Does it matter how we rank genes for GSEA?

What is it about?

Gene set enrichment analysis (GSEA) is an important tool in biological discovery. In short, we first group genes in classes (gene sets) such as "interferon inducible genes" and then ask whether genes from a class are more likely to be up-regulated in a particular condition. For example, we might wonder whether interferon inducible genes are more likely to be highly expressed in tuberculosis. There are many types of GSEA and many algorithms (a popular one of them, confusingly, is even called "GSEA"). Very successful algorithms belong to the so-called "second generation" of GSEA and rely on creating lists of genes ordered by something, e.g. by p-value from a comparison between two conditions. It is possible to create different types of ordered lists (for example, order genes by the magnitude of effect size, or by a novel metrics that we invented, which we call MSD).

Featured Image

Why is it important?

Few analysis methods are as commonly used in transcriptomics as gene set enrichment analysis. It is therefore important to know which approaches are best. In this paper, we look at one part of the process: the ranking of the genes, and we show that it really does matter which type of ranking one chooses.

Perspectives