As described above, at the 0.01 permutation threshold we detected 3,130 genes that had a significant cis- association, 1,074 (34%) of which had a significant cis-eQTL in at least two populations. We evaluated whether the extent to which pairs of populations shared significant cis- associations was related to their distance as defined by SNPs, with the goal of assessing the degree of sharing of functional variation across populations of varying ancestry. Qualitatively, we observed that more closely-related populations tend to share more cis- associated genes than more distantly-related populations (Figure S7). We further considered sharing using parsimony. Assessing eQTLs discovered at the 0.01 permutation threshold and shared at least at the 0.1 permutation threshold, we used an estimate of the general population structure of these eight populations to identify significant subpopulation sharing between CHB+JPT, CEU+GIH and CEU+GIH+JPT+CHB (Figure S8). For eQTLs estimated as ancestral (or recurrent) to all populations using this methodology we did not find any enrichment in particular functional categories (no GO terms had a 0.05 significance after Bonferroni correction).
