EMA - Easy Microarray data Analysis

N. Servant 1,2,3 , E. Gravier 1,2,3,4 , P. Gestraud 1,2,3 , C. Laurent 1,2,3,6 , C. Paccard 1,2,3 ,
A. Biton 1,2,3,5 , I. Brito 1,2,3 , J. Mandel 1,2,3 , B. Asselain 1,2,3 , E. Barillot 1,2,3 , P. Hupé 1,2,3,5 .
1
Institut Curie, Paris F-75248, France,
2
INSERM U900, Paris F-75248, France,
3
Ecole des Mines ParisTech, Fontainebleau, F-77300 France,
4
Institut Curie, Deṕartement de Transfert, Paris F-75248, France,
5
CNRS, UMR144, Paris F-75248, France
6
CNRS, UMR146, Paris F-75248, France

[email protected] - http://bioinfo.curie.fr/projects/ema/

March 21, 2012

Contents
1 Introduction 2

2 Data Input 2

3 Quality Control 2

4 Data pre-processing 3
4.1 Normalisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
4.2 Data filtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

5 Exploratory Analysis 5
5.1 Principal Component Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
5.2 Agglomerative Hierarchical Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . 8

6 Differential Analysis 9

7 Gene annotation 10

8 Gene-Set analysis 11

9 Functional enrichment assessment 11

10 Supervised Classification 12

11 Survival analysis 12

12 System Performance 13

13 Session Info 14

1
1 Introduction
The increasing number of methodologies and tools currently available to analyse gene expression
microarray data can be confusing for non specialist users. Based on the experience of biostatisticians
of Institut Curie, we propose both a clear analysis protocol and the tools to investigate the data. It
provides a useful starting point for many microarrays users.
All the functions included in the powerful and free EMA (Easy Microarray data Analysis) package
were tested and chosen according to their performance and their suitability.
Most of these functions have also been improved to facilitate their use, the visualization and the
interpretation of the results in the specific field of microarray data analysis.
EMA supports entire analysis process for any type of gene expression arrays, from normalisation,
unsupervised and supervised analysis to functional analysis. Exploratory analyses are also proposed
in the context of censored data.
Below, we present the analysis strategy used in the Institut Curie for a class comparison study. This
strategy is applied to a cancer microarrays dataset (Marty et al. (2008) - GSE13787) which com-
pares 12 Basal-like breast carcinomas (BLCs) and 11 HER2 positive breast carcinomas (HER2+).
The mRNA profiles of these 23 samples were analysed using HG-U133plus2 Affymetrix chips that
measure the expression values of 54613 probesets (after removal of AFFX probesets).
The aim is to detect differentially expressed genes between the two groups of tumours, in order to
discover potential therapeutic targets in human BLCs. A brief application to simulated survival
data is also provided.

2 Data Input
The main functions of the EMA package work with an expression matrix as input (expression levels
-from e.g genes- measured in tissue samples and log2 transformed). Based on this simple input, the
EMA package can be used for any gene expression data such as Affymetrix data (HG-U133a/b, HG-
U95av2, HG-U95a/b/c/d/e, HG-U133plus2, MouseGenome4302/430a2, MurineGenome11k/19k),
one-colour experiments like Illumina data (HumanWGv1/2/3, MouseWG1/2, RatRSv1), two-colour
data or exon data (HuEx1.0stv2). The annotation data must be available from the BioConductor
project (Gentleman et al. (2004)) or the BrainArray website (Dai et al. (2005)) to work with data
annotation and to perform the functional analysis. Some of the EMA functions can also be used
for other applications like miRNAs expression data. The only limiting step which depends on the
providers and the data type is the normalisation (see section 4.1).

3 Quality Control
Data quality control is a major concern in microarray analysis. This step aims to detect problematic
raw probe-level data (array with spatial artifacts or with poor RNA quality for example) to facilitate
the decision of whether to remove this array from further analysis. However, raw probe-level data
artefacts can also be removed thanks to normalisation procedures (see section 4.1).
This is why we recommend to combine the results obtained by both the quality control step and
the normalisation process to make a decision. For quality assessment, we recommend to use the
arrayQualityMetrics package (Kauflmann et al. (2009)) which proposes powerful and comprehensive
tools and an automatic report generation. This package can be used for Affymetrix datasets as well
as for two-colour or non Affymetrix one colour experiments.

2
4 Data pre-processing
4.1 Normalisation
Microarray normalisation is a fundamental step which aims at removing systematic bias and noise
variability caused by technical and experimental artefacts.
This is also the only step which depends on the provider and the data type. For each type of arrays,
the raw data files are different and need a specific pre-processing.
Several packages already exist to normalise Illumina, one or two-colour data.
The lumi package (Du et al. (2008)) supports the Illumina raw data output of the Illumina Bead
Studio toolkit and can easily normalise them.
> ## ILLUMINA CHIPS
> ## From lumi1.14.0 help page
> ## load example data
> require(lumi)
> data(example.lumi)
> ## Do all the default preprocessing in one step
> lumi.N <- lumiExpresso(example.lumi)
> ## Convert data into matrix for EMA
> lumi.mat <- exprs(lumi.N)
The vsn (Huber et al. (2002)) package can be used to normalise data from spotted two-colour or one-
colour microarrays as described in the following example. It also includes normalisation techniques
for Affymetrix arrays.
Other packages like limma Smith et al. (2003) or more recently BABAR (Alston et al. (2010)) also
provide several normalisation techniques for these data.

> ## cDNA CHIPS

> ## From vsn3.16.0 help page
> ## VSN normalisation
> require(vsn)
> data(lymphoma)
> lym <- justvsn(lymphoma)
> ## Convert data into matrix for EMA
> lym.mat <- exprs(lym)
EMA offers a function to normalise the Affymetrix GeneChip data using MAS5.0 (Affymetrix
Inc. (2002)), RMA (Irizarry et al. (2003)) or GCRMA (Wu et al. (2004)) methods.
We recommend the use of GCRMA method because it outperforms MAS5.0 by ignoring the MM
intensities (observed to detect some specific signal) and it improves RMA by taking into account
the sequence information to describe background noise. However, GCRMA can be more time
and memory consuming for a large number of arrays. In that case, RMA could be an interesting
alternative. The performances of the GCRMA method are shown in the ”System performance”
section.
> ##Load EMA package
> require(EMA)
> ## GCRMA Normalisation
> ## Not run because cel files are not available from this package
> cel.path=paste(getwd(),"/Data/E-GEOD-13787", sep="")
> marty<-normAffy(celfile.path=cel.path, method="GCRMA")

3
4.2 Data filtering
The aim of the filtering step is to discard the probesets with very low expression across the samples
(and that provide no biological information) in order to reduce noise in data and to avoid wrong
interpretations of the final results.
In differential analysis for example, the number of hypothesis to test is therefore reduced (which is
very useful in the multiple testing context).
Another problem also occurs in case of data scaling (for example in PCA) where the biological signal
is completely shrinked by the probesets with very low variability.
Most of the probesets with a low variability have either a very low expression level or, on the
contrary, have reached their saturation level. However, as shown in the following histogram, those
with a low expression level are much more frequent. That’s why by discarding the probesets with
a low expression level for all the samples, and with a low variability, we expect to easily reduce the
problem of noise and data scaling. Other common filtering methods include filtering by variance,
inter-quartile range (IQR) or MAS detection call (for Affymetrix arrays).

> ##Load GCRMA Normalised data

> ##marty.type.cl=Her2+ corresponds to Her2+ breast cancer
> ##marty.type.cl=Basal corresponds to Basal-Like carcinoma
> data(marty)
> ##And discard probesets with a maximum log2 expression value below 3.5
> marty.f<-expFilter(marty)

Keep probes with at least 1 sample(s) with an expression level higher than 3.5

> dim(marty.f)

[1] 33409 23

Data distribution
80000
60000
Frequency

40000
20000
0

3.5

5 10 15

Expression Level

4
 By displaying a bimodal distribution, this histogram supports the choice to use the threshold
value of 3.5 that divides low and high expression values.

5 Exploratory Analysis
The aim of exploratory analysis is to explore the gene expression data structure. First, it is rec-
ommended to look at the distribution of arrays and/or genes to give a simple summary of the data
structure. Among exploratory techniques, PCA or clustering are widely used as they provide easy-
to-interpret graphical representations. They make it possible to detect potential outliers and/or
bias like batch effect or any non relevant effects. In addition, clusters of similar gene profiles and
similar sample profiles can also be identified.

5.1 Principal Component Analysis

The principal component analysis (PCA) is a standard tool for extracting relevant information from
a dataset. The goal is to reduce the dataset to a lower dimension and to reveal the underlying struc-
ture of the data. As mentioned in the filtering step, scaling probesets with very low variability can
introduce a bias in the results (see below), that is why we recommend to perform PCA after filtering.

> acp<-runPCA(t(marty.f), scale=FALSE,lab.sample=marty.type.cl,

+ plotSample=FALSE, plotInertia=FALSE)
> plotInertia(acp)

Inertia percentage of components

15

10

0
Dim1

Dim2

Dim3

Dim4

Dim5

> ## Individual map (axe 1 and 2)

> plotSample(acp,axes=c(1,2),lab=marty.type.cl)

5
 Sample representation
Axes 1 and 2
100
● Her2+
GSM346885 ● Basal
GSM346892
●

GSM346902
GSM346900
GSM346899 GSM346886
GSM346889
50

GSM346903
●
●
GSM346882
● ●●
●
●

GSM346891
Dim 2 (9.76%)

GSM346883
GSM346897 ●
GSM346898
GSM346894
●
●
0

●
● ●
GSM346884
●
GSM346887
GSM346893
● ●
GSM346895
GSM346904
−50

●●
GSM346890
GSM346896 ●
●
GSM346901
●
−100

GSM346888
●

−150 −100 −50 0 50 100 150 200

Dim 1 (18.77%)

> ## Or create a pdf report with selected plots

> runPCA(t(marty.f), scale=FALSE, pdfname="PCA.pdf",lab.sample=marty.type.cl)

The two groups of tumours (HER2+ in red and BLCs in blue) are clearly separated by their
gene expression profile.
In the following example, we show how the filtering step affects the PCA results. In the case of
scaled data, the weights associated to genes with low expression across samples become high and
can lead to some wrong interpretations.

6
 Sample representation
Axes 1 and 2

Her2+
200
●

● Basal

GSM346884
●

GSM346888
100

GSM346891 GSM346882 GSM346890

●
Dim 2 (8.82%)

●
GSM346883 ●
●

●
GSM346892 GSM346887
GSM346893
GSM346889
GSM346886 ●
● ●
●
GSM346885 ●

●
0

GSM346902
GSM346903
GSM346899 GSM346898 ●
GSM346901
GSM346896
GSM346900GSM346894 GSM346897 ●
●
●
GSM346904
● GSM346895
● ●● ●
−100

●
●

−300 −200 −100 0 100 200

Dim 1 (30.57%)

Variable Representation on axes 1 and 2

1.0

● High
● Low
0.5
Dim 2 (8.82%)

0.0

●
−0.5
−1.0

−1.5 −1.0 −0.5 0.0 0.5 1.0 1.5

Dim 1 (30.57%)

The main information provided by the data (summarised by axe 1) is not associated with the
two groups of tumours. Only Axe 2 provides this information. It can be shown that Axe 1 is
associated with the low expressed genes (high scaled signal level) rather than biological signal.

7
5.2 Agglomerative Hierarchical Clustering
As the PCA, the agglomerative hierarchical clustering is a useful method to identify clusters or to
highlight experimental bias.
Clustering algorithm is based on the determination of two distances.
The first (called ’metric’) defines the similarity between two elements. The second (called ’linkage
criterion’) define the similarity between two clusters of elements.
The choice of these distances is very important and depends on the similarities we want to detect
(for example, the clustering of genes using the Pearson correlation or the Euclidean distance can
lead to very different results). This kind of algorithm has shown to be sensitive to perturbations of
the data. Thus, it is possible to assess the quality of the clustering method (based on a given metric
and linkage criterion) by measuring its stability under perturbation or resampling of the data.

> ## Sample Hierarchical Clustering (Pearson's correlation coefficient and Ward method)
> c.sample<-clustering(data=marty.f, metric="pearson", method="ward")
> clustering.plot(tree=c.sample, lab=marty.type.cl,
+ title="GCRMA Data - filtered")

GCRMA Data − filtered

Hierarchical Clustering : ward pearson
0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14

Her2+
Basal
ward linkage

GSM346896
GSM346904
GSM346901
GSM346895
GSM346899
GSM346903
GSM346902
GSM346897
GSM346894
GSM346900
GSM346898
GSM346884
GSM346890
GSM346888
GSM346882
GSM346886
GSM346889
GSM346892
GSM346885
GSM346883
GSM346891
GSM346893
GSM346887

A heatmap is a graphical representation to visualise the level of expression of genes across

the samples, and therefore allowing a better understanding of the clusters (genes and/or samples)
detected. To facilitate the visualisation and the interpretation of the results, we propose to perform
a gene selection (keep the genes with the highest interquartile range - IQR values) before performing
the heatmap.

8
> ## Heatmap performed on the 100 probesets with the highest IQR values
> mvgenes<-genes.selection(marty.f, thres.num=100)
> c.sample<-clustering(data=marty.f[mvgenes,], metric="pearson", method="ward")
> c.gene<-clustering(data=t(marty.f[mvgenes,]), metric="pearsonabs", method="ward")
> clustering.plot(tree=c.sample, tree.sup=c.gene, data=marty.f[mvgenes,],
+ names.sup=FALSE, lab= marty.type.cl, trim.heatmap=0.99)

Her2+ > 1.7

Basal

0.1

< −1.4

Hierarchical Clustering : ward pearsonabs

GSM346896
GSM346904
GSM346901
GSM346894
GSM346897
GSM346900
GSM346895
GSM346898
GSM346899
GSM346902
GSM346903
GSM346884
GSM346889
GSM346891
GSM346882
GSM346893
GSM346890
GSM346892
GSM346883
GSM346885
GSM346886
GSM346887
GSM346888

Hierarchical Clustering : ward pearson

The two groups of tumours (HER2+ in red and BLCs in blue) are clearly separated by their
gene expression profile, as also shown by the PCA.

>
> ### MFA
>
> ### BiClustering - BiCare package

6 Differential Analysis
A standard approach to detect Differentially Expressed Genes (DEG) is to test the difference be-
tween means of the two groups (with Student or Mann-Whitney tests for example) and then to
adjust for multiple testing by applying the Benjamini and Hochberg (BH) procedure (Benjamini
and Hochberg, 1995)

9
> ### Student test with BH correction
> marty.type.num <- ifelse(marty.type.cl=="Her2+",0,1)
> rt<-runTtest(marty.f, labels=marty.type.num, algo="t.equalvar", q=0.05, plot=FALSE)

[1] "typeFDR= FDR-BH"

> head(rt)

probeID Stat RawpValue AdjpValue

1 117_at 0.05985479 0.95283725 0.9758069
2 121_at 0.75142912 0.46073334 0.6481549
3 177_at 0.94302249 0.35639290 0.5661514
4 243_g_at -1.11306304 0.27826051 0.5100864
5 266_s_at -0.39507776 0.69676987 0.8213388
6 320_at 1.88783936 0.07294414 0.2481408

Detection of the DEG can also be performed by the Significance Analysis of Microarrays (SAM)
algorithm (Tusher at al., 2001) (with t-test as statistic test for example). This algorithm is less
stringent than the Benjamini and Hochberg procedure because it estimates the proportion of truly
not differentially expressed genes.
By using permutations this approach takes into account the correlations between genes and avoids
making assumptions about their distribution (which is usually problematic in microarray experi-
ments due to a small sample size). This is an advantage over other techniques (for example t-test
with Benjamini and Hochberg correction) which assume normal distribution with equal variance
and independence of genes. In addition, test statistics of genes with low variability tend to be large,
which increases false positives. SAM corrects these statistics by adding a small value (called the
’fudge factor’) to their denominator.
In case of more complex design experiments we also recommend the use of the limma package, very
powerful tool to assess differential expression by using linear models.

> rs<-runSAM(marty.f, labels=marty.type.num)

> head(rs)

7 Gene annotation
Working with probesets name is often not the easiest way to interpret the data in term of bi-
ology. Annotations, as gene symbol are often very useful to understand the results. BioMart
(http://www.biomart.org/) is a very powerful tool and gives direct access to several annotations
databases (Ensembl, Uniprot, etc). A R package dedicated to interface R and the BioMart databases
was created by Durinck S, Huber W, and Davis S.
EMA allows to query BioMart in a easiest way, without any knowledge of the biomaRt object. The
connexion and the proxy configuration have to be checked before running biomaRt.

> ##Look at the annotation of the DEG

> rt.sign <- rt[order(rt$AdjpValue),]
> rt.annot<-bioMartAnnot(rt.sign, inputTypeId ="affy_hg_u133_plus_2",
+ outputTypeId=c("entrezgene","hgnc_symbol"),
+ dataset=c("hsapiens_gene_ensembl"),database = "ensembl")

Annotations are an important issue of the microarray analysis. The probes used for the design of
the array relied on earlier genome and transcriptome annotation and can be different from current

10
knowledge. The first consequence of this problem is that many conclusions derived from analyses
may be significantly flawed.
Here, all the analyses were done with the Affymetrix annotation package, but the users can also
used their own probe set definition for their analysis.
EMA proposes a simple function to check the signal of the 11 probes summarized in each probeset.

> ## Not run because cel files are not available from this package
> filenames <- list.files("Data/E-GEOD-13787", pattern=".CEL", ignore.case=TRUE)
> rawdata <- ReadAffy(filenames=filenames, celfile.path="Data/E-GEOD-13787", cdfname=NULL)
> probePlots(rawdata, pbsList=rt.annot$affy_hg_u133_plus2[1:10])

8 Gene-Set analysis
Instead of working at the gene level, several recent studies discussed the problem of identifying
differentially expressed groups of genes from a set of microarray experiments.
One of the most famous approach is the Gene Set Enrichment Analysis (GSEA) procedure of Sub-
ramanian et al. (Subramanian et al. (2006)). Here we decided to use the GSA package, presented in
the paper as an improvement of the GSEA approach. It differs from a Gene Set Enrichment Analysis
(Subramanian et al 2006) in its use of the ”maxmean” statistic: this is the mean of the positive or
negative part of gene scores in the gene set, whichever is larger in absolute value. Moreover, GSA
does ”restandardisation” of the genes, in addition of the permutation of columns, to estimate the
false discovery rate.

> ## http://www.broad.mit.edu/gsea/msigdb/msigdb_index.html
> ## You have to register first and then download the gmt file from their site
> gsaOUT <- runGSA(marty, marty.type.num ,
+ gmtfile="c2.kegg.v2.5.symbols.gmt",
+ chip="hgu133plus2")

9 Functional enrichment assessment

The goal is to interpret the results provided by the differential analysis. One of the most classical
way to answer this question, is to assess if the DEG list is enriched for biologically relevant gene-sets
compared to a reference gene-set (called ’universe’, typically all the genes of the array). Functional
enrichment is based on the GeneOntology and KEGG pathways annotation terms.
The hyper-geometric test of the GOstats package (Falcon et al. (2007)) is used to test the over-
representation of the functional terms in the gene list.
The choice of the universe could have a significant impact on the results. It is well discussed in
the vignette of the GOstats package. Here, we decided to apply a non-specific filtering procedure
different from the one proposed by Falcon and Gentleman. Since not all genes will be expressed
under all conditions in our data, we can ask the question of defining the universe only with the
expressed genes or with all the genes of the array. Actually, we are not able to distinguish the genes
which are biologically non expressed, from the ones of low quality. That’s why we think that the
non-expressed probesets could be biologically relevant, as well as the ones with a little variation
across samples, and we decided to first defined the universe with all the genes of the array. Then,
we just remove probe sets that have no Entrez Gene identifier in our annotation data or no GO
annotation. Finally, the Hypergeometric test is performed on the unique EntrezId of the gene list,
and the unique EntrezId of the universe. The pvalues in output are not corrected from multiple
testing. Note that because of the existing dependence structure (between genes, and GO terms)

11
it is difficult to do any multiple testing correction. Moreover the most interesting genesets are not
necessarily the ones with the smallest pvalues. Nodes that are interesting are typically those with
a reasonable number of genes (10 or more) and small pvalues.
Another interesting method is provided by the globaltest package. It tests the null hypothesis
that no gene in the gene set is associated with the class label rather than the null hypothesis that
the genes in the gene set are not more associated with the class label than random genes on the
microarray (followed by GSEA or methods based on hyper-geometric test). The model used for
gene set testing is a regression model with random effects. It allows groups of genes of different size
to be tested (from a single gene to all genes of the chip).

> ## GO and KEGG analysis on the DEG by the SAM procedure

> runHyperGO(list=rownames(rt.sign), pack.annot="hgu133plus2.db", name="HyperGO_type")
> runHyperKEGG(list=rownames(rt.sign), pack.annot="hgu133plus2.db", name="HyperKEGG_type")

Two types of reports are generated. The first one is an HTML report (see GOStats pack-
age) which describes the significant genesets (Id, pvalue, Odd Ratio, Expected Count, Count, Size
and GO/KEGG term) The second (text format) describes the description of the content of each
significant genesets, with the associated genes.

10 Supervised Classification
The classification is a prediction or learning problem from which each object is characterised by
a response variable (class label) and a set of measurements (genes expression). One of the main
goal of the supervised classification, especially in microarray experiments, is to find the best rule to
classify/distinguish the objects.
One of the most famous application of this type of method is the gene signature published by Van’t
veer as a predictor for metastasis (Van’t Veer et al. (2002)). For all these approaches, we suggest to
use the CMA package (Slawski et al (2008)) including the most popular machine learning and gene
selection algorithms.

11 Survival analysis
Kaplan Meier and log-rank test analyses proposed by the EMA package helps for example to compare
survival between patients with low expression value and patients with high expression value of a
given gene.

> set.seed(5000)
> gene<-rnorm(100)
> gene[51:100]<-gene[51:100]+2
> group<-ifelse(gene<=median(gene),"Low gene expression","High gene expression")
> time<-abs(rnorm(100))
> time[51:100]<-time[51:100]+2
> status<-sample(c(0,1),size=100,replace=TRUE)
> res<-km(time,status,group,main="Kaplan Meier curve")
> res$fit.km

Call: survfit(formula = Surv(time, status) ~ group, na.action = na.omit)

records n.max n.start events median 0.95LCL 0.95UCL

group=High gene expression 50 50 50 23 2.85 2.643 NA
group=Low gene expression 50 50 50 27 1.34 0.845 NA

12
> res$lr

Call:
survdiff(formula = Surv(time, status) ~ group, na.action = na.omit)

N Observed Expected (O-E)^2/E (O-E)^2/V

group=High gene expression 50 23 33.4 3.22 10.2
group=Low gene expression 50 27 16.6 6.45 10.2

Chisq= 10.2 on 1 degrees of freedom, p= 0.00142

> res$p.lr

[1] 0.00142434

Kaplan Meier curve

1.0

High gene expression

Low gene expression
0.8
0.6
0.4

p=0.0014
0.2
0.0

0 1 2 3 4

Time (years)

12 System Performance
We tested the EMA package on a classical desktop (1 cpu Intel Xeon 3.16 Ghz) and the perfor-
mances were reported using the profr package.
The following table details how much CPU time each function used for the analysis workflow de-
scribed above.

13
 function time(s)
data 0.24
expFilter 0.9
runPCA 7.4
plotSample 0.04
runPCA 8.58
plotVariable 11.06
clustering 0.1
clustering.plot 0.2
genes.selection 8.46
clustering 0.02
clustering.plot 0.42
runTtest 4.58
runSAM 77.12
bioMartAnnot 1.06
runHyperGO 186.86
runHyperKEGG 14.1
km 0.08

We also compared the performances of the GCRMA and RMA functions on a dataset of 290
HG-U133a2 Affymetrix arrays.
The normalisation was run on 1 CPU (3 Ghz) with a 64 bits version of R.2.11.1.

function time(s)
RMA 350.961
GCRMA 481.055

As expected, GCRMA is more time computationally intensive than RMA on a large dataset. In
this case, RMA can be an interesting alternative.

13 Session Info
ˆ R version 2.15.0 beta (2012-03-19 r58787), x86_64-unknown-linux-gnu

ˆ Locale: LC_CTYPE=fr_FR.UTF-8, LC_NUMERIC=C, LC_TIME=fr_FR.UTF-8, LC_COLLATE=C,

LC_MONETARY=fr_FR.UTF-8, LC_MESSAGES=fr_FR.UTF-8, LC_PAPER=C, LC_NAME=C,
LC_ADDRESS=C, LC_TELEPHONE=C, LC_MEASUREMENT=fr_FR.UTF-8, LC_IDENTIFICATION=C

ˆ Base packages: base, datasets, grDevices, graphics, methods, splines, stats, utils

ˆ Other packages: AnnotationDbi 1.17.27, Biobase 2.15.4, BiocGenerics 0.1.14,

BiocInstaller 1.3.7, Category 2.21.2, DBI 0.2-5, EMA 1.3.2, FactoMineR 1.18,
GOstats 2.21.1, GSA 1.03, MASS 7.3-17, RSQLite 0.11.1, RankProd 2.27.1, affy 1.33.6,
biomaRt 2.11.2, cluster 1.14.2, ellipse 0.3-7, gcrma 2.27.2, graph 1.33.1, heatmap.plus 1.3,
lattice 0.20-6, multtest 2.11.4, rgl 0.92.858, scatterplot3d 0.3-33, siggenes 1.29.7,
survival 2.36-12, xtable 1.7-0

ˆ Loaded via a namespace (and not attached): Biostrings 2.23.6, GO.db 2.7.1,
GSEABase 1.17.3, IRanges 1.13.30, RBGL 1.31.1, RCurl 1.91-1, XML 3.9-4, affyio 1.23.2,

14
annotate 1.33.5, genefilter 1.37.0, grid 2.15.0, preprocessCore 1.17.7, stats4 2.15.0,
tools 2.15.0, zlibbioc 1.1.1

15
References
Affymetrix Inc., (2002) Statistical Algorithms Description Document.
http://www.affymetrix.com/support/technical/whitepapers.affx

Alston M.J., Seers J., Hinton J.CD., Lucchini S., BABAR: an R package to simplify the normalisa-
tion of common reference design microarray-based transcriptomic datasets, BMC Bioinformatics
2010, 11:73

Dai M., Wang P., Boyd A.D., Kostov G., Athey B., Jones E.G., Bunney W.E., Myers R.M., Speed
T.P., Akil H., Watson S.J., Meng F., Evolving gene/transcript definitions significantly alter the
interpretation of GeneChip data, Nucleic Acids Research, 33(20):e175.

Du P., Kibbe W.A., Lin S.M., lumi:a pipeline for processing Illumina microarray, Bioinformatics.
2008;24(13):1547-1548

Falcon S., Gentleman R, Using GOstats to test gene lists for GO term association, Bioinformatics.
2007 Jan 15;23(2):257-8

Huber W., Von Heydebreck A., Sültmann H., Poustka A. and Vingron M. (2002) Variance stabi-
lization applied to microarray data calibration and to the quantification of differential expres-
sion.Bioinformatics 18: suppl. 1 (2002), S96-S104 (ISMB 2002).

Gentleman, R. C. and Carey, V. J. and Bates, D. M. and Bolstad, B. and Dettling, M. and Dudoit,
S. and Ellis, B. and Gautier, L. and Ge, Y. and Gentry, J. and Hornik, K. and Hothorn, T. and
Huber, W. and Iacus, S. and Irizarry, R. and Leisch, F. and Li, C. and Maechler, M. and Rossini,
A. J. and Sawitzki, G. and Smith, C. and Smyth, G. and Tierney, L. and Yang, J. Y. and Zhang, J.,
Bioconductor: open software development for computational biology and bioinformatics, Genome
Biology, 2004, 5, R80.

Irizarry R.A., Hobbs,B. et al. (2003) Exploration, normalisation, and summaries of high density
oligonucleotide array probe level data, Biostatistics, 4, 249-254

Kauflmann,A., Gentleman R., Huber W. (2009) ArrayQualityMetrics-a bioconductor package for

quality assessment of microarray data, Bioinformatics,vol. 25, no3.pp. 415-416

Marty, B., et al. (2008) Frequent PTEN genomic alterations and activated phosphatidylinositol
3-kinase pathway in basal-like breast cancer cells,Breast Cancer Res., 10, R101.

Slawski M., Daumer M., Boulesteix A.L., CMA - A comprehensive Bioconductor package for super-
vised classification with high dimensional data, BMC Bioinformatics, Vol. 9, No. 1.

Subramanian A., et al. Gene set enrichment analysis: A knowledge-based approach for interpreting
genome-wide expression profiles, Proc. Natl. Acad. Sci. USA, 102, 15545-15550.

Smyth, G. K., and Speed, T. P. (2003). Normalization of cDNA microarray data. Methods 31,
265-273.

Van ’t Veer,L.J., et al. (2002) Gene expression profiling predicts clinical outcome of breast cancer,
Nature, 415, 530-536.

Wu,Z., Irizarry R.A., et al. (2004) A Model-Based Background Adjustment for Oligonucleotide
Expression Arrays, Journal of the American Statistical Association, 99, 909.

16