Gene filtering by PCA for Affymetrix GeneChips

Due to the nature of the array experiments which examine the expression of tens of thousands of genes (or probesets) simultaneously, the number of null hypotheses to be tested is large. Hence multiple testing correction is often necessary to control the number of false positives. However, multiple testing correction can lead to low statistical power in detecting genes that are truly differentially expressed. Filtering out non-informative genes allows for a reduction of the number of hypotheses, which potentially can reduce the impact of multiple testing corrections. While several filtering methods have been suggested [1], the best practice to filtering is still under debate. We propose a new filtering statistic for Affymetrix GeneChips, based on principal component analysis (PCA) on the probe-level gene expression data. Given that all the probes in a probeset are designed to target one or a common cluster of transcripts, the measurements of probes in a probeset