 25920000 double
grp 216x1 25056 cell

There are three variables: MZ, Y, grp. MZ is the mass/charge vector, Y is the intensity values for all
216 patients (control and cancer), and grp holds the index information as to which of these samples
represent cancer patients and which ones represent normal patients.

Initialize some variables that will be used through out the example.

N = numel(grp); % Number of samples

Cidx = strcmp('Cancer',grp); % Logical index vector for Cancer samples
Nidx = strcmp('Normal',grp); % Logical index vector for Normal samples
Cvec = find(Cidx); % Index vector for Cancer samples
Nvec = find(Nidx); % Index vector for Normal samples
xAxisLabel = 'Mass/Charge (M/Z)'; % x label for plots
yAxisLabel = 'Ion Intensity'; % y label for plots

6-39
6 Mass Spectrometry and Bioanalytics

Visualizing Some of the Samples

You can plot some data sets into a figure window to visually compare profiles from the two groups; in
this example five spectrograms from cancer patients (blue) and five from control patients (green) are
displayed.

figure; hold on;

hC = plot(MZ,Y(:,Cvec(1:5)),'b');
hN = plot(MZ,Y(:,Nvec(1:5)),'g');
xlabel(xAxisLabel); ylabel(yAxisLabel);
axis([2000 12000 -5 60])
legend([hN(1),hC(1)],{'Control','Ovarian Cancer'})
title('Multiple Sample Spectrograms')

Zooming in on the region from 8500 to 8700 M/Z shows some peaks that might be useful for
classifying the data.

axis([8450,8700,-1,7])

6-40
 Identifying Significant Features and Classifying Protein Profiles

Another way to visualize the whole data set is to look at the group average signal for the control and
cancer samples. You can plot the group average and the envelopes of each group.

mean_N = mean(Y(:,Nidx),2); % group average for control samples

max_N = max(Y(:,Nidx),[],2); % top envelopes of the control samples
min_N = min(Y(:,Nidx),[],2); % bottom envelopes of the control samples
mean_C = mean(Y(:,Cidx),2); % group average for cancer samples
max_C = max(Y(:,Cidx),[],2); % top envelopes of the control samples
min_C = min(Y(:,Cidx),[],2); % bottom envelopes of the control samples

figure; hold on;

hC = plot(MZ,mean_C,'b');
hN = plot(MZ,mean_N,'g');
gC = plot(MZ,[max_C min_C],'b--');
gN = plot(MZ,[max_N min_N],'g--');
xlabel(xAxisLabel); ylabel(yAxisLabel);
axis([8450,8700,-1,7])
legend([hN,hC,gN(1),gC(1)],{'Control Group Avg.','Ovarian Cancer Group Avg',...
'Control Envelope','Ovarian Cancer Envelope'},...
'Location', 'NorthWest')
title('Group Average and Group Envelopes')

6-41
6 Mass Spectrometry and Bioanalytics

Observe that apparently there is no single feature that can discriminate both groups perfectly.

Ranking Key Features

A simple approach for finding significant features is to assume that each M/Z value is independent
and compute a two-way t-test. rankfeatures returns an index to the most significant M/Z values,
for instance 100 indices ranked by the absolute value of the test statistic. This feature selection
method is also known as a filtering method, where the learning algorithm is not involved on how the
features are selected.

[feat,stat] = rankfeatures(Y,grp,'CRITERION','ttest','NUMBER',100);

The first output of rankfeatures can be used to extract the M/Z values of the significant features.

sig_Masses = MZ(feat);
sig_Masses(1:7)' %display the first seven

ans =

1.0e+03 *

8.1009 8.1016 8.1024 8.1001 8.1032 7.7366 7.7359

The second output of rankfeatures is a vector with the absolute value of the test statistic. You can
plot it over the spectra using yyaxis.

6-42
 Identifying Significant Features and Classifying Protein Profiles

figure;

yyaxis left
plot(MZ, [mean_N mean_C]);
ylim([-1,20])
xlim([7950,8300])
title('Significant M/Z Values')
xlabel(xAxisLabel);
ylabel(yAxisLabel);

yyaxis right
plot(MZ,stat);
ylim([-1,22])
ylabel('Test Statistic');

legend({'Control Group Avg.','Ovarian Cancer Group Avg.', 'Test Statistics'})

Notice that there are significant regions at high M/Z values but low intensity (~8100 Da.). Other
approaches to measure class separability are available in rankfeatures, such as entropy based,
Bhattacharyya, or the area under the empirical receiver operating characteristic (ROC) curve.

Blind Classification Using Linear Discriminant Analysis (LDA)

Now that you have identified some significant features, you can use this information to classify the
cancer and normal samples. Due to the small number of samples, you can run a cross-validation using
the 20% holdout to have a better estimation of the classifier performance. cvpartition allows you

6-43
6 Mass Spectrometry and Bioanalytics

to set the training and test indices for different types of system evaluation methods, such as hold-out,
K-fold and Leave-M-Out.

per_eval = 0.20; % training size for cross-validation

rng('default'); % initialize random generator to the same state
% used to generate the published example
cv = cvpartition(grp,'holdout',per_eval)

cv =

Hold-out cross validation partition

NumObservations: 216
NumTestSets: 1
TrainSize: 173
TestSize: 43

Observe that features are selected only from the training subset and the validation is performed with
the test subset. classperf allows you to keep track of multiple validations.

cp_lda1 = classperf(grp); % initializes the CP object

for k=1:10 % run cross-validation 10 times
cv = repartition(cv);
feat = rankfeatures(Y(:,training(cv)),grp(training(cv)),'NUMBER',100);
c = classify(Y(feat,test(cv))',Y(feat,training(cv))',grp(training(cv)));
classperf(cp_lda1,c,test(cv)); % updates the CP object with current validation
end

After the loop you can assess the performance of the overall blind classification using any of the
properties in the CP object, such as the error rate, sensitivity, specificity, and others.

cp_lda1

Label: ''
Description: ''
ClassLabels: {2x1 cell}
GroundTruth: [216x1 double]
NumberOfObservations: 216
ControlClasses: 2
TargetClasses: 1
ValidationCounter: 10
SampleDistribution: [216x1 double]
ErrorDistribution: [216x1 double]
SampleDistributionByClass: [2x1 double]
ErrorDistributionByClass: [2x1 double]
CountingMatrix: [3x2 double]
CorrectRate: 0.8488
ErrorRate: 0.1512
LastCorrectRate: 0.8837
LastErrorRate: 0.1163
InconclusiveRate: 0
ClassifiedRate: 1
Sensitivity: 0.8208
Specificity: 0.8842
PositivePredictiveValue: 0.8995
NegativePredictiveValue: 0.7962
PositiveLikelihood: 7.0890
NegativeLikelihood: 0.2026

6-44
 Identifying Significant Features and Classifying Protein Profiles

Prevalence: 0.5581
DiagnosticTable: [2x2 double]

This naive approach for feature selection can be improved by eliminating some features based on the
regional information. For example, 'NWEIGHT' in rankfeatures outweighs the test statistic of
neighboring M/Z features such that other significant M/Z values can be incorporated into the subset
of selected features

cp_lda2 = classperf(grp); % initializes the CP object

for k=1:10 % run cross-validation 10 times
cv = repartition(cv);
feat = rankfeatures(Y(:,training(cv)),grp(training(cv)),'NUMBER',100,'NWEIGHT',5);
c = classify(Y(feat,test(cv))',Y(feat,training(cv))',grp(training(cv)));
classperf(cp_lda2,c,test(cv)); % updates the CP object with current validation
end
cp_lda2.CorrectRate % average correct classification rate

ans =

0.9023

PCA/LDA Reduction of the Data Dimensionality

Lilien et al. presented in [2] an algorithm to reduce the data dimensionality that uses principal
component analysis (PCA), then LDA is used to classify the groups. In this example 2000 of the most
significant features in the M/Z space are mapped to the 150 principal components

cp_pcalda = classperf(grp); % initializes the CP object

for k=1:10 % run cross-validation 10 times
cv = repartition(cv);
% select the 2000 most significant features.
feat = rankfeatures(Y(:,training(cv)),grp(training(cv)),'NUMBER',2000);
% PCA to reduce dimensionality
P = pca(Y(feat,training(cv))');
% Project into PCA space
x = Y(feat,:)' * P(:,1:150);
% Use LDA
c = classify(x(test(cv),:),x(training(cv),:),grp(training(cv)));
classperf(cp_pcalda,c,test(cv));
end
cp_pcalda.CorrectRate % average correct classification rate

ans =

0.9814

Randomized Search for Subset Feature Selection

Feature selection can also be reinforced by classification, this approach is usually referred to as a
wrapper selection method. Randomized search for feature selection generates random subsets of
features and assesses their quality independently with the learning algorithm. Later, it selects a pool
of the most frequent good features. Li et al. in [3] apply this concept to the analysis of protein

6-45
6 Mass Spectrometry and Bioanalytics

expression patterns. The randfeatures function allows you to search a subset of features using LDA
or a k-nearest neighbor classifier over randomized subsets of features.

Note: the following example is computationally intensive, so it has been disabled from the example.
Also, for better results you should increase the pool size and the stringency of the classifier from the
default values in randfeatures. Type help randfeatures for more information.

if 0 % <== change to 1 to enable. This may use extensive time to complete.

cv = repartition(cv);
[feat,fCount] = randfeatures(Y(:,training(cv)),grp(training(cv)),...
'CLASSIFIER','da','PerformanceThreshold',0.90);
else
load randFeatCancerDetect
end

Assess the Quality of the Selected Features with the Evaluation Set

The first output from randfeatures is an ordered list of indices of MZ values. The first item occurs
most frequently in the subsets where good classification was achieved. The second output is the
actual counts of the number of times each value was selected. You can use hist to look at this
distribution.

figure;
hist(fCount,max(fCount)+1);

You will see that most values appear at most once in a selected subset. Zooming in gives a better idea
of the details for the more frequently selected values.

6-46
 Identifying Significant Features and Classifying Protein Profiles

axis([0 80 0 100])

Only a few values were selected more than 10 times. You can visualize these by using a stem plot to
show the most frequently selected features.

figure; hold on;

sigFeats = fCount;
sigFeats(sigFeats<=10) = 0;
plot(MZ,[mean_N mean_C]);
stem(MZ(sigFeats>0),sigFeats(sigFeats>0),'r');
axis([2000,12000,-1,80])
legend({'Control Group Avg.','Ovarian Cancer Group Avg.','Significant Features'}, ...
'Location', 'NorthWest')
xlabel(xAxisLabel); ylabel(yAxisLabel);

6-47
6 Mass Spectrometry and Bioanalytics

These features appear to clump together in several groups. You can investigate further how many of
the features are significant by running the following experiment. The most frequently selected
feature is used to classify the data, then the two most frequently selected features are used and so on
until all the features that were selected more than 10 times are used. You can then see if adding more
features improves the classifier.

nSig = sum(fCount>10);
cp_rndfeat = zeros(20,nSig);
for i = 1:nSig
for j = 1:20
cv = repartition(cv);
P = pca(Y(feat(1:i),training(cv))');
x = Y(feat(1:i),:)' * P;
c = classify(x(test(cv),:),x(training(cv),:),grp(training(cv)));
cp = classperf(grp,c,test(cv));
cp_rndfeat(j,i) = cp.CorrectRate; % average correct classification rate
end
end
figure
plot(1:nSig, [max(cp_rndfeat);mean(cp_rndfeat)]);
legend({'Best CorrectRate','Mean CorrectRate'},'Location', 'SouthEast')

6-48
 Identifying Significant Features and Classifying Protein Profiles

From this graph you can see that for as few as three features it is sometimes possible to get perfect
classification. You will also notice that the maximum of the mean correct rate occurs for a small
number of features and then gradually decreases.

[bestAverageCR, bestNumFeatures] = max(mean(cp_rndfeat));

You can now visualize the features that give the best average classification. You can see that these
actually correspond to only three peaks in the data.

figure; hold on;

sigFeats = fCount;
sigFeats(sigFeats<=10) = 0;
ax_handle = plot(MZ,[mean_N mean_C]);
stem(MZ(feat(1:bestNumFeatures)),sigFeats(feat(1:bestNumFeatures)),'r');
axis([7650,8850,-1,80])
legend({'Control Group Avg.','Ovarian Cancer Group Avg.','Significant Features'})
xlabel(xAxisLabel); ylabel(yAxisLabel);

6-49
6 Mass Spectrometry and Bioanalytics

Alternative Statistical Learning Algorithms

There are many classification tools in MATLAB® that you can also use to analyze proteomic data.
Among them are support vector machines (fitcsvm), k-nearest neighbors (fitcknn), neural
networks (Deep Learning Toolbox™), and classification trees (fitctree). For feature selection, you
can also use sequential subset feature selection (sequentialfs) or optimize the randomized search
methods by using a genetic algorithm (Global Optimization Toolbox). For example, see “Genetic
Algorithm Search for Features in Mass Spectrometry Data” on page 6-71.

References
[1] Conrads, T P, V A Fusaro, S Ross, D Johann, V Rajapakse, B A Hitt, S M Steinberg, et al. “High-
Resolution Serum Proteomic Features for Ovarian Cancer Detection.” Endocrine-Related
Cancer, June 2004, 163–78.

[2] Lilien, Ryan H., Hany Farid, and Bruce R. Donald. “Probabilistic Disease Classification of
Expression-Dependent Proteomic Data from Mass Spectrometry of Human Serum.” Journal of
Computational Biology 10, no. 6 (December 2003): 925–46.

[3] Li, L., D. M. Umbach, P. Terry, and J. A. Taylor. “Application of the GA/KNN Method to SELDI
Proteomics Data.” Bioinformatics 20, no. 10 (July 1, 2004): 1638–40.

[4] Petricoin, Emanuel F, Ali M Ardekani, Ben A Hitt, Peter J Levine, Vincent A Fusaro, Seth M
Steinberg, Gordon B Mills, et al. “Use of Proteomic Patterns in Serum to Identify Ovarian
Cancer.” The Lancet 359, no. 9306 (February 2002): 572–77.

6-50
 Identifying Significant Features and Classifying Protein Profiles

See Also
msnorm | rankfeatures | classperf

Related Examples
• “Batch Processing of Spectra Using Sequential and Parallel Computing” on page 6-79

6-51
6 Mass Spectrometry and Bioanalytics

Differential Analysis of Complex Protein and Metabolite

Mixtures using Liquid Chromatography/Mass Spectrometry
(LC/MS)

This example shows how the SAMPLEALIGN function can correct nonlinear warping in the
chromatographic dimension of hyphenated mass spectrometry data sets without the need for full
identification of the sample compounds and/or the use of internal standards. By correcting such
warping between a pair (or set) of biologically related samples, differential analysis is enhanced and
can be automated.

Introduction

The use of complex peptide and metabolite mixtures in LC/MS requires label-free alignment
procedures. The analysis of this type of data requires searching for statistically significant differences
between biologically related data sets, without the need for a full identification of all the compounds
in the sample (either peptides/proteins or metabolites). Comparing compounds requires alignment in
two dimensions, the mass-charge dimension and the retention time dimension [1]. In the examples
“Preprocessing Raw Mass Spectrometry Data” on page 6-2 and “Visualizing and Preprocessing
Hyphenated Mass Spectrometry Data Sets for Metabolite and Protein/Peptide Profiling” on page 6-19,
you can learn how to use the MSALIGN, MSPALIGN, and SAMPLEALIGN functions to warp or calibrate
different type of anomalies in the mass/charge dimension. In this example, you will learn how to use
the SAMPLEALIGN function to also correct the nonlinear and unpredicted variations in the retention
time dimension.

While it is possible to implement alternative methods for aligning retention times, other approaches
typically require identification of compounds, which is not always feasible, or manual manipulations
that thwart attempts to automate for high throughput data analysis.

Data Set Description

This example uses two samples in PAe000153 and PAe000155 available from Peptide Atlas [2]. The
samples are LC-ESI-MS scans of four salt protein fractions from the saccharomyces cerevisiae each
containing more than 1000 peptides. Yeast samples were treated with different chemicals (glycine
and serine) in order to get two biologically diverse samples. Time alignment of these two data sets is
one of the most challenging cases reported in [3]. The data sets are not distributed with MATLAB®.
You must download the data sets to a local directory or your own repository. Alternatively, you can try
other data sets available in public databases for protein data, such as Sashimi Data Repository. If you
receive any errors related to memory or java heap space, try increasing your java heap space as
described here. LC/MS data analysis requires extended amounts of memory from the operating
system; if you receive "Out of memory" errors when running this example, try increasing the
virtual memory (or swap space) of your operating system or try setting the 3GB switch (32-bit
Windows® XP only). For details, see “Resolve “Out of Memory” Errors”.

Read and extract the lists of peaks from the XML files containing the intensity data for the sample
treated with Serine and the sample treated with Glycine.
ser = mzxmlread('005_1.mzXML')
[ps,ts] = mzxml2peaks(ser,'level',1);
gly = mzxmlread('005a.mzXML')
[pg,tg] = mzxml2peaks(gly,'level',1);

ser =

6-52
Differential Analysis of Complex Protein and Metabolite Mixtures using Liquid Chromatography/Mass Spectrometry (LC/MS)

struct with fields:

scan: [5610×1 struct]

mzXML: [1×1 struct]
index: [1×1 struct]

gly =

struct with fields:

scan: [5518×1 struct]

mzXML: [1×1 struct]
index: [1×1 struct]

Use the MSPPRESAMPLE function to resample the data sets while preserving the peak heights and
locations in the mass/charge dimension. Data sets are resampled to have both a common grid with
5,000 mass/charge values. A common grid is desirable for comparative visualization, and for
differential analysis.

[MZs,Ys] = msppresample(ps,5000);
[MZg,Yg] = msppresample(pg,5000);

Use the MSHEATMAP function to visualize both samples. When working with heat maps it is a common
technique to display the logarithm of the ion intensities, which enhances the dynamic range of the
colormap.

fh1 = msheatmap(MZs,ts,log(Ys),'resolution',0.15);
title('Serine Treatment')
fh2 = msheatmap(MZg,tg,log(Yg),'resolution',0.15);
title('Glycine Treatment')

6-53
6 Mass Spectrometry and Bioanalytics

6-54
Differential Analysis of Complex Protein and Metabolite Mixtures using Liquid Chromatography/Mass Spectrometry (LC/MS)

Detailed Inspection of the Misalignment Problems

Notice you can visualize the data sets separately; however, the time vectors have different size, and
therefore the heat maps have different number of rows (or Ys and Yg have different number of
columns). Moreover, the sampling rate is not constant and the shift between the time vectors is not
linear.

whos('Ys','Yg','ts','tg')
figure
plot(1:numel(ts),ts,1:numel(tg),tg)
legend('Serine','Glycine','Location','NorthWest')
title('Time Vectors')
xlabel('Spectrum Index')
ylabel('Retention Time (seconds)')

Name Size Bytes Class Attributes

Yg 5000x921 18420000 single

Ys 5000x937 18740000 single
tg 921x1 7368 double
ts 937x1 7496 double

6-55
6 Mass Spectrometry and Bioanalytics

To observe the same region of interest in both data sets, you need to calculate the appropriate row
indices in each matrix. For example, to inspect the peptide peaks in the 480-520 Da MZ range and
1550-1900 seconds retention time range, you need to find the closest matches for this range in the
time vectors and then zoom in each figure:

ind_ser = samplealign(ts,[1550;1900]);
figure(fh1);
axis([480 520 ind_ser'])

ind_gly = samplealign(tg,[1550;1900]);
figure(fh2);
axis([480 520 ind_gly'])

6-56
Differential Analysis of Complex Protein and Metabolite Mixtures using Liquid Chromatography/Mass Spectrometry (LC/MS)

6-57
6 Mass Spectrometry and Bioanalytics

Even though you zoomed in the same range, you can still observe that the top-right peptide in the
axes is shifted in the retention time dimension. In the sample treated with serine, the center of this
peak appears to occur at approximately 1595 seconds, while in the sample treated with glycine the
putative same peptide occurs at approximately 1630 seconds. This will prevent you from a accurate
comparative analysis, even if you resample the data sets to the same time vector. In addition to the
shift in the retention time, the data set seems to be improperly calibrated in the mass/charge
dimension, because the peaks do not have a compact shape in contiguous spectra.

Mass/Charge Calibration and Enhancement of the Matrices

Before correcting the retention time, you can enhance the samples using an approach similar to the
one described in the example “Visualizing and Preprocessing Hyphenated Mass Spectrometry Data
Sets for Metabolite and Protein/Peptide Profiling” on page 6-19. For brevity, we only display the
MATLAB code without any further explanation:
SF = @(x) 1-exp(-x./5e7); % scaling function
DF = @(R,S) sqrt((SF(R(:,2))-SF(S(:,2))).^2 + (R(:,1)-S(:,1)).^2);
CMZ = (315:.10:680)'; % Common Mass/Charge Vector with a finer grid

% Align peaks of the serine sample in the MZ direction

LAI = zeros(size(CMZ));
for i = 1:numel(ps)
if ~rem(i,250), fprintf(' %d...',i); end
[k,j] = samplealign([CMZ,LAI],double(ps{i}),'band',1.5,'gap',[0 2],'dist',DF);
LAI = LAI*.25;
LAI(k) = LAI(k) + ps{i}(j,2);
psa{i,1} = [CMZ(k) ps{i}(j,2)];

6-58
Differential Analysis of Complex Protein and Metabolite Mixtures using Liquid Chromatography/Mass Spectrometry (LC/MS)

end

% Align peaks of the glycine sample in the MZ direction

LAI = zeros(size(CMZ));
for i = 1:numel(pg)
if ~rem(i,250), fprintf(' %d...',i); end
[k,j] = samplealign([CMZ,LAI],double(pg{i}),'band',1.5,'gap',[0 2],'dist',DF);
LAI = LAI*.25;
LAI(k) = LAI(k) + pg{i}(j,2);
pga{i,1} = [CMZ(k) pg{i}(j,2)];
end

% Peak-preserving resample
[MZs,Ys] = msppresample(psa,5000);
[MZg,Yg] = msppresample(pga,5000);

% Gaussian Filtering
Gpulse = exp(-.5*(-10:10).^2)./sum(exp(-.05*(-10:10).^2));
Ysf = convn(Ys,Gpulse,'same');
Ygf = convn(Yg,Gpulse,'same');

% Visualization
fh3 = msheatmap(MZs,ts,log(Ysf),'resolution',0.15);
title('Serine Treatment (Enhanced)')
axis([480 520 ind_ser'])
fh4 = msheatmap(MZg,tg,log(Ygf),'resolution',0.15);
title('Glycine Treatment (Enhanced)')
axis([480 520 ind_gly'])

250... 500... 750... 250... 500... 750...

6-59
6 Mass Spectrometry and Bioanalytics

6-60
Differential Analysis of Complex Protein and Metabolite Mixtures using Liquid Chromatography/Mass Spectrometry (LC/MS)

Chromatographic Alignment

At this point, you have mass/charge calibrated and smoothed the two LC/MS data sets, but you are
still unable to perform a differential analysis because the data sets have a small misalignment along
the retention time axis.

You can use SAMPLEALIGN to correct the drift in the chromatographic domain. First, you should
inspect the data and look for the worst case shift, this helps you to estimate the BAND constraint. By
panning over both heat maps you can observe that common peptide peaks are not shifted more than
50 seconds. Use the input argument SHOWCONSTRAINTS to display the constraint space for the time
warping operation and assess if the Dynamic Programming (DP) algorithm can handle this problem
size. In this case you have less than 12,000 nodes. By omitting the output arguments, SAMPLEALIGN
displays only the constraints without running the DP algorithm. Also note that the input signals are
the filtered and enhanced data sets, but these have been upsampled to 5,000 MZ values, which are
very computationally demanding if you use all. Therefore, use the function MSPALIGN to obtain a
reduced list of mass/charge values (RMZ) indicating where the most intense peaks are, then use the
SAMPLEALIGN function also to find the indices of MZs (or MZg) that best match the reduced mass/
charge vector:

RMZ = mspalign([ps;pg])';
idx = samplealign(MZs,RMZ,'width',1); % with these input parameters this
% operation is equivalent to find the
% nearest neighbor for each RMZ in
% MZs.

samplealign([ts Ysf(idx,:)'],[tg Ygf(idx,:)'],'band',50,'showconstraints',true)

6-61
6 Mass Spectrometry and Bioanalytics

6-62
Differential Analysis of Complex Protein and Metabolite Mixtures using Liquid Chromatography/Mass Spectrometry (LC/MS)

SAMPLEALIGN uses the Euclidean distance as default to score matched pairs of samples. In LC/MS
data sets each sample corresponds to a spectrum at a given time, therefore, the cross-correlation
between each pair of matched spectra provides a better distance metric. SAMPLEALIGN allows you to
define your own metric to calculate the distance between spectra, it is also possible to envision a
metric that rewards more spectra pairs that match high ion intensity peaks rather than low ion
intensity noisy peaks. Use the input argument WEIGHT to remove the first column from the inputs,
which represents the retention time, so the scoring metric between spectra is based only on the ion
intensities.
cc = @(Xu,Yu) (mean(Xu.*Yu,2).^2)./mean(Xu.*Xu,2)./mean(Yu.*Yu,2);
ub = @(X) bsxfun(@minus,X,mean(X,2));
df = @(x,y) 1-cc(ub(x),ub(y));

[i,j] = samplealign([ts Ysf(idx,:)'],[tg Ygf(idx,:)'],'band',50,...

'distance',df ,'weight',[false true(1,numel(idx))]);

fh5 = figure;
plot(ts(i),tg(j)); grid
title('Warp Function')
xlabel('Retention Time in Data Set with Serine')
ylabel('Retention Time in Data Set with Glycine')

fh6 = figure; hold on

plot((ts(i)+tg(j))/2,ts(i)-tg(j))
title('Shift Function')
xlabel('Retention Time')
ylabel('Retention Time Shift between Data Sets')

6-63
6 Mass Spectrometry and Bioanalytics

6-64
Differential Analysis of Complex Protein and Metabolite Mixtures using Liquid Chromatography/Mass Spectrometry (LC/MS)

Because it is expected that the real shift function between both data sets is continuous, you can
smooth the observed shifts or regress a continuous model to create a warp model between the two
time axes. In this example, we simply regress the drifting to a polynomial function:

[p,p_struct,mu] = polyfit((ts(i)+tg(j))/2,ts(i)-tg(j),20);
sf = @(z) polyval(p,(z-mu(1))./mu(2));
figure(fh6)
plot(tg,sf(tg),'r')
legend('Observed shifts','Estimated shift curve','Location','SouthEast')

6-65
6 Mass Spectrometry and Bioanalytics

Comparative Analysis

To carry out a comparative analysis, resample the LC/MS data sets to a common time vector. When
resampling we use the estimated shift function to correct the retention time dimension. In this
example, each spectrum in both data sets is shifted half the distance of the shift function. In the case
of multiple alignment of data sets, it is possible to calculate the pairwise shift functions between all
data sets, and then apply the corrections to a common reference in such a way that the overall shifts
are minimized [4].

t = (max(min(tg),min(ts)):mean(diff(tg)):min(max(tg),max(ts)))';

% Visualization
fh7 = msheatmap(MZs,t,log(interp1q(ts,Ysf',t+sf(t)/2)'),'resolution',0.15);
title('Serine Treatment (Enhanced & Aligned)')
fh8 = msheatmap(MZg,t,log(interp1q(tg,Ygf',t-sf(t)/2)'),'resolution',0.15);
title('Glycine Treatment (Enhanced & Aligned)')

6-66
Differential Analysis of Complex Protein and Metabolite Mixtures using Liquid Chromatography/Mass Spectrometry (LC/MS)

6-67
6 Mass Spectrometry and Bioanalytics

To interactively or programmatically compare regions between two enhanced and time aligned data
sets, you can link the axes of two heat maps. Because the heat maps now use a regularly spaced time
vector, you can zoom in by using the AXIS function without having to search the appropriate row
indices of the matrices.

linkaxes(findobj([fh7 fh8],'Tag','MSHeatMap'))
axis([480 520 1550 1900])

6-68
Differential Analysis of Complex Protein and Metabolite Mixtures using Liquid Chromatography/Mass Spectrometry (LC/MS)

6-69
6 Mass Spectrometry and Bioanalytics

References

[1] Listgarten, J. and Emili, A., "Statistical and computational methods for comparative proteomic
profiling using liquid chromatography tandem mass spectrometry", Molecular and Cell Proteomics,
4(4):419-34, 2005.

[2] Desiere, F., et al., "The Peptide Atlas Project", Nucleic Acids Research, 34:D655-8, 2006.

[3] Prince, J.T. and Marcotte, E.M., "Chromatographic alignment of ESI-LC-MS proteomics data sets
by ordered bijective interpolated warping", Analytical Chemistry, 78(17):6140-52, 2006.

[4] Andrade L. and Manolakos E.S., "Automatic Estimation of Mobility Shift Coefficients in DNA
Chromatograms", Neural Networks for Signal Processing Proceedings, 91-100, 2003.

6-70
 Genetic Algorithm Search for Features in Mass Spectrometry Data

Genetic Algorithm Search for Features in Mass Spectrometry

Data

This example shows how to use the Global Optimization Toolbox with the Bioinformatics
Toolbox™ to optimize the search for features to classify mass spectrometry (SELDI) data.

Introduction

Genetic algorithms optimize search results for problems with large data sets. You can use the
MATLAB® genetic algorithm function to solve these problems in Bioinformatics. Genetic algorithms
have been applied to phylogenetic tree building, gene expression and mass spectrometry data
analysis, and many other areas of Bioinformatics that have large and computationally expensive
problems. This example searches for optimal features (peaks) in mass spectrometry data. We will look
for specific peaks in the data that distinguish cancer patients from control patients.

Global Optimization Toolbox

First familiarize yourself with the Global Optimization Toolbox. The documentation describes how a
genetic algorithm works and how to use it in MATLAB. To access the documentation, use the doc
command.

doc ga

Preprocess Mass Spectrometry Data

The original data in this example is from the FDA-NCI Clinical Proteomics Program Databank. It is a
collection of samples from 121 ovarian cancer patients and 95 control patients. For a detailed
description of this data set, see [1] and [2].

This example assumes that you already have the preprocessed data
OvarianCancerQAQCdataset.mat. However, if you do not have the data file, you can recreate by
following the steps in the example “Batch Processing of Spectra Using Sequential and Parallel
Computing” on page 6-79.

Alternatively, you can run the script msseqprocessing.m.

addpath(fullfile(matlabroot,'examples','bioinfo','main')) % Make sure the supporting files are on

type msseqprocessing

% MSSEQPROCESSING Script to create OvarianCancerQAQCdataset.mat (used in

% CANCERDETECTDEMO). Before running this file initialize the variable
% "repository" to the full path where you placed you mass-spectrometry
% files. For Example:
%
% repository = 'F:/MassSpecRepository/OvarianCD_PostQAQC/';
%
% or
%
% repository = '/home/username/MassSpecRepository/OvarianCD_PostQAQC/';
%
% The approximate time of execution is 18 minutes (Pentium 4, 4GHz). If you
% have the Parallel Computing Toolbox refer to BIODISTCOMPDEMO to see
% how you can speed this analysis up.

6-71
6 Mass Spectrometry and Bioanalytics

% Copyright 2003-2008 The MathWorks, Inc.

repositoryC = [repository 'Cancer/'];

repositoryN = [repository 'Normal/'];

filesCancer = dir([repositoryC '*.txt']);

NumberCancerDatasets = numel(filesCancer);
fprintf('Found %d Cancer mass-spectrograms.\n',NumberCancerDatasets)
filesNormal = dir([repositoryN '*.txt']);
NumberNormalDatasets = numel(filesNormal);
fprintf('Found %d Control mass-spectrograms.\n',NumberNormalDatasets)

files = [ strcat('Cancer/',{filesCancer.name}) ...

strcat('Normal/',{filesNormal.name})];
N = numel(files); % total number of files

fprintf('Total %d mass-spectrograms to process...\n',N)

[MZ,Y] = msbatchprocessing(repository,files);

disp('Finished; normalizing and saving to OvarianCancerQAQCdataset.mat.')

Y = msnorm(MZ,Y,'QUANTILE',0.5,'LIMITS',[3500 11000],'MAX',50);

grp = [repmat({'Cancer'},size(filesCancer));...
repmat({'Normal'},size(filesNormal))];

save OvarianCancerQAQCdataset.mat Y MZ grp

Load Mass Spectrometry Data into MATLAB®

Once you have the preprocessed data, you can load it into MATLAB.

load OvarianCancerQAQCdataset
whos

Name Size Bytes Class Attributes

MZ 15000x1 120000 double

Y 15000x216 25920000 double
grp 216x1 25056 cell

There are three variables: MZ, Y, grp. MZ is the mass/charge vector, Y is the intensity values for all
216 patients (control and cancer), and grp holds the index information as to which of these samples
represent cancer patients and which ones represent normal patients. To visualize this data, see the
example “Identifying Significant Features and Classifying Protein Profiles” on page 6-38.

Initialize the variables used in the example.

[numPoints, numSamples] = size(Y); % total number of samples and data points

id = grp2idx(grp); % ground truth: Cancer=1, Control=2

Create a Fitness Function for the Genetic Algorithm

A genetic algorithm requires an objective function, also known as the fitness function, which
describes the phenomenon that we want to optimize. In this example, the genetic algorithm

6-72
 Genetic Algorithm Search for Features in Mass Spectrometry Data

machinery tests small subsets of M/Z values using the fitness function and then determines which
M/Z values get passed on to or removed from each subsequent generation. The fitness function
biogafit is passed to the genetic algorithm solver using a function handle. In this example, biogafit
maximizes the separability of two classes by using a linear combination of 1) the a-posteriori
probability and 2) the empirical error rate of a linear classifier (classify). You can create your own
fitness function to try different classifiers or alternative methods for assessing the performance of the
classifiers.

type biogafit

function classPerformance = biogafit(thePopulation,Y,id)

%BIOGAFIT The fitness function for BIOGAMSDEMO
%
% This function uses the classify function to measure how well mass
% spectrometry data is grouped using certain masses. The input argument
% thePopulation is a vector of row indices from the mass spectrometry
% data Y. Classification performance is a linear combination of the error
% rate and the posteriori probability of the classifier.

% Copyright 2003-2013 The MathWorks, Inc.

thePopulation = round(thePopulation);
try
[c,~,p] = classify(Y(thePopulation,:)',Y(thePopulation,:)',double(id),'linear');
cp = classperf(id,c);
classPerformance = 100*cp.ErrorRate + 1 - mean(max(p,[],2));
catch
% In case pooled covariance matrix is not positive definite we try a
% naive-Bayes classifier:
try
[c,~,p] = classify(Y(thePopulation,:)',Y(thePopulation,:)',double(id),'diaglinear');
cp = classperf(id,c);
classPerformance = 100*cp.ErrorRate + 1 - mean(max(p,[],2));
catch
classPerformance = Inf;
end
end

Create an Initial Population

Users can change how the optimization is performed by the genetic algorithm by creating custom
functions for crossover, fitness scaling, mutation, selection, and population creation. In this example
you will use the biogacreate function written for this example to create initial random data points
from the mass spectrometry data. The function header requires specific input parameters as specified
by the GA documentation. There is a default creation function in the toolbox for creating initial
populations of data points.

type biogacreate

function pop = biogacreate(GenomeLength,~,options,Y,id)

%BIOGACREATE Population creation function for MSGADEMO
%

6-73
6 Mass Spectrometry and Bioanalytics

% This function creates a population matrix with dimensions of

% options.PopulationSize rows by the number of independent variables
% (GenomeLength) columns. These values are integers that correspond to
% randomly selected rows of the mass spectrometry data Y. Each row of the
% population matrix is a random sample of row indices of the mass spec
% data.
% Note: This function's input arguments are required by the GA function.
% See GA documentation for further detail.

% Copyright 2005-2013 The MathWorks, Inc.

pop = zeros(options.PopulationSize,GenomeLength);
npop = numel(pop);
ranked_features = rankfeatures(Y,id,'NumberOfIndices',npop,'NWeighting',.5);
pop(:) = randsample(ranked_features,npop);

Set Genetic Algorithm Options

The GA function uses an options structure to hold the algorithm parameters that it uses when
performing a minimization with a genetic algorithm. The optimoptions function will create this
options structure. For the purposes of this example, the genetic algorithm will run only for 50
generations. However, you may set 'Generations' to a larger value.

options = optimoptions('ga','CreationFcn',{@biogacreate,Y,id},...
'PopulationSize',120,...
'Generations',50,...
'Display','iter')

options =

ga options:

Set properties:
CreationFcn: {1x3 cell}
Display: 'iter'
MaxGenerations: 50
PopulationSize: 120

Default properties:
ConstraintTolerance: 1.0000e-03
CrossoverFcn: []
CrossoverFraction: 0.8000
EliteCount: '0.05*PopulationSize'
FitnessLimit: -Inf
FitnessScalingFcn: @fitscalingrank
FunctionTolerance: 1.0000e-06
HybridFcn: []
InitialPopulationMatrix: []
InitialPopulationRange: []
InitialScoresMatrix: []
MaxStallGenerations: 50
MaxStallTime: Inf
MaxTime: Inf
MutationFcn: []
NonlinearConstraintAlgorithm: 'auglag'
OutputFcn: []
PlotFcn: []

6-74
 Genetic Algorithm Search for Features in Mass Spectrometry Data

PopulationType: 'doubleVector'
SelectionFcn: []
UseParallel: 0
UseVectorized: 0

Run GA to Find 20 Discriminative Features

Use ga to start the genetic algorithm function. 100 groups of 20 datapoints each will evolve over 50
generations. Selection, crossover, and mutation events generate a new population in every
generation.

% initialize the random generators to the same state used to generate the
% published example
rng('default')
nVars = 12; % set the number of desired features
FitnessFcn = {@biogafit,Y,id}; % set the fitness function
feat = ga(FitnessFcn,nVars,options); % call the Genetic Algorithm

feat = round(feat);
Significant_Masses = MZ(feat)

cp = classperf(classify(Y(feat,:)',Y(feat,:)',id),id);
cp.CorrectRate

Single objective optimization:

12 Variable(s)

Options:
CreationFcn: @biogacreate
CrossoverFcn: @crossoverscattered
SelectionFcn: @selectionstochunif
MutationFcn: @mutationgaussian

Best Mean Stall

Generation Func-count f(x) f(x) Generations
1 240 2.827 8.928 0
2 354 2.827 8.718 1
3 468 0.9663 8.001 0
4 582 0.9516 7.249 0
5 696 0.9516 6.903 1
6 810 0.4926 6.804 0
7 924 0.4926 6.301 1
8 1038 0.02443 5.215 0
9 1152 0.02443 4.77 1
10 1266 0.02101 4.084 0
11 1380 0.02101 3.792 1
12 1494 0.01854 3.437 0
13 1608 0.01606 3.44 0
14 1722 0.01372 2.768 0
15 1836 0.01218 2.74 0
16 1950 0.01204 2.471 0
17 2064 0.01204 2.649 1
18 2178 0.01189 2.326 0
19 2292 0.01189 2.003 0
20 2406 0.0118 2.341 0
21 2520 0.01099 1.714 0

6-75
6 Mass Spectrometry and Bioanalytics

22 2634 0.01094 1.828 0

23 2748 0.01094 1.94 1
24 2862 0.01094 2.285 2
25 2976 0.009843 2.026 0
26 3090 0.009843 1.899 1
27 3204 0.009183 1.802 0
28 3318 0.007877 1.5 0
29 3432 0.007788 1.793 0
30 3546 0.007788 1.756 1

Best Mean Stall

Generation Func-count f(x) f(x) Generations
31 3660 0.007091 1.719 0
32 3774 0.006982 1.598 0
33 3888 0.006982 1.269 1
34 4002 0.006732 1.279 0
35 4116 0.005008 1.229 0
36 4230 0.004325 1.179 0
37 4344 0.004325 1.534 1
38 4458 0.003982 1.15 0
39 4572 0.003982 0.9602 1
40 4686 0.003982 0.8547 2
41 4800 0.003891 0.9083 0
42 4914 0.003683 0.7409 0
43 5028 0.003683 0.516 1
44 5142 0.003364 0.5153 0
45 5256 0.003172 0.4218 0
46 5370 0.003172 0.3783 1
47 5484 0.002997 0.1883 0
48 5598 0.002675 0.1297 0
49 5712 0.002611 0.04382 0
50 5826 0.002519 0.007859 0
Optimization terminated: maximum number of generations exceeded.

Significant_Masses =

1.0e+03 *

7.6861
7.9234
8.9834
8.6171
7.1808
7.3057
8.1132
8.5241
7.0527
7.7600
7.7442
7.7245

ans =

6-76
 Genetic Algorithm Search for Features in Mass Spectrometry Data

Display the Features that are Discriminatory

To visualize which features have been selected by the genetic algorithm, the data is plotted with peak
positions marked with red vertical lines.

xAxisLabel = 'Mass/Charge (M/Z)'; % x label for plots

yAxisLabel = 'Relative Ion Intensity'; % y label for plots
figure; hold on;
hC = plot(MZ,Y(:,1:15) ,'b');
hN = plot(MZ,Y(:,141:155),'g');
hG = plot(MZ(feat(ceil((1:3*nVars )/3))), repmat([0 100 NaN],1,nVars),'r');
xlabel(xAxisLabel); ylabel(yAxisLabel);
axis([1900 12000 -1 40]);
legend([hN(1),hC(1),hG(1)],{'Control','Ovarian Cancer', 'Discriminative Features'}, ...
'Location', 'NorthWest');
title('MS Data with Discriminative Features found by Genetic Algorithm');

Observe the interesting peak around 8100 Da., which seems to be shifted to the right on healthy
samples.

axis([8082 8128 -.5 4])

6-77
6 Mass Spectrometry and Bioanalytics

References
[1] Conrads, T P, V A Fusaro, S Ross, D Johann, V Rajapakse, B A Hitt, S M Steinberg, et al. “High-
Resolution Serum Proteomic Features for Ovarian Cancer Detection.” Endocrine-Related
Cancer, June 2004, 163–78.

[2] Petricoin, Emanuel F, Ali M Ardekani, Ben A Hitt, Peter J Levine, Vincent A Fusaro, Seth M
Steinberg, Gordon B Mills, et al. “Use of Proteomic Patterns in Serum to Identify Ovarian
Cancer.” The Lancet 359, no. 9306 (February 2002): 572–77.

See Also
msnorm

Related Examples
• “Batch Processing of Spectra Using Sequential and Parallel Computing” on page 6-79
• “Identifying Significant Features and Classifying Protein Profiles” on page 6-38

6-78
 Batch Processing of Spectra Using Sequential and Parallel Computing

Batch Processing of Spectra Using Sequential and Parallel

Computing

This example shows how you can use a single computer, a multicore computer, or a cluster of
computers to preprocess a large set of mass spectrometry signals. Note: Parallel Computing
Toolbox™ and MATLAB® Parallel Server™ are required for the last part of this example.

Introduction

This example shows the required steps to set up a batch operation over a group of mass spectra
contained in one or more directories. You can achieve this sequentially, or in parallel using either a
multicore computer or a cluster of computers. Batch processing adapts to the single-program
multiple-data (SPMD) parallel computing model, and it is best suited for Parallel Computing
Toolbox™ and MATLAB® Parallel Server™.

The signals to preprocess come from protein surface-enhanced laser desorption/ionization-time of

flight (SELDI-TOF) mass spectra. The data in this example are from the FDA-NCI Clinical Proteomics
Program Databank. In particular, the example uses the high-resolution ovarian cancer data set that
was generated using the WCX2 protein array. For a detailed description of this data set, see [1] and
[2].

Setting the Repository for the Data

This example assumes that you have downloaded and uncompressed the data sets into your
repository. Ideally, you should place the data sets in a network drive. If the workers all have access to
the same drives on the network, they can access needed data that reside on these shared resources.
This is the preferred method for sharing data, as it minimizes network traffic.

First, get the name and full path to the data repository. Two strings are defined: the path from the
local computer to the repository and the path required by the cluster computers to access the same
directory. Change both accordingly to your network configuration.

local_repository = 'C:/Examples/MassSpecRepository/OvarianCD_PostQAQC/';
worker_repository = 'L:/Examples/MassSpecRepository/OvarianCD_PostQAQC/';

For this particular example, the files are stored in two subdirectories: 'Normal' and 'Cancer'. You can
create lists containing the files to process using the DIR command,

cancerFiles = dir([local_repository 'Cancer/*.txt'])

normalFiles = dir([local_repository 'Normal/*.txt'])

cancerFiles =

121×1 struct array with fields:

name
folder
date
bytes
isdir
datenum

6-79
6 Mass Spectrometry and Bioanalytics

normalFiles =

95×1 struct array with fields:

name
folder
date
bytes
isdir
datenum

and put them into a single variable:

files = [ strcat('Cancer/',{cancerFiles.name}) ...

strcat('Normal/',{normalFiles.name})];
N = numel(files) % total number of files

N =

216

Sequential Batch Processing

Before attempting to process all the files in parallel, you need to test your algorithms locally with a
for loop.

Write a function with the sequential set of instructions that need to be applied to every data set. The
input arguments are the path to the data (depending on how the machine that will actually do the
work sees them) and the list of files to process. The output arguments are the preprocessed signals
and the M/Z vector. Because the M/Z vector is the same for every spectrogram after the
preprocessing, you need to store it only once. For example:

type msbatchprocessing

function [MZ,Y] = msbatchprocessing(repository,files)

% MSBATCHPROCESSING Example function for BIODISTCOMPDEMO
%
% [MZ,Y] = MSBATCHPROCESSING(REPOSITORY,FILES) Preprocesses the
% spectrogram in files FILES and returns the mass/charge (MZ) and ion
% intensities (Y) vectors.
%
% Hard-coded parameters in the preprocessing steps have been adjusted to
% deal with the high-resolution spectrograms of the example.

% Copyright 2004-2012 The MathWorks, Inc.

K = numel(files);
Y = zeros(15000,K); % need to preset the size of Y for memory performance
MZ = zeros(15000,1);

parfor k = 1:K

file = [repository files{k}];

6-80
 Batch Processing of Spectra Using Sequential and Parallel Computing

% read the two-column text file with mass-charge and intensity values
fid = fopen(file,'r');
ftext = textscan(fid, '%f%f');
fclose(fid);
signal = ftext{1};
intensity = ftext{2};

% resample the signal to 15000 points between 2000 and 11900

mzout = (sqrt(2000)+(0:(15000-1))'*diff(sqrt([2000,11900]))/15000).^2;
[mz,YR] = msresample(signal,intensity,mzout);

% align the spectrograms to two good reference peaks

P = [3883.766 7766.166];
YA = msalign(mz,YR,P,'WIDTH',2);

% estimate and adjust the background

YB = msbackadj(mz,YA,'STEP',50,'WINDOW',50);

% reduce the noise using a nonparametric filter

Y(:,k) = mslowess(mz,YB,'SPAN',5);

% the mass/charge vector is the same for all spectra after the resample
if k==1
MZ(:,k) = mz;
end

end

Note inside the function MSBATCHPROCESSING the intentional use of PARFOR instead of FOR. Batch
processing is generally implemented by tasks that are independent between iterations. In such case,
the statement FOR can indifferently be changed to PARFOR, creating a sequence of MATLAB®
statements (or program) that can run seamlessly on a sequential computer, a multicore computer, or
a cluster of computers, without having to modify it. In this case, the loop executes sequentially
because you have not created a Parallel Pool (assuming that in the Parallel Computing Toolbox™
Preferences the checkbox for automatically creating a Parallel Pool is not checked, otherwise
MATLAB will execute in parallel anyways). For the example purposes, only 20 spectrograms are
preprocessed and stored in the Y matrix. You can measure the amount of time MATLAB® takes to
complete the loop using the TIC and TOC commands.

tic
repository = local_repository;
K = 20; % change to N to do all

[MZ,Y] = msbatchprocessing(repository,files(1:K));

disp(sprintf('Sequential time for %d spectrograms: %f seconds',K,toc))

Sequential time for 20 spectrograms: 7.725275 seconds

Parallel Batch Processing with Multicore Computers

If you have Parallel Computing Toolbox™, you can use local workers to parallelize the loop iterations.
For example, if your local machine has four-cores, you can start a Parallel Pool with four workers
using the default 'local' cluster profile:

POOL = parpool('local',4);

6-81
6 Mass Spectrometry and Bioanalytics

tic
repository = local_repository;
K = 20; % change to N to do all

[MZ,Y] = msbatchprocessing(repository,files(1:K));

disp(sprintf('Parallel time with four local workers for %d spectrograms: %f seconds',K,toc))

Starting parallel pool (parpool) using the 'local' profile ...

Connected to the parallel pool (number of workers: 4).
Parallel time with four local workers for 20 spectrograms: 3.549382 seconds

Stop the local worker pool:

delete(POOL)

Parallel Batch Processing with Distributed Computing

If you have Parallel Computing Toolbox™ and MATLAB® Parallel Server™ you can also distribute the
loop iterations to a larger number of computers. In this example, the cluster profile
'compbio_config_01' links to 6 workers. For information about setting up and selecting parallel
configurations, see "Cluster Profiles and Computation Scaling" in the Parallel Computing Toolbox™
documentation.

Note that if you have written your own batch processing function, you should include it in the
respective cluster profile by using the Cluster Profile Manager. This will ensure that MATLAB®
properly transmits your new function to the workers. You access the Cluster Profile Manager using
the Parallel pull-down menu on the MATLAB® desktop.

POOL = parpool('compbio_config_01',6);

tic
repository = worker_repository;
K = 20; % change to N to do all

[MZ,Y] = msbatchprocessing(repository,files(1:K));

disp(sprintf('Parallel time with 6 remote workers for %d spectrograms: %f seconds',K,toc))

Starting parallel pool (parpool) using the 'compbio_config_01' profile ...

Connected to the parallel pool (number of workers: 6).
Parallel time with 6 remote workers for 20 spectrograms: 3.541660 seconds

Stop the cluster pool:

delete(POOL)

Asynchronous Parallel Batch Processing

The execution schemes described above all operate synchronously, that is, they block the MATLAB®
command line until their execution is completed. If you want to start a batch process job and get
access to the command line while the computations run asynchronously (async), you can manually
distribute the parallel tasks and collect the results later. This example uses the same cluster profile as
before.

Create one job with one task (MSBATCHPROCESSING). The task runs on one of the workers, and its
internal PARFOR loop is distributed among all the available workers in the parallel configuration.

6-82
 Batch Processing of Spectra Using Sequential and Parallel Computing

Note that if N (number of spectrograms) is much larger than the number of available workers in your
parallel configuration, Parallel Computing Toolbox™ automatically balances the work load, even if
you have a heterogeneous cluster.

tic % start the clock

repository = worker_repository;
K = N; % do all spectrograms
CLUSTER = parcluster('compbio_config_01');
JOB = createCommunicatingJob(CLUSTER,'NumWorkersRange',[6 6]);
TASK = createTask(JOB,@msbatchprocessing,2,{repository,files(1:K)});

submit(JOB)

When the job is submitted, your local MATLAB® prompt returns immediately. Your parallel job starts
once the parallel resources become available. Meanwhile, you can monitor your parallel job by
inspecting the TASK or JOB objects. Use the WAIT method to programmatically wait until your task
finishes:

wait(TASK)
TASK.OutputArguments

ans =

1×2 cell array

{15000×1 double} {15000×216 double}

MZ = TASK.OutputArguments{1};
Y = TASK.OutputArguments{2};
destroy(JOB) % done retrieving the results
disp(sprintf('Parallel time (asynchronous) with 6 remote workers for %d spectrograms: %f seconds'

Parallel time (asynchronous) with 6 remote workers for 216 spectrograms: 68.368132 seconds

Postprocessing

After collecting all the data, you can use it locally. For example, you can apply group normalization:

Y = msnorm(MZ,Y,'QUANTILE',0.5,'LIMITS',[3500 11000],'MAX',50);

Create a grouping vector with the type for each spectrogram as well as indexing vectors. This
"labelling" will aid to perform further analysis on the data set.

grp = [repmat({'Cancer'},size(cancerFiles));...
repmat({'Normal'},size(normalFiles))];
cancerIdx = find(strcmp(grp,'Cancer'));
numel(cancerIdx) % number of files in the "Cancer" subdirectory

ans =

121

normalIdx = find(strcmp(grp,'Normal'));
numel(normalIdx) % number of files in the "Normal" subdirectory

6-83
6 Mass Spectrometry and Bioanalytics

ans =

95

Once the data is labelled, you can display some spectrograms of each class using a different color
(the first five of each group in this example).

h = plot(MZ,Y(:,cancerIdx(1:5)),'b',MZ,Y(:,normalIdx(1:5)),'r');
axis([7650 8200 -2 50])
xlabel('Mass/Charge (M/Z)');ylabel('Relative Intensity')
legend(h([1 end]),{'Ovarian Cancer','Control'})
title('Region of the pre-processed spectrograms')

Save the preprocessed data set, because it will be used in the examples “Identifying Significant
Features and Classifying Protein Profiles” on page 6-38 and “Genetic Algorithm Search for Features
in Mass Spectrometry Data” on page 6-71.

save OvarianCancerQAQCdataset.mat Y MZ grp

6-84
 Batch Processing of Spectra Using Sequential and Parallel Computing

Disclaimer

TIC - TOC timing is presented here as an example. The sequential and parallel execution time will
vary depending on the hardware you use.

References
[1] Conrads, T P, V A Fusaro, S Ross, D Johann, V Rajapakse, B A Hitt, S M Steinberg, et al. “High-
Resolution Serum Proteomic Features for Ovarian Cancer Detection.” Endocrine-Related
Cancer, June 2004, 163–78.

[2] Petricoin, Emanuel F, Ali M Ardekani, Ben A Hitt, Peter J Levine, Vincent A Fusaro, Seth M
Steinberg, Gordon B Mills, et al. “Use of Proteomic Patterns in Serum to Identify Ovarian
Cancer.” The Lancet 359, no. 9306 (February 2002): 572–77.

See Also
msnorm | msresample | msbackadj | mslowess | msalign

6-85