Differential expression analysis for multiple conditions

TL;DR

This paper extends the DESeq method to analyze differential gene expression across three or more conditions simultaneously, using a Monte Carlo approach for efficient p-value approximation, demonstrated on simulated and real bacterial data.

Contribution

The paper introduces a novel extension of DESeq for multiple conditions, enabling simultaneous analysis and efficient computation with Monte Carlo methods.

Findings

Effective analysis of multiple conditions demonstrated on simulated data.

Application to C. jejuni data shows practical utility.

Monte Carlo approach provides fast p-value estimation.

Abstract

As high-throughput sequencing has become common practice, the cost of sequencing large amounts of genetic data has been drastically reduced, leading to much larger data sets for analysis. One important task is to identify biological conditions that lead to unusually high or low expression of a particular gene. Packages such as DESeq implement a simple method for testing differential signal when exactly two biological conditions are possible. For more than two conditions, pairwise testing is typically used. Here the DESeq method is extended so that three or more biological conditions can be assessed simultaneously. Because the computation time grows exponentially in the number of conditions, a Monte Carlo approach provides a fast way to approximate the $p$-values for the new test. The approach is studied on both simulated data and a data set of {\em C. jejuni}, the bacteria responsible for most food poisoning in the United States.