Getting the numbers comparable Normalization Getting the numbers comparable

The DNA Array Analysis Pipeline Question Experimental Design Array design Probe design Sample Preparation Hybridization Buy Chip/Array Image analysis Normalization Expression Index Calculation Comparable Gene Expression Data Statistical Analysis Fit to Model (time series) Advanced Data Analysis Clustering PCA Classification Promoter Analysis Meta analysis Survival analysis Regulatory Network

Intensities are not just mRNA concentrations Tissue contamination RNA degradation RNA purification Reverse transcription Amplification efficiency Dye effect (cy3/cy5) Spotting DNA-support binding Other issues related to array manufacturing ‘Background’ correction Image segmentation Hybridization efficiency and specificity Spatial effects

Gene-specific variation Two kinds of variation Global variation Amount of RNA in the biopsy Efficiencies of: RNA extraction Reverse transcription amplification Labeling Photodetection Gene-specific variation Spotting efficiency, Spot size Spot shape Cross-/unspecific hybridization Biological variation Effect Noise Systematic Stochastic

Stochastic noise we use statistics to deal with PCA Plot of 34 patients, 8973 dimensions (genes) reduced to 2

...like we will see later PCA for 100 most significant genes reduced to 2 dimensions

Gene-specific variation: Sources of variation Global variation: Similar effect on many measurements Corrections can be estimated from data Gene-specific variation: Too random to be explicitly accounted for “noise” Systematic Stochastic Normalization Statistical testing

Calibration = Normalization = Scaling

Nonlinear normalization

The Qspline method From the empirical distribution, a number of quantiles are calculated for each of the channels to be normalized (one channel shown in red) and for the reference distribution (shown in black) A QQ-plot is made and a normalization curve is constructed by fitting a cubic spline function As reference one can use an artificial “median array” for a set of arrays or use a log-normal distribution, which is a good approximation.

Accumulating quantiles Once again…qspline Accumulating quantiles When many microarrays are to be normalized to each other an average array can be used as target

Lowess Normalization M A * * * * One of the most commonly utilized normalization techniques is the LOcally Weighted Scatterplot Smoothing (LOWESS) algorithm.

Invariant set normalization (Li and Wong) A invariant set of probes is used Probes that does does not change intensity rank between arrays A piecewise linear median line is calculated This curve is used for normalization

Spatial normalization After intensity normalization After spatial normalization Raw data After intensity After intensity Spatial bias estimate After spatial After spatial normalization normalization normalization normalization

The DNA Array Analysis Pipeline Question Experimental Design Array design Probe design Sample Preparation Hybridization Buy Chip/Array Image analysis Normalization Expression Index Calculation Comparable Gene Expression Data Statistical Analysis Fit to Model (time series) Advanced Data Analysis Clustering PCA Classification Promoter Analysis Meta analysis Survival analysis Regulatory Network

Expression index value Some microarrays have multiple probes addressing the expression of the same gene Affymetrix chips have 11-20 probe pairs pr. Gene - Perfect Match (PM) - MisMatch (MM) PM: CGATCAATTGCACTATGTCATTTCT MM: CGATCAATTGCAGTATGTCATTTCT However for downstream analysis we often want to deal with only one value pr. gene. Therefore we want to collapse the intensities from many probes into one value: a gene expression index value

Expression index calculation Simplest method? Median But more sophisticated methods exists: dChip, RMA and MAS 5 (from Affymetrix)

dChip (Li & Wong) Model: PMij = qifj + eij Outlier removal: Identify extreme residuals Remove Re-fit Iterate Distribution of errors eij assumed independent of signal strength (Li and Wong, 2001)

RMA Robust Multi-array Average (RMA) expression measure (Irizarry et al., Biostatistics, 2003) For each probe set, re-write PMij = qifj as: log(PMij)= log(qi ) + log(fj) Fit this additive model by iteratively re-weighted least-squares or median polish

MAS. 5 MicroArray Suite version 5 uses MM* is an adjusted MM that is never bigger than PM Tukey biweight is a robust average procedure with weights and outlier rejection

Std Dev of gene measures from 20 replicate arrays Methods compared on expression variance Std Dev of gene measures from 20 replicate arrays Std Dev of gene measures from 20 replicate arrays Expression level Blue and Red: RMA; Black: dChip; Green: MAS5.0 From Terry speed

Robustness MAS5.0 MAS 5.0 Log fold change estimate from 20ug cRNA (Irizarry et al., Biostatistics, 2003) MAS 5.0 Log fold change estimate from 1.25ug cRNA Log fold change estimate from 20ug cRNA

Robustness dChip dChip Log fold change estimate from 20ug cRNA (Irizarry et al., Biostatistics, 2003) dChip Log fold change estimate from 20ug cRNA Log fold change estimate from 1.25ug cRNA

Robustness RMA RMA Log fold change estimate from 20ug cRNA (Irizarry et al., Biostatistics, 2003) RMA Log fold change estimate from 20ug cRNA Log fold change estimate from 1.25ug cRNA

All of this is implemented in… R In the BioConductor packages ‘affy’ (Gautier et al., 2003).

References Li and Wong, (2001). Model-based analysis of oligonucleotide arrays: Model validation, design issues and standard error application. Genome Biology 2:1–11. Irizarry, Bolstad, Collin, Cope, Hobbs and Speed, (2003) Summaries of Affymetrix GeneChip probe level data. Nucleic Acids Research 31(4):e15.) Affymetrix. Affymetrix Microarray Suite User Guide. Affymetrix, Santa Clara, CA, version 5 edition, 2001. Gautier, Cope, Bolstad, and Irizarry, (2003). affy - an r package for the analysis of affymetrix genechip data at the probe level. Bioinformatics