A flexible model for correlated count data, with application to analysis of gene expression differences in multi-condition experiments

Detecting differences in gene expression is an important part of RNA sequencing (RNA-seq) experiments, and many statistical methods have been developed for this aim. Most differential expression analyses focus on comparing expression between two groups (e.g., treatment vs. control). But there is increasing interest in multi-condition differential expression analyses in which expression is measured in many conditions, and the aim is to accurately detect and estimate expression differences in all conditions. We show that directly modeling the RNA-seq counts in all conditions simultaneously, while also inferring how expression differences are shared across conditions, leads to greatly improved estimates of expression differences, particularly when the power to detect expression differences is low in the individual conditions (e.g., due to small sample sizes). We illustrate the potential of this new multi-condition differential expression analysis in analyzing data from a single-cell experiment for studying the effects of cytokine stimulation on gene expression. We call our new method “Poisson multivariate adaptive shrinkage,” and it is implemented in the R package poisson.mash.alpha, available at https://github.com/stephenslab/poisson.mash.alpha. Overall design: Droplet-based scRNA-seq on PBMCs from unstimulated mice and mice injected with various cytokines and chemokines.