Posterior Inference of Differential Expression (DE) State and Evolutionary Distance

Abstract

Abstract To address how gene regulation plays a key role in phenotypic innovations through high throughput transcriptomes, it is desirable to develop statistically-sound methods that enable researchers to study the pattern of transcriptome evolution. On the other hand, statistical detection of differentially expressed (DE) genes between two samples is probably the most common practice in the transcriptome analysis. In this paper, the author developed a new evolutionary approach, based on the genome-wide p-value profile arising from statistical testing of DE genes between species. First, the widely-used BUM (beta-uniform mixture) model for the genome-wide p-value distribution was invoked to calculate the posterior probability of a gene being DE denoted by Q = P(DE|p). Then the author develop a simple method to estimate the transcriptome distance for any number (m) of orthologous genes based on their Q-values. Computer simulations showed the statistical performance of these new methods are generally satisfactory.