SVAPLSseq: A Method to correct for hidden sources of variability in differential gene expression studies based on RNAseq data

RNAseq technology has revolutionized the face of gene expression profiling by generating read count data measuring the transcript abundances for each queried gene. But on the other side, the underlying technical artefacts generate a wide variety of hidden effects that may potentially distort the primary signals of differential expression between two sample groups. This is in addition to the factors of unwanted biological variability may give rise to a highly complicated pattern of residual expression heterogeneity in the data. Standard normalization techniques fail to correct for these latent variables and leads to a substantial reduction in the power of common statistical tests for differential expression. Here I introduce a novel method SVAPLSseq that aims to capture the traces of hidden variability in the data and incorporate them in a regression framework to re-estimate the primary signals for finding the truly positive genes. Application on both simulated and real-life RNAseq data shows the superior performance of the method compared to other available techniques. The method is provided as an R package SVAPLSseq that has been submitted to Bioconductor.