To determine whether the significant trans-eQTL SNPs were enriched for functional regions on the genome, we annotated the trans-eQTL SNPs using SNPInfo53, SNPNexus54,55, and HaploReg56, which integrate multiple data sources (such as ENCODE project data31, Ensembl57, and several micro-RNA databases). We limited these analyses to those trans-eQTL SNPs that were previously shown to be associated with complex traits at genome-wide significance levels (‘trait associated SNPs’, reported P < 5 × 10-8). These SNPs were subsequently pruned (using PLINK's --clump command, using an r2 < 0.2). We used the permuted trans-eQTL data to get realistic null-distributions for each of these tools: we selected equally sized sets of unlinked SNPs (r2 < 0.2 in the Western-European subpopulations of the 1000 genomes project25, 2011-05-21 release, 286 individuals, excluding Finnish individuals) that showed the highest significance in the permuted data, ensuring that only trait-associated SNPs are included in the null-distribution, as it is known that trait-associated SNPs in general already have different functional properties than randomly selected SNPs58 (e.g. trait-associated SNPs typically map in closer proximity to genes than random SNPs). We also ensured
