Microarray Data Analysis of Illumina Data Using R/Bioconductor Reddy Gali, Ph.D. rgali@hms.harvard.edu submit-c3-bioinformatics@rt.med.harvard.edu http://catalyst.harvard.edu

Agenda Introduction to microarrays Workflow of a gene expression microarray experiment Microarray experimental design Public microarray databases Microarray preprocessing - Quality control and Diagnostic analysis

Agenda Introduction to R/Bioconductor Installation of R and Bioconductor Packages General data analysis and strategies Data analysis using lumi package Data analysis using limma package 2

Workflow of Gene Expression Biological question Experimental design QC Tissue / sample preparation Extraction of Total RNA Probe amplification & labeling Microarray hybridization & processing Image analysis Data analysis Expression measures - Normalization - Statistical Filtering - Clustering - Pathway analysis Biological Verification

Pitfalls of Microarray Experiment Gene expression changes detected by microarray analysis cannot be validated by other methods - Inadequate design Data quality is low - Statistical approach is not adequate - Expression level of gene is below detection limit - Change in gene expression is small - Microarray detection probe is not specific or not sensitive 4

Questions usually asked What kind of technology or microarrays I have to use How many replicates do I need What is a real replicate Do I need statistical advice Should I do technical replicate Should I pool my samples How do I analyze my dataset What software should I use 5

Design of Microarray Experiment Replicates Goal, resources, technology, quality, design and analysis Two fold change – 3 replicates Smaller change – 5 replicates Technical replicates and Biological replicates Sample pooling Amount of sample Replicates of pooled sample No way to find variance between samples 6

Gene Expression Omnibus- GEO 7

Public Microarray Databases BodyMap - http://bodymap.ims.u-tokyo.ac.jp/ SMD - http://genome-www5.stanford.edu/ RIKEN - http://read.gsc.riken.go.jp/ MGI - http://www.informatics.jax.org/ GEO - http://www.ncbi.nlm.nih.gov/geo/ CIBEX - http://cibex.nig.ac.jp/index.jsp ArrayExpress - http://www.ebi.ac.uk/microarray-as/ae/ 8

Microarray Platforms Agilent Microarrays 60-mer format Codelink Bioarrays 30-mer format Affymetrix GeneChips 25-mer format Illumina Beadchips NimbleGen 60-mer format 9

Illumina Bead Array Technology Silica Beads Each bead is covered with hundreds of thousands of copies of a specific oligonucleotide 10

Some Facts Each bead carries copies of probes with, on average, 30 replicates of every bead type per array Around 105 copies of a particular DNA sequence of interest are covalently attached to each bead DNA sequences (oligonucleoties) attached to the beads are 75 base pairs in length, with 25 base pairs used for decoding and 50 base pairs used for target hybridization A pool of different bead types is created, beads of the same type having the same probe sequence attached

Box Plots of unnormalized data 12

Raw vs Normalized data Raw Data Normalized Data 13

Histograms of unnormalized data 14

Why Normalize It adjusts the individual hybridization intensities to balance them appropriately so that meaningful biological comparisons can be made. Unequal quantities of starting RNA Differences in labeling or detection efficiencies between the fluorescent dyes used Systematic biases in the measured expression levels. Sample preparation Variability in hybridization Spatial effects Scanner settings Experimenter bias 15

Free Software – Data analysis Bioconductor is an open source and open development software project to provide tools for the analysis and comprehension of genomic data. TMEV 4.0 is an application that allows the viewing of processed microarray slide representations and the identification of genes and expression patterns of interest. 16

R / Bioconductor R and Bioconductor packages R (http://cran.r-project.org/ )is a comprehensive statistical environment and programming language for professional data analysis and graphical display. Bioconductor (http://www.bioconductor.org/) is an open source and open development software project for the analysis of microarray, sequence and genome data. More 300 Bioconductor packages. http://faculty.ucr.edu/~tgirke/Documents/R_BioCond/R_BioCondManual.html 17

R / Bioconductor - Installation 18

Preparing R for analysis

Preparing R for analysis

Preparing R for analysis

Preparing R for analysis

Preparing R for analysis

Analysis using lumi R package - Loading data into R/Bioconductor >lumi_data <- lumiR(‘worshop_data.csv') Summary of the loaded data >lumi_data - Quality control of loaded data >summary(lumi_data, 'QC')

>density(lumi.Rdata)

>boxplot(lumi.Rdata)

>MAplot(lumi.Rdata)

>> plot(lumi.Rdata, what='sampleRelation') >> plot(lumi.Rdata, what=‘cv') >> plot(lumi.Rdata, what=‘outlier')

Variance Stabilization > lumi.Tdata <- lumiT(lumi.Rdata) > lumi.VSdata <- plotVST(lumi.Tdata)

> lumi.Ndata <- lumiN(lumi.Tdata) Normalization > lumi.Ndata <- lumiN(lumi.Tdata) Or Do all the default preprocessing in one step > lumi.N.Q <- lumiExpresso(lumi.Rdata) Background Correction: bgAdjust Variance Stabilizing Transform method: vst Normalization method: quantile Perform all the QC again > summary(lumi.Ndata, 'QC')

Differential expression >design <- model.matrix(~ -1 + factor(c(1, 1, 1,1, 2, 2, 2,2))) >colnames(design) = c("control","affected") >fit <- lmFit(lumi.Ndata, design) >cont.matrix <- makeContrasts(signature = affected - control,levels=design) >fit2 <- contrasts.fit(fit, cont.matrix) >ebFit <- eBayes(fit2) >results <- topTable(ebFit, number=100, sort.by="B", resort.by="M") >print(results) >write.table(topTable(ebFit, coef=1, adjust="fdr", sort.by="B", number=25000), file="results.xls", row.names=F, sep="\t")

Thank you http://catalyst.harvard.edu Reddy Gali, Ph.D. rgali@hms.harvard.edu Phone: 617 432 7471 http://catalyst.harvard.edu