Differential gene expression DGE studies aim at identifying genes whose mean expression level differs significantly between two known populations. The state of the art approach to this problem consists in performing one test per gene, followed by a multiple testing correction in order to control the False Discovery Rate (FDR), that is, the expected proportion of errors among selected genes. The obtained gene list is then typically refined by further selecting genes with a large effect size (as in volcano plots). However, such downstream selections generally have no associated statistical guarantees. This problem can be overcome by post hoc inference, which provides guarantees on the number or on the proportion of false discoveries among arbitrary gene selections.

In this talk we propose to give a survey of the use of permutation-based post hoc inference for DGE studies, based on a study dedicated to two-group DGE and a recent extension to more general study designs. These developments are implemented in dedicated R and python packages, and (most of them are) available from IIDEA, a dedicated interactive R/shiny application deployed at https://shiny-iidea-sanssouci.apps.math.cnrs.fr/.