Differential expression analysis

Differential expression analysis identifies which genes are transcribed at significantly different levels between two or more conditions, for example treated versus untreated cells, or diseased versus healthy tissue, using RNA sequencing or microarray data. The result is a ranked list of genes with a fold-change, how much expression differs, and a statistical significance value for each.

Because a single study tests expression across tens of thousands of genes simultaneously, differential expression analysis is a large-scale multiple hypothesis testing problem: a raw p-value threshold of 0.05 applied gene by gene would produce huge numbers of false positives, so results are corrected using a false discovery rate method, most commonly the Benjamini-Hochberg procedure, rather than a simple significance cutoff. This differs from genomics analysis of DNA sequence itself, which looks at what genetic variants are present rather than which genes are actively being expressed.

Differential expression analysis is central to identifying disease biomarkers, drug mechanisms and developmental pathways, and is normally run as a documented, versioned scientific workflow given how easily normalization and correction choices change the result. Common pitfalls are too few biological replicates to estimate variability reliably, batch effects confounded with the condition being studied, and reporting fold-change without the accompanying significance and false-discovery correction.