transcript being regulated was inside or outside the hotspot-bounded associated region of the marker. To account for uncertainty in the location of the true causal variant due to linkage disequilibrium, we redefined cis-eQTL signals according to the location of all markers in high linkage disequilibrium (r2 > 0.8) with the sentinel marker (Fig. 6c and Supplementary Table 3). Even when we restricted our attention to signals that replicated in one of our three replication data sets, we found that 44.1% of cis-eQTL signals lay entirely outside their target gene, and 88.0% lay entirely outside the associated region of the target gene. These findings point to one possible explanation for why functional characterization of some GWAS hits has been difficult.
