A primer to phylogenetic analysis using

the PHYLIP package

Jarno Tuimala
Fifth Edition
All rights reserved. The PDF version of this book or parts of it can be used in Finnish
universities as course material, provided that this copyright notice is included. However, this
publication may not be sold or included as part of other publications without permission of
the publisher.

© The author and

CSC – Scientific Computing Ltd.
2006

ISBN 952-5520-02-1

2
Index
Index .............................................................................................................................................................. 3
Preface ........................................................................................................................................................... 4
Introduction .................................................................................................................................................. 5
What is PHYLIP? ................................................................................................................................... 5
Installation .............................................................................................................................................. 5
User interface.......................................................................................................................................... 6
Getting started – datafiles and programs................................................................................................... 6
Always keep records............................................................................................................................... 6
Sequence alignment ................................................................................................................................ 7
Font files ................................................................................................................................................. 8
Running PHYLIP programs ................................................................................................................... 8
Essential programs.................................................................................................................................. 9
Quick start................................................................................................................................................... 11
Distance methods.................................................................................................................................. 11
Tree drawing......................................................................................................................................... 13
Amino acid sequences .......................................................................................................................... 14
Basic analyses in more detail ..................................................................................................................... 15
Distance methods.................................................................................................................................. 15
Parsimony methods............................................................................................................................... 20
Maximum likelihood methods .............................................................................................................. 25
Resampling procedure .......................................................................................................................... 29
Drawing the tree ................................................................................................................................... 33
Advanced topics .......................................................................................................................................... 35
User trees .............................................................................................................................................. 35
Estimating the transition/transversion ratio .......................................................................................... 37
Estimating base frequencies ................................................................................................................. 38
Testing molecular clock........................................................................................................................ 38
Inferring ancient states of sequence sites.............................................................................................. 39
Statistical tests of trees.......................................................................................................................... 41
LogDet-distance.................................................................................................................................... 42
Computing topological distances between trees ................................................................................... 42
Weighting ............................................................................................................................................. 44
Dnaml, HMM, gamma distribution and rate heterogeneity .................................................................. 46
Multiple outgroups ............................................................................................................................... 47
Error messages...................................................................................................................................... 47
Scripting ............................................................................................................................................... 49
Recommendations....................................................................................................................................... 51
Some pragmatic warnings..................................................................................................................... 51
PHYLIP programs ..................................................................................................................................... 53
Flow charts .................................................................................................................................................. 54

3
Preface

The purpose of this tutorial is to demonstrate how to use PHYLIP, a collection of

phylogenetic analysis software, and some of the options that are available. This tutorial is not
intented to be a course in phylogenetics, although some phylogenetic concepts will be
discussed briefly. There are other books available which cover the theoretical sides of the
phylogenetic analysis, but the actual data analysis work is less well covered.

Here we will mostly deal with molecular sequence data analysis in the current PHYLIP
version 3.66.

Commands that the user should type are written with 12 pt Courier font. File names are
written with 12 pt Courier New font. Output from the programs is represented with
10 pt Courier font.

I want to thank Joe Felsenstein for his extensive and constructive comments on this text. He
has continued to offer feedback for several years, and his efforts are acknowledged with great
appreciation.

11th July 2006

Jarno Tuimala

4
Introduction

What is PHYLIP?

PHYLIP is a comprehensive phylogenetic analysis package created by Joseph Felsenstein at

the University of Washington. This package can do many of the phylogenetic analyses
available in the literature today. Methods that are available in the package include parsimony,
distance matrix, and likelihood methods, including bootstrapping and consensus trees. Data
types that can be handled include molecular sequences, gene frequencies, restriction sites,
distance matrices, and 0/1 (binary) discrete characters.

Installation

PHYLIP is freely available from:

http://evolution.genetics.washington.edu/phylip.html.

It ships with a comprehensive manual covering the usage of different programs. Manual is
written in HTML-format, so you can read it using a web browser.

If you’re using a Windows machine, installation is easy. Download the three zip-files
(phylip.exe, phylipwx.exe, phylipwy.exe), and extract them to a preferred
folder. The subfolder exe contains all the programs. Manual can be found from the subfolder
doc.

For Macintosh OS X you may download the packaged disk image (Phylip3.66.dmg). It is
compressed, so you need to expand it, and copy the resulting folder to a desired location.
Alternatively, you may compile the programs from their sources as outlined in the UNIX
installation below. There are source codes and ready made compilations available for older
Macintosh systems, Mac OS 8 or 9, also.

Installation for UNIX systems is also quite straight-forward. These instruction apply for
RedHat-based Linux systems. Installation on main frame can require tweaking of the
Makefile. Download the source code and documentation package (phylip-3.66.tar.gz) into a
suitable folder. Unzip the package with gzip utility (gzip –d phylip-3.66.tar.gz)
and expand the tar ball (tar xvf phylip-3.66.tar). Move to the newly formed
folder containing the source codes (cd phylip3.6/src). The folder contains a file called
Makefile. Installation of the PHYLIP programs is done simply by typing make install.
The default Makefile usually works fine.

The draw programs (Drawgram, and Drawtree) need an installation of the X Windows
development environment (Athena Widgets), and without it you’ll get some error messages
during installation, e.g.,
ld32: WARNING 84 : /usr/lib32/libX11.so is not used for resolving any symbol.
ld32: WARNING 84 : /usr/lib32/libXaw.so is not used for resolving any symbol.
ld32: FATAL 12 : Expecting n32 objects: draw.o is of unknown type.

5
Thus, you might need to change the path to the Widgets on the following lines in the
Makefile:

# for Linux with X Windows development packages installed

# or for MacOS X with X Windows installed
DFLAGS = $(CFLAGS) -DX -I/usr/X11R6/include

# if the Xlib library for the X windowing system is somewhere

# unexpected, you may have to change the path /usr/X11R6/lib in this one
#
# For gcc for Linux with X windows development packages installed
# or for MacOS X with X windows installed
DLIBS= -L/usr/X11R6/lib/ -lX11 -lXaw -lXt

After a successful installation, two new directories appear under the phylip3.66 folder. These
contain documentation (doc) and executables (exe).

There is more advice on installing the PHYLIP package in the UNIX environment in the
manual, so read it meticulously before proceeding. If everything else fails, you might want to
check the site http://www.biolinux.org/phylip.html for ready compilations.
However, a fresh compilation on your machine might be more up-to-date.

User interface

The programs are controlled through a menu, which asks the users which options they want
to set, and allows them to start the computation. The data are read into the program from a
text file, which the user can prepare using any word processor or text editor (but it is
important that this text file not be in the special format of that word processor - it should
instead be in flat ASCII or Text Only format). Some sequence alignment programs, like
ClustalX and T-Coffee, can write data files in the PHYLIP format.

Most of the programs look for the data in a file called infile. If they do not find this file
they then ask the user to type in the file name of the data file. Output is written onto special
files with names like outfile and outtree. Trees written onto outtree are in the
Newick format, an informal standard agreed to in 1986 by authors of a number of major
phylogeny packages (Felsenstein, PHYLIP documentation).

Getting started – datafiles and programs

Always keep records

It is very important to keep record of lab-procedures you have done, but it is even more so
with computer analyses. You might easily get confused with many many result files,
especially, if you have not given them informative names. And, the computer can crash, or
the hard-drive may become corrupt, and you can lose your work. After such incidents it is
easier to recover the work you have done, if you have kept a good analysis record.

So, always keep records. It indicates you have been working.

6
Sequence alignment

PHYLIP programs read the aligned sequences in PHYLIP-format. Usually you can recognize
the files in this format from the .phy extension associated with the files. Aligned sequences in
the suitable format can be produced, e.g., with the program ClustalX, which is freely
available from:

http://www-igbmc.u-strasbg.fr/BioInfo/ClustalX/Top.html.

Just be sure that you save the aligned sequences in PHYLIP-format.

If you need to edit the alignment (with text-editor) or to do some analyses in the other
programs that do not read PHYLIP-format, save the alignment also in the .aln format
(Clustal-format). The editing of the Clustal alignment format is easier than the editing of
PHYLIP-format, and Clustal will readily read in the .aln format, if you later need to convert
the edited sequences into some other format.

Any multiple sequence alignment can also be manually reformatted with a text editor. The
format requirements for PHYLIP are rather stringent, and any deviation will result in a
program that hangs, usually with the error message Unable to allocate memory.

The file must confirm to the following (Felsenstein, PHYLIP documentation):

1. The file begins with the information about the number of sequences and the number of
nucleotides or amino acids in the alignment.
2. The sequence names must be exactly 10 characters long. Spaces can be added to the end of
shorter names to make them this length. Do not use Tab characters for this.
3. Gaps must be indicated by - .
4. Missing data or missing information (no sequence) is indicated by ?.
5. Spaces between the alignment blocks are allowed. This normally makes the alignment
more readable. Spaces are usually inserted into the alignment every 10 bases or amino acids.
6. Blanks will be ignored, and so will numerical digits. This allows GENBANK and EMBL
sequence entries to be read with minimum editing.

Example of the formatted sequences:

5 100
Rabbit ?????????? ?????????C CAATCTACAC ACGGG-GTAG GGATTACATA
Human AGCCACACCC TAGGGTTGGC CAATCTACTC CCAGGAGCAG GGAGGGCAGG
Opossum AGCCACACCC CAACCTTAGC CAATAGACAT CCAGAAGCCC AAAAGGCAAG
Chicken GCCCGGGGAA GAGGAGGGGC CCGGCGG-AG GCGATAAAAG TGGGGACACA
Frog GGATGGAGAA TTAGAGCACT TGTTCTTTTT GCAGAAGCTC AGAATAAACG

TTTGGATGGT AG---GATAT GGGCCTACCA TGGCGTTAAC GGGT-AACGY

TTTCGACGGT AA---GGTAT TGGCTTACCG TGGCAATGAC AGGT-GACGG
TTTCGACGGT AA---GGTAT TGGCTTACCG TGGCAATGAC AGGT-GACGY
TTTCGACGGT AA---GGTAT TGGCTTACCG TGGCAATGAC AGGT-GACGG
TTTCGATGGT AA---GGTAT TGGCTTACCG TGGCAATGAC AGGT-GACGG

Possible ambiguities (such as N, Y or R nucleotides) are also handled correctly, and do not
cause trouble.

7
Font files

In order to be able to use the tree-drawing tools, the font files need to be in the same folder as
the Drawtree or Drawgram program(s). If you are using PHYLIP on a PC from the same
folder it was installed in, you should not encounter any troubles. However, this is not strictly
necessary, just remember to copy the font files with tree drawing programs to the same
folder. Or, better still, copy and rename your favorite fontfile as fontfile and keep only it
with the tree drawing programs. There are six different fonts available:

font1 simple sans-serif Roman

font2 medium quality sans-serif Roman
font3 high quality sans-serif Roman
font4 medium quality sans-serif Italic
font5 high quality sans-serif Italic
font6 Russian Cyrillic

Running PHYLIP programs

The programs are used in a sequential way. The output from the first program is used as an
input in the next program. The trick is to know how to use the programs in suitable
combinations. See the flow charts in the end of this book for some suggestions.

In Windows, the PHYLIP programs can be invoked by double-clicking on the icon or by

typing the name of the program on the command line. It is advisable to use programs from
the command line, because then you will be better able to see, e.g., the error messages that
might appear. In NT-line Windows versions (NT, 2000 and XP) the DOS prompt, i.e.,
command line, can be invoked from Start -> All Programs -> Accessories -> Command
Prompt.

Most PHYLIP programs run in the same way. The input for a program is taken from a file
called infile - if the program does not find this file it then asks the user to type in the file
name of the data file. The results are written in a file called outfile. Some programs may
write both outfile and a file called outtree or plotfile.

Because most of the programs use the default names for the input and output files, you need
to be sure to rename the files you want to save before proceeding to further analysis.
Otherwise you risk losing your results. For example, you get a distance matrix (outfile)
from the program Dnadist, but you want to try different settings for the matrix calculations.
Then, before doing the matrix calculation again, rename outfile to Dnadist_out_F84
or something similar, so that you can tell different analysis results apart after you have ceased
to work.

8
Essential programs

Here is a list of the programs that can be used for the molecular sequence data analysis. The
programs are divided into the method categories. The choice of the correct analysis method is
left for the user.

Distance methods

These programs are intended to be used sequentially. First a distance matrix is calculated by
Dnadist or Protdist program from the multiple sequence alignment. The matrix is then
transformed into a tree by Fitch, Kitsch or Neighbor program. Programs Dnadist and Protdist
create a file outfile. Before running Fitch, Kitsch ot Neighbor, outfile should be
renamed, either as infile ot with another file name. Fitch, Kitsch and Neighbor programs
create both outfile and outtree.

Dnadist DNA distance matrix calculation

Protdist Protein distance matrix calculation
Fitch Fitch-Margoliash tree drawing method without molecular clock
Kitsch Fitch-Margoliash tree drawing method with molecular clock
Neighbor Neighbor-Joining and UPGMA tree drawing method

Character based methods

These programs read in the sequence alignment, and produce either one or multiple trees in
the output files outfile and outtree.

Dnapars DNA parsimony

Dnapenny DNA parsimony using branch-and-bound
Dnaml DNA maximum likelihood without molecular clock
Dnamlk DNA maximum likelihood with molecular clock
Protpars Protein parsimony
Proml Protein maximum likelihood

Resampling tool

This program reads in a sequence alignment, and generates a specified number of random
samples into a file outfile. These random samples are usually used in subsequent analysis
as a sequence alignment file with the option M (“use multiple datasets”) turned on.

Seqboot Generates random samples by bootstrapping or jack-knifing

Tree drawing

These programs draw a tree from the specifications in the Newick-format. For example, the
specification can be in a file produced by the program Dnaml. The Newick file outtree

9
produced by Dnaml should be renamed to intree before visualizing the tree. Drawgram
and Drawtree produce a file plotfile, whereas Retree saves the result in a file outtree.

Drawgram Draws a rooted tree

Drawtree Draws an unrooted tree
Retree Interactive tree-rearrangement

Consensus trees

This program constructs a consensus tree from multiple trees. For example, Dnapars can
produce multiple trees, which can be summarized by the program Consense. Also the results
of bootstrapping are summarized by the program Consense as a majority rule tree.

Consense Draws consensus trees from multiple trees

Tree distances

This program computes, e.g., a topology-based distance between two or more trees. The
distance can be used to assess or compare the results from different analyses.

Treedist Computes distances between trees based on tree topology

10
Quick start

Here a DNA sequence data is used as an example. In the example using PHYLIP in Windows
operating system, three programs are used to construct and plot a tree by Neighbor joining (a
distance method) using the F84 evolutionary model. Details about other methods are
available in the succeeding sections.

Distance methods

Align your DNA sequences and save the alignment in PHYLIP-format as

alignment.phy. Start the program Dnadist by typing Dnadist to the command prompt
or double clicking on the program’s icon.

First Dnadist (and all the other programs also) checks whether there is a file infile in the
folder you started the program in. If it does not find infile it asks you to type in the name
of the sequence alignment file.

Dnadist: can't find input file "infile"

Please enter a new file name> alignment.phy

Note that the programs are easiest to use if both the programs and the datafiles are in the
same folder as in the example above. If datafiles are in a different folder, you can type in the
whole path to the file, e.g., if the files were in the folder D:\data you would type
Dnadist: can't find input file "infile"
Please enter a new file name> D:\data\alignment.phy

All PHYLIP programs are menu-driven. Below is the menu written by Dnadist. Every line in
the menu starts with a capital letter or number. You can change the settings of the program by
typing in the letter or the number in front of the option you would like to change. For
example, typing “d” and pressing Enter, would cycle through different evolutionary models
implemented in Dnadist. After you are satisfied with the settings (for this quick start, do not
change any options), you should type in “y” and press Enter. This starts the run.
Nucleic acid sequence Distance Matrix program, version 3.66

Settings for this run:

D Distance (F84, Kimura, Jukes-Cantor, LogDet)? F84
G Gamma distributed rates across sites? No
T Transition/transversion ratio? 2.0
C One category of substitution rates? Yes
W Use weights for sites? No
F Use empirical base frequencies? Yes
L Form of distance matrix? Square
M Analyze multiple data sets? No
I Input sequences interleaved? Yes
0 Terminal type (IBM PC, ANSI, none)? ANSI
1 Print out the data at start of run No
2 Print indications of progress of run Yes

Y to accept these or type the letter for one to change

y

11
Dnadist prints indications of the run (below). After it has finished calculating all the pairwise
distances between the sequences, it tells you so (Done.). These pairwise distances are saved
in a file outfile. The file contains just plain text, and you may want rename the file as
outfile.txt so that it opens automatically in Notepad when you double-click it.
Distances calculated for species
Rabbit ....
Human ...
Opossum ..
Chicken .
Frog

Distances written to file "outfile"

Done.

Next rename outfile as infile, and run the program Neighbor (type in Neighbor).
The next menu should appear. Now Neighbor has read the pairwise distances from the file
infile, and does not ask you for a new filename. You can again modify the settings to your
liking, but for this quick start just type in y and press Enter.

Neighbor-Joining/UPGMA method version 3.66

Settings for this run:

N Neighbor-joining or UPGMA tree? Neighbor-joining
O Outgroup root? No, use as outgroup species 1
L Lower-triangular data matrix? No
R Upper-triangular data matrix? No
S Subreplicates? No
J Randomize input order of species? No. Use input order
M Analyze multiple data sets? No
0 Terminal type (IBM PC, ANSI, none)? ANSI
1 Print out the data at start of run No
2 Print indications of progress of run Yes
3 Print out tree Yes
4 Write out trees onto tree file? Yes

Y to accept these or type the letter for one to change

y

Like Dnadist, Neighbor also prints out indications of the run. After completing the analysis,
the program tells you so (Done.).
Cycle 2: OTU 4 ( 0.62698) joins OTU 5 ( 0.95492)
Cycle 1: OTU 3 ( 0.73871) joins node 4 ( 0.17009)
last cycle:
OTU 1 ( 0.05116) joins OTU 2 ( 0.23064) joins node 3 ( 0.12944)

Output written on file "outfile"

Tree written on file "outtree"

Done.

The tree is now contained in the files outfile and outtree. You can view the graphical
tree in outfile by opening it in some text editor. Neighbor has now drawn the following tree
from our example data set.

12
 +---------------------Opossum
+---2
! ! +------------------Chicken
! +----1
! +----------------------------Frog
!
3-Rabbit
!
+------Human

Note that the tree should be viewed using a font such as Courier, where all the letters take the
same amount of space. Otherwise the nodes and branches might become disconnected.

Tree drawing

Next you can draw a nicer looking graphical tree from the file outtree using the program
Drawgram. First, rename the file outtree as intree, and start the program Drawgram.
Drawgram first searches for a file called fontfile from the current folder, and if it is
unable to find it (if you have not renamed one of the original font files as fontfile), it
asks for the name of the font file. You should then specify which font to use by typing in its
name (font1 – font6). After specifying that, the menu of the program appears. Now you
should change the final plotting devise as MS-Windows Bitmap using the option P. Postscript
output format is ideally suitable for publication images, but it is slightly complicated to use
on a basic Windows machine. The program also asks for the dimensions of the tree – you
might initially try 640 x 400. The settings are accepted by typing in “y” and pressing Enter.

DRAWGRAM from PHYLIP version 3.66

Reading tree ...
Tree has been read.
Loading the font ...
Drawtree: can't find font file "fontfile"
Please enter a new file name> font1
Font loaded.

Rooted tree plotting program version 3.66

Here are the settings:

0 Screen type (IBM PC, ANSI): IBM PC
P Final plotting device: Postscript printer
V Previewing device: MS Windows display
H Tree grows: Horizontally
S Tree style: Phenogram
B Use branch lengths: Yes
L Angle of labels: 90.0
R Scale of branch length: Automatically rescaled
D Depth/Breadth of tree: 0.53
T Stem-length/tree-depth: 0.05
C Character ht / tip space: 0.3333
A Ancestral nodes: Weighted
F Font: Times-Roman
M Horizontal margins: 1.65 cm
M Vertical margins: 2.16 cm
# Pages per tree: one page per tree

Y to accept these or type the letter for one to change

y

13
Drawgram opens a new window, where you can see a preview of the tree. If you’re satisfied
with the results, select from the File menu (in the same newly-opened window) “plot”. A new
file (plotfile) should now appear in the current folder. If you rename it as
plotfile.bmp you would be able to open it in some graphics package for more
modifications.

The final picture looks like this:

The length of the branch is the number of nucleotide or amino acid changes that are expected
in that particular branch in the tree. The number of changes is estimated using the
evolutionary model specified in the Dnadist program.

Amino acid sequences

The sequence of the programs is similar to the one presented here one with one important
exception. When using amino acid sequences for inferring phylogenies with distance
methods, the distance matrix is calculated using program Protdist, not Dnadist.

14
Basic analyses in more detail

Before proceeding further, please read the previous sections. They contain relevant
information, if you haven’t used PHYLIP earlier.

There are three different ways to analyze DNA or amino acid sequence data in PHYLIP.
Parsimony and maximum likelihood are character based methods, which means that they
treat every single site of the multiple sequence alignment independently. Distance methods
summarize the differences between sequences by calculating a pairwise distance measure
between all aligned sequences. In either case, after the data analysis and tree-drawing, the
validity of data can (or should, according to some) be assessed by a bootstrap analysis.

The programs can be invoked by double-clicking on their icons. If you have done some
analyses before within the same folder, the program detects that outfile already exists,
and it ask you to:

Dnadist: the file "outfile" that you wanted to

use as output file already exists.
Do you want to Replace it, Append to it,
write to a new File, or Quit?
(please type R, A, F, or Q)

Replace means to overwrite, Append adds the new results to the end of an existing file, and
File asks for a new file name. You can also Quit the program.

Distance methods

For the distance method analysis you'll need at least two programs. Dnadist or Protdist
calculates a matrix of pairwise distances between every sequence in the file. The tree is
inferred from those distances by Neighbor, Fitch or Kitsch program.

DNA data

The Dnadist program writes out a menu:

Nucleic acid sequence Distance Matrix program, version 3.66

Settings for this run:

D Distance (F84, Kimura, Jukes-Cantor, LogDet)? F84
G Gamma distributed rates across sites? No
T Transition/transversion ratio? 2.0
C One category of substitution rates? Yes
W Use weights for sites? No
F Use empirical base frequencies? Yes
L Form of distance matrix? Square
M Analyze multiple data sets? No
I Input sequences interleaved? Yes
0 Terminal type (IBM PC, ANSI, none)? ANSI
1 Print out the data at start of run No
2 Print indications of progress of run Yes

Y to accept these or type the letter for one to change

15
The settings can be changed by typing in the letter before the option, and pressing Return.
For example, typing "d" and return, would cycle through the different distance calculation
methods. These distances are also called evolutionary models. Evolutionary models are
mathematical formulas which try to compensate for the multiple substitution problem and
transition-transversion bias.

Briefly, Jukes-Cantor distance assumes that all substitutions are equally likely to happen.
Kimura distance has two different change rates (rate parameters), one for transitions and the
other for transversions. These models also assume that the equilibrium frequencies of all the
bases are 0.25. F84 distance has two rate parameters, one for transitions, and the other for
transversions, but also allows the equilibrium frequencies of the bases to differ from each
other. The LogDet distance should be used if there are (large) base frequency differences
between sequences in the tree. The LogDet distance cannot cope with ambiguity codes. It
must have completely defined sequences.

The transition/transversion ratio (Ts/Tv) can be modified if more detailed data is available on
an Ts/Tv ratio. If not, the Ts/Tv ratio of 1-2 is often a good approximation to the situation
with most mammalian nuclear genes.

Usually the phylogenetic methods assume that all the sites are evolving at the same speed.
This is clearly an unrealistic assumption. For example, third codon positions evolve with a
higher speed than second positions. That is because most of the substitutions in the third
position are selectively neutral whereas substitutions in the second codon position always
lead to amino acid changes. This rate heterogeneity can be treated by using gamma
distribution. The shape of the gamma distribution is defined by a parameter called alpha (α).
If you activate option "g" you'll be prompted to enter this shape parameter:

Coefficient of variation of substitution rate among sites (must be

positive)

In gamma distribution parameters, this is 1/(square root of alpha)

If you know alpha, you can calculate the prompted CV by the equations (Felsenstein,
PHYLIP documentation):

1 / √ (alpha) = CV
alpha = (1 / CV)2

A good approximation of the alpha for many protein coding genes is 0.5, but you can also
estimate this value using programs such as TREE-PUZZLE, which is not a part of PHYLIP
package. For a maximum likelihood solution using PHYLIP programs, you can apply the
example in the section Estimating transversion/transition ratio.

Another layer of rate variation also is available. The option "c" allows user-defined rate
categories. The user is prompted for the number of user-defined rates, and for the rates
themselves, which cannot be negative but can be zero. These numbers are defined relative to
each other, so that if rates for three categories are set to 1 : 3 : 2.5 this would have the same
meaning as setting them to 2 : 6 : 5. The assignment of rates to sites is then made by reading
a file whose default name is "categories". It should contain a string of digits 1 through 9. A
new line or a blank can occur after any character in this string.

16
Thus the categories file might look like this:

1222311111224111551155333333444

“The user can assign categories of rates to each site (for example, we might want first,
second, and third codon positions in a protein coding sequence to be three different
categories. This is done with the categories input file and the option "c". We then specify
(using the menu) the relative rates of evolution of sites in the different categories. For
example, we might specify that first, second, and third positions evolve at relative rates of
1.0, 0.8, and 2.7.” (Felsenstein, PHYLIP documentation).

“The weights-option ("w") allows us to specify weights on the individual characters. The
weights cause a character to be counted as if it were n characters, where n is the weight. By
use of the weights we can give weight to some characters, and drop others from the analysis.
In the molecular sequence programs only two values of the weights, 0 or 1 are allowed,
except for Dnapars”. (Felsenstein, PHYLIP documentation) An exception to this is Dnapars
which accept weights from 0-9…35 (0-9 with numbers and 10-35 with letters). For more
information on weighting, see the section Weighting in advanced topics chapter.

If you want to specify the base frequencies by yourself, you can do that by invoking the
option "f". The program then prompts you to type in the frequencies of different bases.

After you have modified the settings to your liking, you can calculate the distance matrix by
typing "y" and pressing return.

This will produce a distance matrix with default name outfile:

5
Rabbit 0.0000 0.2818 0.9386 0.9830 1.2617
Human 0.2818 0.0000 1.0795 1.1496 1.5312
Opossum 0.9386 1.0795 0.0000 1.5380 1.8615
Chicken 0.9830 1.1496 1.5380 0.0000 1.5819
Frog 1.2617 1.5312 1.8615 1.5819 0.0000

The outfile needs to be renamed, because the programs will go awry if they try both read
and write from and to the same file. After renaming this new file can be used as an input into
the Neighbor, Fitch, or Kitsch program.

17
Neighbor writes out a menu, where you can again change the settings:

Neighbor-Joining/UPGMA method version 3.66

Settings for this run:

N Neighbor-joining or UPGMA tree? Neighbor-joining
O Outgroup root? No, use as outgroup species 1
L Lower-triangular data matrix? No
R Upper-triangular data matrix? No
S Subreplicates? No
J Randomize input order of species? No. Use input order
M Analyze multiple data sets? No
0 Terminal type (IBM PC, ANSI, none)? ANSI
1 Print out the data at start of run No
2 Print indications of progress of run Yes
3 Print out tree Yes
4 Write out trees onto tree file? Yes

Y to accept these or type the letter for one to change

At this point, you can specify the outgroup of the tree. A sister taxon to the ingroup (the taxa
under investigation) is rather often used as an outgroup. For example, if human, chimpanzee
and gorilla are under scrutiny, orangutan could be specified as an outgroup.

Also Fitch writes out a menu, where setting can be modified, but the menu looks a bit
different:
Fitch-Margoliash method version 3.66

Settings for this run:

D Method (F-M, Minimum Evolution)? Fitch-Margoliash
U Search for best tree? Yes
P Power? 2.00000
- Negative branch lengths allowed? No
O Outgroup root? No, use as outgroup species 1
L Lower-triangular data matrix? No
R Upper-triangular data matrix? No
S Subreplicates? No
G Global rearrangements? No
J Randomize input order of species? No. Use input order
M Analyze multiple data sets? No
0 Terminal type (IBM PC, ANSI, none)? ANSI
1 Print out the data at start of run No
2 Print indications of progress of run Yes
3 Print out tree Yes
4 Write out trees onto tree file? Yes

Y to accept these or type the letter for one to change

Fitch-Margoliash is the distance based optimization method, which mean that it searches for
a tree with the smallest squared distance between the distances and their predictions from the
tree.

In Fitch you can also randomize the input order of the sequences with option "j", jumble.
Often the input order of the sequences affects the outcome of the analysis. This can be
assessed by randomizing the input order. The program also asks you to specify the number of
times you want to randomize the input order of the sequences. It is advisable to do jumbling
at least 10 times, because it almost certainly improves the results.

18
You also have the option of using a user-defined tree with option "u" which controls whether
negative branch lengths are allowed. In this case the program will, as default, read the user-
defined tree from the file intree. This also activates a new option "n". If you choose to use
the branch lengths from the user tree, Fitch calculates the Sum of Squares and the Average
Percent Standard Deviation for the user-defined tree.

Confirm the setting by typing "y" and pressing Return. Neighbor creates two new files,
outfile and outtree.

Outfile contains detailed information about the analysis and its results. It also contains the
inferred tree drawn with symbol graphics. Also the estimated branch lengths are reported in
the file.

5 Populations

Neighbor-Joining/UPGMA method version 3.6

Neighbor-joining method

Negative branch lengths allowed

+---------------------Opossum
+---2
! ! +------------------Chicken
! +----1
! +----------------------------Frog
!
3-Rabbit
!
+------Human

remember: this is an unrooted tree!

Between And Length

------- --- ------
3 Human 0.23064
3 2 0.12944
2 Opossum 0.73871
2 1 0.17009
1 Chicken 0.62698
1 Frog 0.95492
3 Rabbit 0.05116

File outtree contains the tree in a computer-readable format. It also reports the branch
lenghts (the numerical values reported after : marks), which are the same as in the file
outfile.

(Human:0.22996,((Frog:0.95134,Chicken:0.63056):0.16672,Opossum:0.74182):0.1
2891,Rabbit:0.05184);

19
Protein data

The protein distance method works more or less similarly as with DNA data. The program
Protdist writes out a menu with modifiable settings. After you have modified them to your
liking, the program produces an outfile, which contains pairwise distances. Those
distances can be transformed into a tree as has been described in the DNA data-section.

Protein distance algorithm, version 3.66

Settings for this run:

P Use JTT, PMB, PAM, Kimura, categories model? Jones-Taylor-Thornton
G Gamma distribution of rates among positions? No
C One category of substitution rates? Yes
W Use weights for positions? No
M Analyze multiple data sets? No
I Input sequences interleaved? Yes
0 Terminal type (IBM PC, ANSI)? ANSI
1 Print out the data at start of run No
2 Print indications of progress of run Yes

Are these settings correct? (type Y or the letter for one to change)

There are five options of evolutionary models to choose from: JTT (Jones-Taylor-Thornton),
PAM (Dayhoff), PMB, Kimura and categories. The PAM model uses the DCMut version of
Dayhoff’s original PAM model for calculations. JTT and PAM are widely used amino acid
substitution matrices (models), and PMB is a much resent model derived from conserved
blocks in the Blocks database. The Kimura model corrects for multiple hits, and categories is
a model put together by Felsenstein. See the PHYLIP documentation for more information.
Computing distances:
CAS1_HUMAN
CAS1_RABIT .
CAS1_MOUSE ..
CAS1_BOVIN ...
CAS1_SHEEP ....
CAS1_PIG .....

Output written to output file

Parsimony methods

DNA data

The programs available for DNA parsimony are Dnapars and Dnapenny. The differences
between these DNA parsimony programs are in the algorithm. Dnapars searches for the most
parsimonious tree by a heuristic algorithm which does not guarantee that the shortest tree is
found. Program Dnapenny uses the branch-and-bound algorithm, which guarantees that the
shortest tree is found, but takes quite much computer time.

20
Program Dnapars is controlled through this menu:
DNA parsimony algorithm, version 3.66

Setting for this run:

U Search for best tree? Yes
S Search option? More thorough search
V Number of trees to save? 10000
J Randomize input order of sequences? No. Use input order
O Outgroup root? No, use as outgroup species 1
T Use Threshold parsimony? No, use ordinary parsimony
N Use Transversion parsimony? No, count all steps
W Sites weighted? No
M Analyze multiple data sets? No
I Input sequences interleaved? Yes
0 Terminal type (IBM PC, ANSI, none)? ANSI
1 Print out the data at start of run No
2 Print indications of progress of run Yes
3 Print out tree Yes
4 Print out steps in each site No
5 Print sequences at all nodes of tree No
6 Write out trees onto tree file? Yes

Y to accept these or type the letter for one to change

Invoking the option "s" will allow you to specify more or less thorough search. The more
thorough search will save multiple equally parsimonious trees without collapsing those
branches that do not have support (no evidence of any change happened in that branch), and
it does rearrangements on all parts of those trees. Less thorough search collapses non-
supported branches before rearrangements. This leads to fewer rearrangements and faster
analysis. You can even decide to do rearrangements on one tree only. This means that only a
random one of the multiple equally parsimonious trees is rearranged. The search is then much
more restricted, which is not necessarily a good situation.

Often parsimony analysis produces many equally parsimonious trees, which can then be
analyzed more closely with, e.g., maximum likelihood or some other methods. You can
specify the number of trees to save with option "v".

Sometimes it is a good practice to limit the number of changes one site can contribute to the
tree. This can be accomplished with option "t". With "t" you can specify a threshold above
which all the changes are regarded as not counting further.

Transversion parsimony (option "n") uses only transversions for the parsimony analysis. This
tries to remove the bias resulting from the more rapid accumulation of transition substitutions
to the DNA.

The Option "5" can be used to infer ancient states of sites. If you choose to print sequences at
all nodes of the tree, the outfile will contain more information than normally:

21
From To Any Steps? State at upper node
. means same as in the node below it on tree
1 GTYCAGGGCT -GGGCATAAA AGGCAGAGCA GGGCCAGCTR
1 2 yes ...NR..... .?........ .......... A.RT......
2 3 yes ..C..CB... ..CA.VDGHD T.V...CV.. ........A.
3 Frog yes C..AA.TTTG GC..TGG.TT ..A....A.. T.A..GT..G
3 Chicken yes .A.GG.C... .-...CA.CG ..CT.T.C.C .CGGG....A
2 Opossum yes ..TTG...GC CA..G..... .........T ..A.T..T.T
1 Human yes AGC....... .......... ..T...G... .A....T..A
1 Rabbit yes ..T....A.. T......... .......... ....-....G

1 CTGCTTACAH
1 2 yes .W.....A.M
2 3 yes .....C....
3 Frog yes .T.A....CA
3 Chicken yes GA..C..G.C
2 Opossum yes .A..A.C.TA
1 Human yes T........T
1 Rabbit maybe .........C

Program Dnapenny has a different kind of menu:

Penny algorithm for DNA, version 3.66

branch-and-bound to find all most parsimonious trees

Settings for this run:

H How many groups of 100 trees: 1000
F How often to report, in trees: 100
S Branch and bound is simple? Yes
O Outgroup root? No, use as outgroup species 1
T Use Threshold parsimony? No, use ordinary parsimony
W Sites weighted? No
M Analyze multiple data sets? No
I Input sequences interleaved? Yes
0 Terminal type (IBM PC, ANSI, none)? ANSI
1 Print out the data at start of run No
2 Print indications of progress of run Yes
3 Print out tree Yes
4 Print out steps in each site No
5 Print sequences at all nodes of tree No
6 Write out trees onto tree file? Yes

Are these settings correct? (type Y or the letter for one to change)

You can specify ("h") how many hundreds of trees you want to search, and how often the
report is printed on the screen ("f").

You can also cause the branch-and-bound algorithm to reconsider the order of the species
with option "s". This will cause the analysis to take a longer time, but according to
Felsenstein “it might prove of use on some data sets [of intermediate messiness]”.

Program Dnapars gives only a little information during the run:

Adding species:
1. Rabbit
2. Human
3. Opossum
4. Chicken
5. Frog

22
Doing global rearrangements on all trees tied for best

!---------!
.........
.........

Output written to output file

Trees also written onto tree file

Done.

Dnapenny gives information about the search in real time, so you can estimate how long the
run is going to take:

How many
trees looked Approximate
at so far Length of How many percentage
(multiples longest tree trees this long searched
of 100): found so far found so far so far
---------- ------------ ------------ ------------
1 89 1 50 %
2 89 1 100 %

Output written to output file

Trees also written onto tree file

Both programs produce files outfile and outtree, but they look a bit different. Dnapars
infers the branching order of the tree and estimates the branch lengths, but Dnapenny only
infers the branching order. The length of the tree is on a line “requires a total of xxxx.xx”. A
smaller value means a more parsimonious tree.

Thus, the outfile and outtree of Dnapars:

DNA parsimony algorithm, version 3.6

One most parsimonious tree found:

+--------------------Frog
+--------3
+----------2 +-----------------Chicken
| |
| +----------------Opossum
|
1----------Human
|
+------------Rabbit

requires a total of 3019.000

between and length

------- --- ------
1 2 0.181782
2 3 0.150139
3 Frog 0.347616
3 Chicken 0.304468
2 Opossum 0.291273
1 Human 0.182778
1 Rabbit 0.219167

23
(((Frog:0.34762,Chicken:0.30447):0.15014,Opossum:0.29127):0.18178,Human:0.1
8278,Rabbit:0.21917);

look different from those produced by Dnapenny:

Penny algorithm for DNA, version 3.6

branch-and-bound to find all most parsimonious trees

requires a total of 3019.000

One most parsimonious tree found:

+-----------Rabbit
!
! +--Frog
1 +--3
! +--2 +--Chicken
! ! !
+--4 +-----Opossum
!
+--------Human

remember: this is an unrooted tree!

(Rabbit,(((Frog,Chicken),Opossum),Human));

Protein data

The protein parsimony program Protpars is comparable to Dnapars, and its menu looks very
similar:
Protein parsimony algorithm, version 3.66

Setting for this run:

U Search for best tree? Yes
J Randomize input order of sequences? No. Use input order
O Outgroup root? No, use as outgroup species 1
T Use Threshold parsimony? No, use ordinary parsimony
C Use which genetic code? Universal
W Sites weighted? No
M Analyze multiple data sets? No
I Input sequences interleaved? Yes
0 Terminal type (IBM PC, ANSI, none)? ANSI
1 Print out the data at start of run No
2 Print indications of progress of run Yes
3 Print out tree Yes
4 Print out steps in each site No
5 Print sequences at all nodes of tree No
6 Write out trees onto tree file? Yes

Are these settings correct? (type Y or the letter for one to change)

You can specify the genetic code used for the analysis by option "c". You have an option to
choose from the universal code and four mitochondrial codes.

24
However, its outfile and outtree are more similar to the output of Dnapenny:
Protein parsimony algorithm, version 3.6

One most parsimonious tree found:

+-----CAS1_PIG
+--5
! ! +--CAS1_SHEEP
+--3 +--4
! ! +--CAS1_BOVIN
+--2 !
! ! +--------CAS1_MOUSE
1 !
! +-----------CAS1_RABIT
!
+--------------CAS1_HUMAN

remember: this is an unrooted tree!

requires a total of 1061.000

((((CAS1_PIG,(CAS1_SHEEP,CAS1_BOVIN)),CAS1_MOUSE),CAS1_RABIT),
CAS1_HUMAN);

Maximum likelihood methods

There are two maximum likelihood programs for DNA data, and two for protein data. Dnaml
uses a maximum likelihood method without molecular clock, and Dnamlk assumes a
molecular clock. Proml does not and promlk does assume molecular clock.

DNA data

Menu of Dnaml:

Nucleic acid sequence Maximum Likelihood method, version 3.66

Settings for this run:

U Search for best tree? Yes
T Transition/transversion ratio: 2.0000
F Use empirical base frequencies? Yes
C One category of sites? Yes
R Rate variation among sites? constant rate
W Sites weighted? No
S Speedier but rougher analysis? Yes
G Global rearrangements? No
J Randomize input order of sequences? No. Use input order
O Outgroup root? No, use as outgroup species 1
M Analyze multiple data sets? No
I Input sequences interleaved? Yes
0 Terminal type (IBM PC, ANSI, none)? ANSI
1 Print out the data at start of run No
2 Print indications of progress of run Yes
3 Print out tree Yes
4 Write out trees onto tree file? Yes
5 Reconstruct hypothetical sequences? No

Y to accept these or type the letter for one to change

25
Most of the options has already been covered in the chapters above. However, there two new
option, "s" and "g" with which you can modify the search parameters a bit.

“The option "s" turns on or off the search method which iterates or does not iterate the branch
lengths in all topologies. Turning this option off ("No, not rough") will cause the program to
run more slowly, but it will also be a bit more likely to find the tree topology of highest
likelihood.” (Felsenstein, PHYLIP documentation)

“The "g" (global search) option causes, after the last species is added to the tree, each
possible group to be removed and re-added. This improves the result, since the position of
every species is reconsidered. It approximately triples the run-time of the program.”
(Felsenstein, PHYLIP documentation) This is equivalent to the rearrangements done in the
parsimony program Dnapars. Specifically, the program uses the SPR (subtree pruning and
regrafting) method for rearranging (or as others say, swapping) the trees.
“If more than one category is specified, then another option, "a", becomes visible in the
menu. This allows us to specify that we want to assume that sites that have the same regional
rate category are expected to be clustered so that there is autocorrelation of rates. The
program asks for the value of the average patch length. This is an expected length of patches
that have the same rate. If it is 1, the rates of successive sites will be independent. If it is, say,
10.25, then the chance of change to a new rate will be 1 / 10.25 after every site. However the
"new rate" is randomly drawn from the mix of rates, and hence could even be the same. So
the actual observed length of patches with the same rate will be a bit larger than 10.25. Note
below that if you choose multiple patches, there will be an estimate in the output file as to
which combination of rate categories contributed most to the likelihood.” (Felsenstein,
PHYLIP documentation)

The menu of the Dnamlk:

Nucleic acid sequence
Maximum Likelihood method with molecular clock, version 3.66

Settings for this run:

U Search for best tree? Yes
T Transition/transversion ratio: 2.0
F Use empirical base frequencies? Yes
C One category of substitution rates? Yes
R Rate variation among sites? constant rate
G Global rearrangements? No
W Sites weighted? No
J Randomize input order of sequences? No. Use input order
M Analyze multiple data sets? No
I Input sequences interleaved? Yes
0 Terminal type (IBM PC, ANSI, none)? ANSI
1 Print out the data at start of run No
2 Print indications of progress of run Yes
3 Print out tree Yes
4 Write out trees onto tree file? Yes
5 Reconstruct hypothetical sequences? No

Are these settings correct? (type Y or the letter for one to change)

The outfile of the Dnaml is quite similar to the outfile of Dnamlk. The first of these is the
outfile from Dnaml analysis, and the latter is the same file from Dnamlk.

26
Nucleic acid sequence Maximum Likelihood method, version 3.6

Empirical Base Frequencies:

A 0.25650
C 0.21951
G 0.22091
T(U) 0.30309

Transition/transversion ratio = 2.000000

+-----Human
|
| +--------------------------Frog
| +-----3
1-----2 +-----------------Chicken
| |
| +--------------------Opossum
|
+--Rabbit

remember: this is an unrooted tree!

Ln Likelihood = -9695.14457

Between And Length Approx. Confidence Limits

------- --- ------ ------- ---------- ------

1 Rabbit 0.09743 ( 0.06975, 0.12511) **

1 Human 0.20090 ( 0.16716, 0.23474) **
1 2 0.18632 ( 0.12823, 0.24451) **
2 3 0.19152 ( 0.11910, 0.26384) **
3 Frog 0.90555 ( 0.78325, 1.02776) **
3 Chicken 0.58309 ( 0.49455, 0.67166) **
2 Opossum 0.69201 ( 0.60518, 0.77894) **

* = significantly positive, P < 0.05

** = significantly positive, P < 0.01

Nucleic acid sequence

Maximum Likelihood method with molecular clock, version 3.6

Empirical Base Frequencies:

A 0.25650
C 0.21951
G 0.22091
T(U) 0.30309

Transition/transversion ratio = 2.000000

+---------------------------------------Frog
--4
! +--------------------------------Chicken
+------3
! +------------------------Opossum
+-------2
! +-------Human
+-----------------1
+-------Rabbit

27
Ln Likelihood = -9714.14764

Ancestor Node Node Height Length

-------- ---- ---- ------ ------
root 4
4 Frog 0.77626 0.77626
4 3 0.13842 0.13842
3 Chicken 0.77626 0.63785
3 2 0.27815 0.13973
2 Opossum 0.77626 0.49812
2 1 0.62602 0.34787
1 Human 0.77626 0.15025
1 Rabbit 0.77626 0.15025

As you can see, Dnaml produces confidence intervals of the branch lengths of the tree, but
Dnamlk doesn't. Dnaml also produces a rough estimate of the p-value that the length of the
branch is significantly greater than zero. This test is done by comparing the tree with the
inferred branch length to a tree where the same branch has been scaled to be zero. This result
is only an approximation, so do not rely on it when interpreting the results.

The reported likelihood (Ln Likelihood) is actually a natural logarithm of the likelihood. The
closer the likelihood value is to zero, the better. It can be used for statistical testing of
hypothesis, as will be discussed in the advanced analysis chapter.

The outtree from both programs is identical in form except that the tree has a three-way
fork at its base:

(Human:0.20090,((Frog:0.90555,Chicken:0.58309):0.19152,
Opossum:0.69201):0.18632,Rabbit:0.09743);

Protein data

The startup menu of the program Proml is similar to Dnaml:

Amino acid sequence Maximum Likelihood method, version 3.66

Settings for this run:

U Search for best tree? Yes
P JTT, PMB or PAM probability model? Jones-Taylor-Thornton
C One category of sites? Yes
R Rate variation among sites? constant rate of change
W Sites weighted? No
S Speedier but rougher analysis? Yes
G Global rearrangements? No
J Randomize input order of sequences? No. Use input order
O Outgroup root? No, use as outgroup species 1
M Analyze multiple data sets? No
I Input sequences interleaved? Yes
0 Terminal type (IBM PC, ANSI, none)? ANSI
1 Print out the data at start of run No
2 Print indications of progress of run Yes
3 Print out tree Yes
4 Write out trees onto tree file? Yes
5 Reconstruct hypothetical sequences? No

Y to accept these or type the letter for one to change

28
The program Proml assumes no molecular clock, and produces output files which are nearly
identical to the output files of Dnaml.

Resampling procedure

The idea behind resampling (bootstrapping and other methods) is to assess how reliable a tree
we can produce with the dataset at hand. Initially, the sequence alignment is analyzed in the
usual way. Then, resampling proceeds (see the flow chart in the end of this book, also) by
first creating a number (100-10000) of random datasets from the original dataset. These
random datasets are analyzed in exactly the same way the original dataset was analyzed, and
the results from the random datasets are summarized by constructing a majority rule
consensus tree (program Consense). Resampling with replacement (bootstrapping) can be
done with the program Seqboot:
Bootstrapping algorithm, version 3.66

Settings for this run:

D Sequence, Morph, Rest., Gene Freqs? Molecular sequences
J Bootstrap, Jackknife, Permute, Rewrite? Bootstrap
% Regular or altered sampling fraction? regular
B Block size for block-bootstrapping? 1 (regular bootstrap)
R How many replicates? 100
W Read weights of characters? No
C Read categories of sites? No
S Write out data sets or just weights? Data sets
I Input sequences interleaved? Yes
0 Terminal type (IBM PC, ANSI, none)? ANSI
1 Print out the data at start of run No
2 Print indications of progress of run Yes

Y to accept these or type the letter for one to change

You have several options to choose from. First, there are three resampling procedures
available in the Seqboot program (option "j"). Bootstrapping creates (with block size=1,
option "b") new data sets, which are of the same size as the original sequence alignment. In
bootstrapping resamples, every sequence site is represented in the alignment a random
number of times, including possibly zero times. The jackknife deletes a random 50% of sites
from the original alignment, and none of the sites can occur in the resamples more than once.
Permutation of species within each site produces data matrices that have the same number
and kinds of characters but no taxonomic structure. It is used for different purposes than the
bootstrap, as it tests not the variation around an estimated tree but the hypothesis that there is
no taxonomic structure in the data: if a statistic such as number of steps is significantly
smaller in the actual data than it is in replicates that are permuted, then we can argue that
there is some taxonomic structure in the data (though perhaps it might be just a pair of sibling
species). The program also converts PHYLIP-formatted data into other formats (Nexus and
XML) using the “rewrite” selection (option J).

With option "r" we set the number of resampled data matrixes produced.

You can also use weights with the bootstrapping procedure (option "w"). If you want to save
hard disk space, you can generate, a file containing sets of weights insetad of new datasets.
These weights can then be used in the same way as the new bootstrapped dataset. Instead of
using multiple datasets in the analysis program, you then need to use these sets of weights.

29
After accepting the settings, Seqboot asks for a random number. Pick an odd random number.
Just keep in mind that using the same random number produces always exactly the same
result. If you need to run bootstrapping several times, change the random number between
the runs.

Accepting the settings produces a file where there are 100 random samples:
6 333
CAS1_MOUSE KLCLAAAFMR RHHHSSNNNN VSSSSQQQQQ QQHHSSE--- --------II IFQQPKYYYL
CAS1_RABIT KLCLTTALRK KHHHLL---- HLLLKLLQQE QQPPSSQEEI ILLKKERLRR RFQQTVPPPL
CAS1_PIG KFCLVVALRK KPPPLL---- HQQQEHHQQN EEPPSRELLF FKKEERKRFF FPVVPLLLLS
CAS1_HUMAN RLCLVVALRK KPPPLL---- YPPPERRQQN PPSSSS---- ---------- ----PIPPPL
CAS1_SHEEP KLCLVVALRK KPPPII---- HQQQGLLPP- ------EVVL LNNEEN-RFF FVAAPFPPPE
CAS1_BOVIN KLCLVVALRK KPPPII---- HQQQGLLQQ- ------EVVL LNNEEN-RFF FFAAPFPPPE
6 333
CAS1_HUMAN MRRLLLLLCL AAARPKKPLL RYPPPRRLQN PESSSE---- ---------- --PPPPPLEE
CAS1_SHEEP MKKLLLLLCL AAARPKKPII KHQQQLLSP- ------EEVN N-LLRFFVVV VVPPPPPEVV
CAS1_MOUSE MKKLLLLLCL AAAMPRRHSS RVSSSQQTQQ QSSSSEEE-- -----IIFFK KKPPYYYLNN
CAS1_PIG MKKLLLFFCL AAARPKKPLL RHQQQHHLQN EDSRREEELR RKFFRFFPPE EEPPLLLSQQ
CAS1_RABIT MKKLLLLLCL AAARHKKHLL GHLLLLLTQE QESSSEQQEE ERKKLRRFFV VVTTPPPLEE
CAS1_BOVIN MKKLLLLLCL AAARPKKPII KHQQQLLPQ- ------EEVN N-LLRFFFFV VVPPPPPEVV
[And so forth...]

Then we proceed with this file as if it was the original sequence alignment. In our case, lets
produce distance trees with the program Protdist. We start by renaming outfile to infile, and
invoking the program:

Protein distance algorithm, version 3.66

Settings for this run:

P Use JTT, PMB, PAM, Kimura, categories model? Jones-Taylor-Thornton
matrix
G Gamma distribution of rates among positions? No
C One category of substitution rates? Yes
W Use weights for positions? No
M Analyze multiple data sets? No
I Input sequences interleaved? Yes
0 Terminal type (IBM PC, ANSI)? ANSI
1 Print out the data at start of run No
2 Print indications of progress of run Yes

Are these settings correct? (type Y or the letter for one to change)
m
How many data sets?
100

We have to tell the program that there are multiple datasets in the same file by changing the
option “m”. Then the distance matrix will be calculated for all the 100 random samples. The
outfile produced contains all these distance matrixes:

6
CAS1_MOUSE 0.00000 0.80622 1.46041 1.14997 2.11928 1.98883
CAS1_RABIT 0.80622 0.00000 1.35772 0.82152 1.76401 1.49899
CAS1_PIG 1.46041 1.35772 0.00000 1.30075 0.61110 0.58505
CAS1_HUMAN 1.14997 0.82152 1.30075 0.00000 1.72499 1.56997
CAS1_SHEEP 2.11928 1.76401 0.61110 1.72499 0.00000 0.08268
CAS1_BOVIN 1.98883 1.49899 0.58505 1.56997 0.08268 0.00000

30
 6
CAS1_HUMAN 0.00000 2.03032 1.01938 1.27690 0.69497 1.95053
CAS1_SHEEP 2.03032 0.00000 2.15271 0.71847 1.79089 0.16285
CAS1_MOUSE 1.01938 2.15271 0.00000 1.47633 0.90248 1.87560
CAS1_PIG 1.27690 0.71847 1.47633 0.00000 1.52777 0.77397
CAS1_RABIT 0.69497 1.79089 0.90248 1.52777 0.00000 1.54938
CAS1_BOVIN 1.95053 0.16285 1.87560 0.77397 1.54938 0.00000

[And so forth]

These distance matrices are then used to infer the tree with the original method, i.e., if the
original data was analysed with F84-distance and Neighbor-joining, the bootstrapping
analysis has to be performed with the same settings.

Again, in the program Neighbor the multiple datasets option (m) has to be chosen. The
resulting outtree contains trees for all the 100 random datasets.
((CAS1_PIG:0.09668,(CAS1_SHEEP:0.11341,CAS1_BOVIN:-0.03073):0.46006):0.75811,
CAS1_HUMAN:0.41697,(CAS1_MOUSE:0.55088,CAS1_RABIT:0.25534):0.16567);
(((CAS1_SHEEP:0.14929,CAS1_BOVIN:0.01356):0.52397,CAS1_PIG:0.14082):0.82720,
CAS1_MOUSE:0.54493,(CAS1_HUMAN:0.38539,CAS1_RABIT:0.30958):0.06851);

[And so forth]

These trees are then combined in a consensus tree with the program Consense.

Consensus tree program, version 3.66

Settings for this run:

C Consensus type (MRe, strict, MR, Ml): Majority rule (extended)
O Outgroup root: No, use as outgroup
species 1
R Trees to be treated as Rooted: No
T Terminal type (IBM PC, ANSI, none): ANSI
1 Print out the sets of species: Yes
2 Print indications of progress of run: Yes
3 Print out tree: Yes
4 Write out trees onto tree file: Yes

Are these settings correct? (type Y or the letter for one to change)

There are four consensus tree types to choose from. Strict consensus creates a tree which only
includes the set of sequences, if it occurs in all the trees. The MR, MRe and M1 all produce a
majority rule trees with slightly different options. The default method (MRe) will include into
the new tree all the groups of sequences which are present in more than 50% of the trees, plus
the most frequent others that are compatibel with these. Ml lets you to specify the percentage.
Note, that the consensus tree from bootstrapping samples should always be drawn with
majority rule method.

The outfile and outtree contain the information on how many times each set of species
was find to be together in the random samples. If this value is under, say 70-95%, the result
should be interpreted with caution. This also implies that there probably is not enough data to
differentiate between topologies when the sets in them have low bootstrap values.

Majority-rule and strict consensus tree program, version 3.61

Species in order:

31
 CAS1 PIG
CAS1 SHEEP
CAS1 BOVIN
CAS1 HUMAN
CAS1 MOUSE
CAS1 RABIT

Sets included in the consensus tree

Set (species in order) How many times out of 100.00

...*** 100.00
.**... 100.00
....** 77.00

Sets NOT included in consensus tree:

Set (species in order) How many times out of 100.00

...*.* 16.00
...**. 7.00

Extended majority rule consensus tree

CONSENSUS TREE:
the numbers at the forks indicate the number
of times the group consisting of the species
which are to the right of that fork occurred
among the trees, out of 100.00 trees

+-------------CAS1 HUMAN
+100.0-|
| | +------CAS1 RABIT
| +-77.0-|
+------| +------CAS1 MOUSE
| |
| | +------CAS1 SHEEP
| +-------100.0-|
| +------CAS1 BOVIN
|
+---------------------------CAS1 PIG

remember: this is an unrooted tree!

The concomitant outtree contains the same information. The bootstrapping values are
reported in the place of branch lenghts.

(((CAS1_HUMAN:100.0,(CAS1_RABIT:100.0,CAS1_MOUSE:100.0):77.0):100.0,(CAS1_S
HEEP:100.0,CAS1_BOVIN:100.0):100.0):100.0,CAS1_PIG:100.0);

The bootstrapping procedure implemented in PHYLIP does not perform the analysis on the
tree inferred from original dataset. Instead, Felsenstein has argued that one reasonable trees
would be the one recovered from the random samples as described above. However, the
original inferred tree and the tree produced by bootstrapping are usually pretty similar.

32
Drawing the tree

We can draw a picture of an unrooted tree from the information contained in the file
"plottree" with the program Drawgram.

Drawgram: can't find input tree file "intree"

Please enter a new file name> plottree
DRAWGRAM from PHYLIP version 3.6
Reading tree ...
Tree has been read.
Loading the font ....
Drawgram: can't find font file "fontfile"
Please enter a new file name> font1
Font loaded.

Rooted tree plotting program version 3.66

Here are the settings:

0 Screen type (IBM PC, ANSI): ANSI
P Final plotting device: Postscript printer
V Previewing device: X Windows display
H Tree grows: Horizontally
S Tree style: Phenogram
B Use branch lengths: Yes
L Angle of labels: 90.0
R Scale of branch length: Automatically rescaled
D Depth/Breadth of tree: 0.53
T Stem-length/tree-depth: 0.05
C Character ht / tip space: 0.3333
A Ancestral nodes: Weighted
F Font: Times-Roman
M Horizontal margins: 1.65 cm
M Vertical margins: 2.16 cm
# Pages per tree: one page per tree

Y to accept these or type the letter for one to change

First, we want to be able to view the file easily in Windows, and we change the Final plotting
device by typing "p" and Return.

33
A new menu opens:

Which plotter or printer will the tree be drawn on?

(many other brands or models are compatible with these)

type: to choose one compatible with:

L Postscript printer file format

M PICT format (for drawing programs)
J HP Laserjet PCL file format
W MS-Windows Bitmap
F FIG 2.0 drawing program format
A Idraw drawing program format
Z VRML Virtual Reality Markup Language file
P PCX file format (for drawing programs)
K TeKtronix 4010 graphics terminal
X X Bitmap format
V POVRAY 3D rendering program file
R Rayshade 3D rendering program file
H Hewlett-Packard pen plotter (HPGL file format)
D DEC ReGIS graphics (VT240 terminal)
E Epson MX-80 dot-matrix printer
C Prowriter/Imagewriter dot-matrix printer
T Toshiba 24-pin dot-matrix printer
O Okidata dot-matrix printer
B Houston Instruments plotter
U other: one you have inserted code for
Choose one:

From this we will choose MS-Windows Bitmap by typing "w" and Return. The program then
asks the desired resolution of the picture (on the next page). After specifying that, you'll be
dropped back to the main menu.

Please select the MS-Windows bitmap file resolution

X resolution?
640
Y resolution?
400

After accepting the settings by typing in "y" and Return, a new window opens. In this
window you can preview the tree, and if it looks good, plot it into file by selecting from File
menu option plot. The tree is plotted in the plotfile. The options of the program Drawtree
are similar, but the program produces a rooted tree.

34
Advanced topics

Here some more advanced analysis options are considered. These options are less often used
by researchers, especially when they do not know about these possibilities.

User trees
User trees have multiple purposes. They can be used for transition/transversion ratio
approximation, likelihood comparisons (statistical comparisons of the trees), etc. These
purposes will be covered in the succeeding sections.

User trees should be in a plain text file named intree. Some programs do not read the
branch lengths or the number of the trees, so this needs a little experimentation. However, the
basic form of the intree file is:

(Frog,(Chicken,(Opossum,(Human,Rabbit))));

The easiest way to produce user-trees is to perform an analysis with some of the programs,
and then modify the file outtree with the program Retree or by some text editor.

Program Retree
Retree reads in a treefile, which is in Newick format. For example, a tree produced with
Dnaml can be directly imported into Retree. In this example we modify the tree:

(Human:0.20090,((Frog:0.90555,Chicken:0.58309):0.19152,
Opossum:0.69201):0.18632,Rabbit:0.09743);

when the program starts, you get a menu:

Tree Rearrangement, version 3.66

Settings for this run:

U Initial tree (arbitrary, user, specify)? User tree from tree
file
N Format to write out trees (PHYLIP, Nexus, XML)? PHYLIP
0 Graphics type (IBM PC, ANSI)? ANSI
W Width of terminal screen, of plotting area? 80, 80
L Number of lines on screen? 24

Are these settings correct? (type Y or the letter for one to change)

After you have confirmed these setting, a new view opens:

,---------------------------1:Frog
,----------------------8
,----------7 `-----------------------------2:Chicken
! !
! !
! `---------------3:Opossum
!
--6------4:Human
!
`-----------------------------------5:Rabbit

NEXT? (Options: R . = U W O M T F D B N H J K L C + ? X Q) (? for Help)

The options of the program can be printed on the screen by typing "?" and Return:

35
. Redisplay the same tree again
= Redisplay the same tree without/with lengths
U Undo the most recent change in the tree
W Write tree to a file
+ Read next tree from file (may blow up if none is there)
R Rearrange a tree by moving a node or group
O select an Outgroup for the tree
M Midpoint root the tree
T Transpose immediate branches at a node
F Flip (rotate) subtree at a node
D Delete or restore nodes
B Change or specify the length of a branch
N Change or specify the name(s) of tip(s)
H Move viewing window to the left
J Move viewing window downward
K Move viewing window upward
L Move viewing window to the right
C show only one Clade (subtree) (might be useful if tree is too big)
? Help (this screen)
Q (Quit) Exit from program
X Exit from program

TO CONTINUE, PRESS ON THE Return OR Enter KEY

Now you can reroot the tree, swap the branches, etc. In this case, we want to remove branch
lengths from all the branches of the tree. This can be done by invoking the option "b":

NEXT? (Options: R . = U W O M T F D B N H J K L C + ? X Q)(? for Help) b

Specify length of which branch (0 = all branches)? 0

(this operation cannot be undone)

enter U to leave the lengths unchanged

OR enter R to remove the lengths from all branches:
r

,-----1:Frog
,-----7
! ! ,--2:Human
! `--8
! `--3:Opossum
!
--6-----------4:Chicken
!
`-----------5:Rabbit

Next we want to save the tree and quit the program:

NEXT? (Options: R . U W O T F D B N H J K L C + ? X Q) (? for Help) x

Do you want to write out the tree to a file? (Y or N) y
Retree: the file "outtree" that you wanted to
use as output tree file already exists.
Do you want to Replace it, Append to it,
write to a new File, or Quit?
(please type R, A, F, or Q)
r

Enter R if the tree is to be rooted

OR enter U if the tree is to be unrooted: u

Tree written to file “outtree”

36
The tree written by program Retree in file outtree is:

(((Frog,Chicken),Opossum),Human,Rabbit);

Estimating the transition/transversion ratio

The estimated ratio depends on the specified outgroup. So, if you have information about the
outgroup you are planning to use in the forthcoming analysis, you should use it in this
estimation process too.

One way to estimate the transition/transversion ratio is to run maximum likelihood program
Dnaml multiple times with different transversion/transition ratios, and try to find the value,
which maximizes the likelihood.

The maximum likelihood method works this way (* = highest likelihood value corresponding
to the maximum likelihood estimate of s/t-ratio):

s/t-ratio likelihood s/t-ratio likelihood

1.25 -9644.49078 1.00 -9643.20009
1.15 -9642.29536 0.95 -9644.9407
1.10 -9641.96561* 0.90 -9647.57138
1.05 -9642.23988 0.85 -9651.21944

You should use the inferred transition/transversion ratio to reanalyze your dataset, if you
have previously used the default settings. In the example, the tree is similar to one inferred
using the default settings:

+-----Human
|
| +-----------------------Frog
| +----3
1-----2 +---------------Chicken
| |
| +------------------Opossum
|
+--Rabbit

But the lengths of the branches are different (compare to the tree inferred using Dnaml in
sections above):
Ln Likelihood = -9641.96561

Between And Length Approx. Confidence Limits

------- --- ------ ------- ---------- ------
1 Rabbit 0.10097 ( 0.07432, 0.12755) **
1 Human 0.18761 ( 0.15605, 0.21917) **
1 2 0.18557 ( 0.13596, 0.23522) **
2 3 0.13641 ( 0.07800, 0.19475) **
3 Frog 0.75496 ( 0.66344, 0.84638) **
3 Chicken 0.51613 ( 0.44578, 0.58657) **
2 Opossum 0.57699 ( 0.50869, 0.64529) **

* = significantly positive, P < 0.05

** = significantly positive, P < 0.01

37
The procedure outlined above can be quite slow, and a good compromise for estimating the
transition/transversion ratio is to acquire a good tree and then estimate the parameters using
the tree. In practice, this can be done using a user tree in program Dnaml and letting the
program to re-estimate the branch lengths. Even faster would be to use the user tree but not
allow Dnaml to re-estimate the branch lengths. Run the analysis with user tree multiple times
and supply different estimates of transition/transversion ratio. Then, as outlined above, pick
the results giving the highest likelihood value.

Estimating base frequencies

Above, a maximum likelihood method for estimating transition/transversion ratio is

presented. A similar method can be used for inferring the maximum likelihood estimators of
base frequencies.

Empirical frequencies are:

A 0.25650
C 0.21951
G 0.22091
T(U) 0.30309

We can start the estimation by option "f", which then prompts for base frequencies. The
frequency values should be separated by blanks:

Base frequencies for A, C, G, T/U (use blanks to separate)?

0.25 0.22 0.22 0.31

Note the likelihood of the old (-9695.14457) and new (-9696.03008) analysis. In this
example, the likelihood got lower after modifying the base frequencies. This indicates that
the empirical frequencies were better. This procedure can be repeated a number of times in
order to get the best estimate, but it requires multiple runs with different base frequencies and
might get tedious.

Testing molecular clock

The molecular clock assumption can be tested by the two maximum likelihood programs,
Dnaml and Dnamlk. The test, which compares two likelihoods is called likelihood ratio test.

First the analysis is run by using the program Dnamlk, which produces an unrooted tree with
molecular clock assumption:

(Frog:0.77626,(Chicken:0.63785,(Opossum:0.49812,(Human:0.15025,
Rabbit:0.15025):0.34787):0.13973):0.13842);

Note the log-likelihood of the tree -9714.14764.

The program Dnaml can read in this treefile, which has first to be named intree.
Remember to run the analysis with the user tree option ("u") turned on. You should also run
the analysis without using the lenghts on the user tree.

38
Note the log-likelihood of tree from the Dnaml analysis -9695.14457

Now calculate the difference between the log-likelihoods: 2 * (-9714.14764 + 9695.14457) =

38.006. This difference is then compared to the chi-square distribution with df (degrees of
freedom) equal to (number of sequences -2). The table consists of dfs on the left side, and p-
values on the upper side. The values in the crossing of the p-value and df is called the critical
value of the distribution. P-value is the risk that we conclude by chance that sequences did
not evolve according to a clock, if they actually did evolve with a clock. Often the p-value is
set to be 0.05.

If the difference is larger than the critical value mentioned in the table below, the sequences
did not evolve according to a molecular clock.

In our case, the difference is 38.006. That is larger than the critical value with df=3 and p-
value=0.05, which is 7.815. Thus, we conclude that we can reject the clock hypothesis, but
by chance we might do that in 1 / 20 cases if the same analysis were repeated (p-value of
0.05) when there actually is a clock.
Probability of exceeding the critical value
df 0.10 0.05 0.025 0.01 0.001
1 2.706 3.841 5.024 6.635 10.828
2 4.605 5.991 7.378 9.210 13.816
3 6.251 7.815 9.348 11.345 16.266
4 7.779 9.488 11.143 13.277 18.467
5 9.236 11.070 12.833 15.086 20.515
6 10.645 12.592 14.449 16.812 22.458
7 12.017 14.067 16.013 18.475 24.322
8 13.362 15.507 17.535 20.090 26.125
9 14.684 16.919 19.023 21.666 27.877
10 15.987 18.307 20.483 23.209 29.588

Inferring ancient states of sequence sites

Why are we interested in reconstructing ancestral character states? It can provide important
information about adaptive radiations and key innovations for these radiations. One
interesting approach has been to use the inferred ancient sequences to produce an inferred
ancient protein. The biochemical properties of this protein could be studied and compared to
the modern proteins.

The ancestral states of sequence sites can be inferred either by parsimony or by maximum
likelihood. It is often thought, that the maximum likelihood method is more accurate than the
parsimony method for inferring ancient sequences.

Ancestral states can be inferred in programs Dnapars, Protpars, Dnaml, Dnamlk, Proml, and
Promlk by turning on the option "5", print sequences at all nodes of tree and reconstruct
hypothetical sequences, respectively.

39
Tree is identical for both methods

+--------------------Frog
+---------------------3
+--2 +-----------------Chicken
| |
| +Human
|
1---------Opossum
|
+Rabbit

[with the maximum maximum likelihood method]

Probable sequences at interior nodes:

node Reconstructed sequence (caps if > 0.95)

1 GTTCAGGACT TGGGCATAAA AGGCAGAGCA GGGCcAGCTG CTGCTTACAC

2 agcCAGGgCT tGGGCATAAA AGtCAGgGCA GaGCcAtCTa tTGCTTACAt
3 rrcrrcyrcy rscacrrgyr ygycarcrcr rrgyyakcrr yygyycarry
Frog CTCAACTTTG GCCATGGGTT TGACAGCACA TGATCGTCAG CTGATCAACA
Chicken GACGGCCGCT rrCACCAGCG TGCTATCCCC ACGGGAGCAA GAGCCCAGAC
Human AGCCAGGGCT rGGGCATAAA AGTCAGGGCA GAGCCATCTA TTGCTTACAT
Opossum GTTTGGGGGC CAGGGATAAA AGGCAGAGCT AGATTAGTTT CAGCATCATA
Rabbit GTTCAGGACT TGGGCATAAA AGGCAGAGCA GGGCrAGCTG CTGCTTACAC

Maximum likelihood method writes the sequence sites with over 95% probability with upper
case, sites with 50-95% probability with lower case, and those with less than 50% probability
with an ambiguity code.

[with the parsimony method]

From To Any Steps? State at upper node

1 GTTCAGGGCT ?GGGCATAAA AGGCAGAGCA RGGY?AGCTD
1 2 yes GTCCAGGGCT -GGGCATAAA AGNCAGVGCA RGGYCAKCTR
2 3 yes GTCVACBGCT -?CACVDGHD TGNCAGCVCA DGGBCAKCAR
3 Frog yes CTCAACTTTG GCCATGGGTT TGACAGCACA TGATCGTCAG
3 Chicken yes GACGGCCGCT --CACCAGCG TGCTATCCCC ACGGGAGCAA
2 Human yes AGCCAGGGCT -GGGCATAAA AGTCAGGGCA GAGCCATCTA
1 Opossum yes GTTTGGGGGC CAGGGATAAA AGGCAGAGCT AGATTAGTTT
1 Rabbit yes GTTCAGGACT TGGGCATAAA AGGCAGAGCA GGGC-AGCTG

1 CTGCTTAMAM
1 2 no CTGCTTAMAM
2 3 yes CTGCTCAVAM
3 Frog yes CTGATCAACA
3 Chicken yes GAGCCCAGAC
2 Human yes TTGCTTACAT
1 Opossum yes CAGCATCATA
1 Rabbit maybe CTGCTTACAC

If the parsimony inferred state is "?" or an ambiguity code, there are multiple equally
parsimonious states; the user has to work these out by hand. In addition, "?" means that a
deletion might or might not have happened. “N” indicates a nondeleted base that is
ambiquous.

40
Statistical tests of trees

Statistical tests can be performed for both parsimony and maximum likelihood trees. The
tests are performed by putting multiple trees in the intree file. Actually, the tests are
automatically performed if there are multiple trees in the intree file:

(((Frog,Chicken),Human),Opossum,Rabbit);
(((Frog,Human),Chicken),Opossum,Rabbit);
(((Frog,Opossum),Human),Chicken,Rabbit);
(((Frog,Chicken),Rabbit),Opossum,Human);
(((Frog,Rabbit),Human),Opossum,Chicken);

To perform the test, load the sequence data, and invoke option U (No, use user trees in input
file ) in a parsimony or maximum likelihood program.

Parsimony programs, for example Dnapars, perform Templeton’s test if there are two trees to
compare, but uses Shimodaira-Hasegawa’s method for more than two trees. Shimodaira-
Hasegawa’s method uses a resampling method to correct the test results for multiple
comparisons. This resampling asks for a random number when performed.

This test finds the best trees among the competing hypothesis. Consult the column
“Significantly worse?”. If it states “No” the tree getting the best score is not significantly
different from the tree it was compared to. Thus, it is not possible to pick the best of trees 3-5
on basis of this test:

Tree Steps Diff Steps P-value Significantly worse?

1 3134.0 38.0 0.001 Yes

2 3194.0 98.0 0.009 Yes
3 3096.0 <------ best
4 3107.0 11.0 0.141 No
5 3124.0 28.0 0.221 No

Maximum likelihood programs perform a Kishino-Hasegawa test for two trees, and
Shimodaira-Hasegawa’s test for more than two trees.

This test finds the best trees among the competing hypothesis. Consult the column
“Significantly worse?”. If it states “No” the tree getting the best score is not significantly
different from the tree it was compared to. Thus, it is not possible to pick the best of trees 1,
4, and 5 on basis of this test:

Tree Ln L Diff Ln L P-value Significantly worse?

1 -9711.73593 -0.49089 0.139 No

2 -9738.53093 -27.28588 0.003 Yes
3 -9730.79910 -19.55405 0.001 Yes
4 -9711.24505 <------ best
5 -9725.82664 -14.58159 0.449 No

41
LogDet-distance

LogDet-distance has been developed to account for the base frequency differences between
lineages. However, the LogDet-distance does not give a reliable tree, when there are large
rate differences between sites in the sequence.

LogDet-distances would be usable when the base frequencies in different lineages are not
constant. In such cases, LogDet distances often outperform maximum likelihood method.

How to test for base frequency heterogeneity? Dnaml writes a table of empirical base
frequencies in the current dataset. The downside is that these frequencies can not be
computed for only two sequences. Here is an example of base frequencies calculated for three
different sets of three taxa:

A 0.28378
C 0.19595
G 0.31081
T(U) 0.20946

A 0.27027
C 0.23649
G 0.27703
T(U) 0.21622

A 0.25850
C 0.29252
G 0.26531
T(U) 0.18367

There are some differences between lineages, but not too large (about 5%), and the usual
distance method should do fine. However, this estimation approach fast becomes complicated
when the number of sequences goes up. One limitation of the LogDet distance is that it may
sometimes be infinite, if there are too many changes between certain pairs of nucleotides.
This can be particularly noticeable with distances computed from bootstrapped sequences.

Computing topological distances between trees

Topological distances are calculated with the program treedist. The symmetric distance
introduced here does not consider the branch lengths, only the tree topologies. The symmetric
distance between two trees is a count of partitions present in the other but not in the another
tree.

To calculate the symmetric distance, you need a standard intree as an input:

(((Frog,Chicken),Human),Opossum,Rabbit);
(((Frog,Human),Chicken),Opossum,Rabbit);
(((Frog,Chicken),Human),Opossum,Rabbit);
(((Frog,Opossum),Human),Chicken,Rabbit);
(((Frog,Chicken),Human),Opossum,Rabbit);
(((Frog,Chicken),Rabbit),Opossum,Human);
(((Frog,Chicken),Human),Opossum,Rabbit);
(((Frog,Rabbit),Human),Opossum,Chicken);

42
The treedist writes a menu:
Tree distance program, version 3.66

Settings for this run:

D Distance Type: Branch Score Distance
O Outgroup root: No, use as outgroup species 1
R Trees to be treated as Rooted: No
T Terminal type (IBM PC, ANSI, none): ANSI
1 Print indications of progress of run: Yes
2 Tree distance submenu: Distance between adjacent pairs

Are these settings correct? (type Y or the letter for one to change)

Invoking the option "2" let's you to specify which way the trees are compared. Choosing
Option P calculates the pairwise topological distances for all the trees in the file intree.

Tree Pairing Submenu:

A Distances between adjacent pairs in tree file.
P Distances between all possible pairs in tree file.
C Distances between corresponding pairs in one tree file and other.
L Distances between all pairs in one tree file and another.

Choose one: (A,P,C,L)

p

Then the program ask what kind of a matrix to return:

Distances output options:

F Full matrix.
V One pair per line, verbose.
S One pair per line, sparse.

Choose one: (F,V,S)

f

As a last step, you need to specify the distance type (Option D):

Tree distance program, version 3.66

Settings for this run:

D Distance Type: Symmetric Difference
O Outgroup root: No, use as outgroup species 1
R Trees to be treated as Rooted: No
T Terminal type (IBM PC, ANSI, none): ANSI
1 Print indications of progress of run: Yes
2 Tree distance submenu: Distances between all possible
pairs in tree file.

Are these settings correct? (type Y or the letter for one to change)

The distances are calculated for all pairs of trees. Results are by default saved in the file
outfile:

43
Tree distance program, version 3.66

Symmetric differences between all pairs of trees in tree file:

1 2 3 4 5 6 7 8
\------------------------------------------------
1 | 0 2 0 4 0 2 0 4
2 | 2 0 2 4 2 4 2 4
3 | 0 2 0 4 0 2 0 4
4 | 4 4 4 0 4 4 4 4
5 | 0 2 0 4 0 2 0 4
6 | 2 4 2 4 2 0 2 4
7 | 0 2 0 4 0 2 0 4
8 | 4 4 4 4 4 4 4 0

Program treedist is handy when there are multiple trees, for example, equally parsimonious
trees, and pairwise comparisons need to be made fast. As is obvious from the example above,
the pairwise distance of 2 means one swapping of two terminal branches (two sequences).

Weighting

Weights can be used to analyze different subsets of characters (by weighting the rest as zero).
For example, it might be of interest to compare the trees inferred from the first, second and
the third codon positions. This can be done using the weights. The weights are saved in a file
named "weights". The file should contain a text string of weights, one weight given to every
sequence alignment position. For example, the weight-pattern given below will use only the
third codon position for inferring the tree:

001001001001001001001001001001001001001001001001001

The weights can continue on several lines, and blanks between the lines will be ignored.

Weights are also handy, if you want to analyze different parts of the sequences, for example,
only conserved areas of the protein. You don't need to edit the original data files if you just
create new weights.

You can also check for the uniformity of the substitution rates of different codon positions by
using weights. Create three weight files, where you specify to include only the first, second
or the third codon positions in the analysis of a 9-site data set there might be:

First positions: 100100100

Second positions: 010010010
Third positions: 001001001

These weight specifications must be exactly as long as your sequence is. Otherwise you will
get into trouble. Then run the program Dnapars with the same material using the option "w"
once with every weight-specifications. Use the topology inferred without weights as user tree
(invoke Option U), and note the number of changes:

44
All positions: 49 changes
First positions: 20 changes
Second positions: 12 changes
Third positions: 17 changes

It now becomes visible that there are different numbers of changes in different codon
positions. It seems that the second position has fewer changes than other positions. This is a
reasonable result, because substitutions in the second codon position more often lead to an
amino acid change than substitutions in the first position.

In the example, the first and third codon position have nearly the same number of changes. It
also reasonable to expect that the third codon position will have more changes than the first,
but this was not supported by the data in our example.

Why should we check for unequal rate of substitution in different positions? One reason is to
check whether there are some evolutionary constraints for substitutions in certain codon
positions. Another reason is to check for saturation of the substitutions in different codon
positions. If there are large differences between the number of changes, especially, if the
number of substitutions in the third codon positions is high, saturation of that position is a
likely explanation. In the case of saturation, DNA sequences that include third position may
give erroneous results, and it could be better to use protein sequences or the DNA sequence
consisting of only the first and the second codon positions.

Another method for testing for the sequence site rate heterogenity is to calculate pairwise
distances using the different codon sites (program Dnadist):

First positions:
5
Rabbit 0.0000 0.1578 1.4088 0.5954 0.8499
Human 0.1578 0.0000 0.8327 0.4690 4.2225
Opossum 1.4088 0.8327 0.0000 0.5029 5.9886
Chicken 0.5954 0.4690 0.5029 0.0000 3.3429
Frog 0.8499 4.2225 5.9886 3.3429 0.0000

Secon positions:
5
Rabbit 0.0000 0.0754 0.2923 0.6328 0.8892
Human 0.0754 0.0000 0.1695 0.4178 0.6464
Opossum 0.2923 0.1695 0.0000 0.2996 0.7112
Chicken 0.6328 0.4178 0.2996 0.0000 0.6150
Frog 0.8892 0.6464 0.7112 0.6150 0.0000

Third positions:
5
Rabbit 0.0000 0.4062 0.5623 0.8167 1.2231
Human 0.4062 0.0000 0.2666 0.4173 1.0609
Opossum 0.5623 0.2666 0.0000 0.1644 0.8987
Chicken 0.8167 0.4173 0.1644 0.0000 1.3639
Frog 1.2231 1.0609 0.8987 1.3639 0.0000

After calculating the pairwise distances, we can check whether the number of substitutions
seems to differ between codon position, which it in our example most certainly does.

Note that, if the species are not closely related (for example, human and frog) the sequences
might have more amino acid changes than silent substitutions.

45
Dnaml, HMM, gamma distribution and rate heterogeneity

Program Dnaml (and Dnamlk, Proml and Promlk) implements two different layers of base
substitution rate heterogenity:

Settings for this run:

U Search for best tree? Yes
T Transition/transversion ratio: 2.0000
F Use empirical base frequencies? Yes
C One category of sites? Yes
R Rate variation among sites? constant rate
W Sites weighted? No
S Speedier but rougher analysis? Yes
G Global rearrangements? No
J Randomize input order of sequences? No. Use input order
O Outgroup root? No, use as outgroup species 1
M Analyze multiple data sets? No
I Input sequences interleaved? Yes
0 Terminal type (IBM PC, ANSI, none)? IBM PC
1 Print out the data at start of run No
2 Print indications of progress of run Yes
3 Print out tree Yes
4 Write out trees onto tree file? Yes
5 Reconstruct hypothetical sequences? No

The first layer models the rate heterogeneity by a hidden Markov model (HMM). This is
done automatically.

“HMM allows us to specify with option "c" to the program that there will be a number of
different possible evolutionary rates, what the prior probabilities of occurrence of each is, and
what the average length of a patch of sites all having the same rate. The program then
computes the likelihood by summing it over all possible assignments of rates to sites,
weighting each by its prior probability of occurrence. There is also a possibility to set that the
rates in adjacent sites are correlated with each other with option "a".”

“Another layer of rate variation is also available. The user can assign categories of rates to
each site (for example, we might want first, second, and third codon positions in a protein
coding sequence to be three different categories. This is done with the categories input file
and the C option. We then specify (using the menu) the relative rates of evolution of sites in
the different categories. For example, we might specify that first, second, and third positions
evolve at relative rates of 1.0, 0.8, and 2.7.” (Felsenstein, PHYLIP documentation)

There is also a possibility to change the gamma-distribution shape parameter with option "r".
At the moment the gamma-parameter can not be directly estimated, but it can be inferred by
an iteration method described above in the section Estimating the transition/transversion
ratio. After the iteration, the best tree can be compared to the tree without rate heterogenity
by the likelihood ratio test with df=1 (see, Testing molecular clock) or by Kishino-Hasegawa
test using user trees (see, Statistical tests of trees).

“If both user-assigned rate categories (with categories file) and regional rate variation (the
Hidden Markov Model rates) are allowed, the program assumes that the actual rate at a site is
the product of the user-assigned category rate and the Hidden Markov Model regional rate.
(This may not always make perfect biological sense: it would be more natural to assume

46
some upper bound to the rate, as we have discussed in the Felsenstein and Churchill paper).
Nevertheless you may want to use both types of rate variation.” (Felsenstein, PHYLIP
documentation)

Multiple outgroups

“It's not a feature but is not too hard to do in many of the programs. In parsimony programs
like mix, for which the “w” (weights) and “a” (Ancestral states) options are available, and
weights can be larger than 1, all you need to do is:”

(a) In mix, make up an extra character with states 0 for all the outgroups and 1 for all the
ingroups. If using Dnapars the ingroup can have (say) "G" and the outgroup "A".
(b) Assign this character an enormous weight (such as Z for 35) using the “w” option, all
other characters getting weight 1, or whatever weight they had before.
(c) If it is available, Use the “a” (Ancestral states) option to designate that for that new
character the state found in the outgroup is the ancestral state.
(d) In mix do not use the “o” (Outgroup) option.
(e) After the tree is found, the designated ingroup should have been held together by the fake
character. The tree will be rooted somewhere in the outgroup (the program may or may not
have a preference for one place in the outgroup over another). Make sure that you subtract
from the total number of steps on the tree all steps in the new character.

“In programs like Dnapars, you cannot use this method as weights of sites cannot be greater
than 1. But you do an analogous trick, by adding a largish number of extra sites to the data,
with one nucleotide state ("A") for the ingroup and another ("G") for the outgroup. You will
then have to use Retree to manually reroot the tree in the desired place.” (Felsenstein,
PHYLIP documentation)

Multiple outgroups as described above cannot be used with maximum likelihood programs.

Error messages

Here are some of the most commonly encountered error messages, and what to do to correct
them. The first three are not actually error messages at all, but an essential part of the normal
function of the programs.

1. Can’t find infile

Dnapars.exe: can't find input file "infile"

Please enter a new file name>

There is no file named infile in the folder you are running the program from, and the
program asks for the name of the input file. This is easily corrected: Type in the name of the
input file (sequence alignment file).

47
2. Outfile already exists

Dnapars.exe: the file "outfile" that you wanted to

use as output file already exists.
Do you want to Replace it, Append to it,
write to a new File, or Quit?
(please type R, A, F, or Q)

There is already an outfile in the same folder you are running the program from. You
have to decide whether to replace (overwrite) it, append it, or quit. You can also specify a
new outfile name.

3. Outtree already exists

Dnapars.exe: the file "outtree" that you wanted to

use as output tree file already exists.
Do you want to Replace it, Append to it,
write to a new File, or Quit?
(please type R, A, F, or Q)

There is already an outtree in the same folder you are running the program from. You
have to decide whether to replace (overwrite) it, append it, or quit. You can also specify a
new outtree name.

4. Wrong data type

ERROR: a function asked for an inappropriate amount of memory: -4 bytes

This can mean one of two things:
1. The input file is incorrect (perhaps it was not saved as Text),
2. There is a bug in the program.
Please check your input file carefully.
If it seems to be a bug, please mail [email protected]
with the name of the program, your computer system type,
a full description of the problem, and with the input data file.
(which should be in the body of the message, not as an Attachment).
Hit Enter or Return to close program.
You may have to hit Enter or Return twice.

There is probably an infile in the same folder you are running the program from. The
problem is that this infile is in a format that the current program can’t use. For example, you
might have renamed a distance matrix as infile when creating Neighbor joining trees. You
might then be using Dnapars program, which does not know how to read a distance matrix,
because it expects to find sequences in the file. Rename or remove the infile, and the error
should disappear.

5. Wrong program

ERROR: bad base: F at site 1 of species 1

Hit Enter or Return to close program.
You may have to hit Enter or Return twice.

48
Or:

The infile is of the wrong type

Hit Enter or Return to close program.
You may have to hit Enter or Return twice.

Or:
Unexpected End of File
Hit Enter or Return to close program.
You may have to hit Enter or Return twice.

You are probably trying to use the wrong kind of data in the current program. This error
message is related to the sequence type: you have used amino acid sequences as input in the
DNA sequence analysis program. Use the correct input file, and the error should disappears.

Scripting

Scripting can be used for automating some analyses, when needed. It is especially attractive
in UNIX / Linux system where jobs can be submitted to a queue. It is also a good idea in
Windows / DOS environment, for example, if several analyses need to be run over the
weekend: script can do the analysis during weekend and organize the results so that they are
easily checked by human eye on Monday morning. Scripting means that a file, which
contains a list directives is created, and instead of running the individual programs, the script
is executed.

Scripting in UNIX, Linux and Mac OS X

Scripting in UNIX is much simpler than in DOS. Let’s create an example script which runs
Dnaml analysis for the dataset alveolata.phy. First, we need to find out, which are the
commands and options we need for running Dnaml. We can do this by first doing a test run
of the program. Let’s also start the work in an empty folder, where outfiles or outtrees are not
present from previous analyses, but which contains the sequence alignment file and the
appropriate PHYLIP program (here Dnaml). When running, Dnaml first asks for a filename,
and then the menu appears. We want to use taxon 8 as an outgroup in this example, so we
invoke option O, and give it a number 8. After that, we want to run the analysis, which starts
in Dnaml by typing Y. Now our script looks like following:

#!bin/csh/
dnaml <<EOF
alveolata.phy
o
8
y
EOF

Save the script in a file called batch (you can modify this). Give the batch file execute
permissions (type chmod u+x batch), and submit it for running by typing, e.g., source
batch in the command prompt.

49
Scripting in DOS

Let’s next do the same analysis using scripts in DOS. In DOS we need two files, the batch
file, which starts the run and another file, which contains all the options to be used as input to
Dnaml. In DOS batch files have an identifier .bat, and make sure you save the file with this
extension. Otherwise DOS will not run the script at all!

So, first we create the file batch.bat in the empty folder containing only the sequence
alignment and the program Dnaml. The file contains just one line, which tells the computer to
run Dnaml from the same folder, and that file input.txt contains the options for the Dnaml
run. Batch.bat looks like following:

dnaml < input.txt

If you would like to save the text Dnaml normally writes to the screen and have it appear in a
file called screenout.txt, use the following line instead:

dnaml < input.txt > screenout.txt

Now the file input.txt contains the same sequence of options as the batch file created for
UNIX analyses, and it looks as follows:

alveolata.phy
o
8
y

You can start the Dnaml -script in DOS by double clicking on its icon.

More advanced scripts

I like using scripts in DOS for doing multiple analyses, because it is easier to modify the
batch-file than to run several programs separately. The following DOS script runs Dnaml for
the dataset (sequence alignment) alveolata.phy, calculates topological distances between all
the best trees (see Computing topological distances between trees) and finally performs the
Shimodaira-Hasegawas’s test (see above the section Statistical tests of trees).

The file batch.bat looks like this:

dnaml < input1.txt > screenout1.txt

copy outtree parsimony-tree-outtree.txt
rename outtree intree
rename outfile parsimony-tree-outfile.txt
treedist <input2.txt > screenout2.txt
rename outfile treedists-outfile.txt
dnaml < input3.txt > screenout3.txt
rename outfile SH-outfile.txt

The settings for individual programs are in the files input1.txt, input2.txt and input3.txt.

50
Input1.txt:
alveolata.phy
O
8
Y

Input2.txt:
2
P
F
Y

Input3.txt:
alveolata.phy
U
O
8
11
Y

After creating these four files, start the run by double-clicking on the batch.bat. The results
appear in four separate files: parsimony-tree-outtree.txt and parsimony-tree-outfile.txt (trees
from initial parsimony analysis), treedists-outfile.txt (topological distances), and SH-
outfile.txt (Shimodaira-Hasegawas’s test).

Recommendations

These recommendations cover some aspects of the actual phylogenetic data analysis that
were not discussed in the examples above. You should adapt the data analysis
recommendations here, because they highlight some of the most commonly encountered
problems.

Some pragmatic warnings

1. Always use several outgroups, if possible. This way you can check that all outgroups
really are outgroups.
2. Each method has different inherent weaknesses, and it might be a good idea to try
several methods, because they have strengths in different areas. Try parsimony,
maximum likelihood, and minimum evolution with LogDet-distances, and compare
the results. If all the methods produce more or less the same tree, then your data
apparently doesn’t have any major pitfalls. Hint: you can compare different trees
visually, but also using program treedist.

“Maximum parsimony can be mislead if there is too much heterogeneity in substitution rates among
lineages (the classic "long edges attract" problem) in the underlying true phylogeny. Minimum
evolution using LogDet distances can be mislead if there is too much site-to-site rate heterogeneity, or
if some of the pairwise distances are undefined (use the "showdist" command to check). Maximum
likelihood under the HKY-gamma model can be mislead if parameters that are assumed to be constant

51
 across the phylogeny (such as the tratio or base frequencies) actually vary among lineages in the true
phylogeny.” (David Swofford, PAUP FAQ, http://paup.csit.fsu.edu/paupfaq/faq.html)

“For example, if there is strong rate heterogeneity in your data (let's say the shape parameter is
estimated to be 0.01), then the LogDet and parsimony trees fall under a certain degree of suspicion
compared to the likelihood tree, which should be relatively immune to this pitfall since the model used
allows for rate heterogeneity. If the parsimony tree differs from the LogDet and likelihood tree, look
for evidence of long branch (edge) attraction in the parsimony tree. If the LogDet tree differs from the
parsimony and likelihood trees, see if the base frequencies vary considerably between tip taxa (a useful
tool for this purpose is the basefreq command).” (David Swofford, PAUP FAQ,
http://paup.csit.fsu.edu/paupfaq/faq.html)

3. Always bootstrap or jackknife your dataset.

4. Try excluding one taxon from the analysis. If the result changes dramatically, then the
excluded taxon causes systematic errors. In ideal case, the trees would be identical
before and after exclusion.
5. If you’re analyzing several genes, different genes might or might not produce the
same results. If the results are discordant, then check your data for non-orthologous
genes (if you’re interested in species phylogenies), because they might cause trouble.
Other sources of systematic errors, like lateral or horizontal gene transfer may
confound the analysis, also.

52
PHYLIP programs

PHYLIP programs classified according to the material and method

Maximum Parsimony Distance Lake's Reliability

likelihood matrix invariants

DNA Dnaml*** Dnapars Dnadist* Dnainvar Dnacomp

Dnamlk**** Dnapenny Fitch*** Seqboot
Dnamove***** Kitsch****
Neighbor**

Protein Proml Protpars Protdist* Seqboot

Promlk Fitch***
Kitsch****
Neighbor**

Restriction Restdist* Seqboot

sites Fitch***
Kitsch****
Neighbor**

Continuous Contml Fitch*** Seqboot

characters Kitsch**** Contrast
Neighbor**

Discrete Pars Clique

characters Mix
Penny
Move*****
Dollop
Dolpenny
Dolmove*****
Factor

Tree drawing Drawgram

Drawtree
Consense
Treedist
Retree

* utilized for matrix calculation

** NJ and UPGMA
*** FM without molecular clock
**** FM with molecular clock
***** interactive tree building

53
Flow charts

The following data flow charts describe some basic analyses. Flow chart A describes a
maximum likelihood analysis for DNA sequences. Flow chart B describes an analysis using
Neighbor joining method for DNA sequences. Flow chart C describes a bootstrapping
analysis for DNA sequences using maximum likelihood as the analysis method. Note that
you might need to rename the files between the analysis steps. For example, the outfile from
Dnadist in flowchart B cannot be directly used in Neighbor. You should rename the outfile
first, for instance, to infile.

In order to assess the reliability of the data using the bootstrapping method, you should first
make the conventional analysis (flow chart A) using whatever analysis method is suitable for
your purposes. After that, you should perform bootstrapping analysis (flow chart C) using
exactly the same analysis method you used for conventional analysis. Here, we have used
program Dnaml for maximum likelihood analyses, but it can be replaced with, e.g., Dnapars
for parsimony analysis or Dnadist + Neighbor for analysis using distance methods. Note that
in bootstrapping analysis using Dnaml, Dnapars etc. it might not be advisable to use as many
jumbles as in the conventional analysis, because the analysis programs will then perform a
specified number of jumbles for every random sequence dataset, and that might take a very
long time. A single jumble per bootstrap dataset is probably enough.

54
PHYLIP is freely available from http://evolution.genetics.washington.edu/phylip.html.

PHYLIP -- Phylogeny Inference Package (Version 3.2).

Felsenstein, J. 1989. Cladistics 5: 164-166

55