the 2 Mb window (Figure 4 and Figure S11). For genes with significant cis-eQTL associations in all eight populations, the distribution is centered on the TSS, with few associations extending beyond +/−200 Kb from the TSS. As population sharing increases from genes with significant associations in only one population to genes with significant associations in all eight populations, we see a gradual tightening of the distribution around the TSS. This is true for single populations (Figure S11) and when examined in aggregate across populations (Figure 4). This is unlikely to be driven simply by false positive associations that fail to replicate, as the tightening pattern is observed even when comparing associations shared in six versus seven versus eight populations, which are themselves unlikely to be false positives. This observation suggests that the genomic distribution of cis-regulatory variants with respect to their contribution to genome function is different even when deviations of allele frequency are taken into account.
