To prevent the number of significantly-enriched gene sets being inflated by clusters of closely-related GO categories/KEGG pathways containing the same set of significant genes, a pruning procedure was then applied to the gene sets. A relatedness measure R was defined for each pair of gene sets as the number of genes in both sets divided by the number of genes in either (or both) sets. Thus identical gene sets would have an R=1 and completely non-overlapping ones have R=0. A criterion K was defined, and the list of gene sets pruned by removing categories with R > K, producing less biased but more significantly enriched gene sets. The same pruning procedure was applied to the lists of enriched gene sets from the simulated studies generated in ALIGATOR, to enable a test of an excess number of enriched gene sets, allowing for pruning, to be performed. In this paper, we utilized two values of K: 0.2 and 0.5.
