Finally, we evaluated whether associated genes were overrepresented in broad gene ontology categories for creQTL. We organized all Gene Ontology [36] annotations into Biological Process (BP), Cellular Component (CC), and Molecular Function (MF) subcategories (Affymetrix, Santa Clara CA). SNPs were assigned to genes, as for cis-eQTL, using the minimum q-value for association across all tested SNPs within a 5 MB distance cutoff (to generate per-gene q-values). For all GO subcategories with at least 10 genes, the two-sample Kolmogorov-Smirnov test (one-tailed) was used to test for enrichment of significant genes in each Gene Ontology category. Specifically, to compute exact p-values for each category, we compared the observed q-value for a given gene set to those calculated from 10,000 same-sized sets of randomly drawn q-values. Because of the large number of categories, many of which contained overlapping genes, resulting p-values (for each GO category) were then adjusted for FDR [31].
