Wald-type test statistics and application to differentially expressed gene detection in RNA-seq experiments.

摘要

Next Generation Sequencing (NGS) technologies are revolutionizing genome research and their application to transcriptomics (RNA-seq) is increasingly being used to detect differential expression (DE) in genes. RNA-seq has been developed in the last decade for simultaneously interrogating RNA transcripts, and thus compare the expression profiles of hundreds of thousands of genes in living cells. The study of differentially expressed genes under varying conditions is very important in both basic and medical research. Our goal in this thesis was to determine the most appropriate test statistics for detecting differentially expressed genes. We developed several Wald-type test methods, derived from an assumption involving the Poisson distribution. Our goal was to successfully detect the largest number of differentially expressed genes. Finally, we compared the Wald-type statistics and the T.Test statistic using both real-life and simulated data. We observe that in most cases the Wald-type statistics perform well, comparing to T.Test, in terms of detecting more differentially expressed genes and controlling type I error.