Given a set of genes selected by some experimental measurement (e.g. because they are differentially expressed between two types of experiments), Marmite checks for significant enrichments in bioentity annotations in this set with respect to the background. The functional enrichment test carried out by Marmite is in many ways conceptually similar to the tests used for classical repositories such as GO, KEGG, etc. (Al-Shahrour et al., 2004; Khatri and Draghici, 2005). The difference in this case is that the functional category to be tested, the bioentity, is considered to be a continuous class. Membership of a gene to a bioentity is therefore defined by a score value, which would reflect the strength of the real relationship gene-bioentity. Therefore, instead of the usual Fisher's or hypergeometric (or similar) tests, we use a Kolmogorov–Smirnov test to compare the distributions of the scores of the co-occurrences between genes and bioentities for each bioentity studied to the background distribution of scores. Since all the bioentities are tested, the P-values assigned to them are adjusted by False Discovery Rate (Benjamini and Hochberg, 1995).