Previous studies have suggested that the majority of human eQTLs map locally and that distant eQTL are less likely to replicate (Dixon et al. 2007; Schadt et al. 2008). We therefore used the proportion of distant effects among highly significant eQTLs as an initial conservative estimate of the false-discovery rate (conservativeFDR). Among the top 1200 eQTLs, both our new analysis and the original analysis detected a similar number of distant eQTLs (Fig. 1). However, as the number of eQTLs increased, the fraction of distant eQTL and the estimated conservativeFDR increased rapidly for our original analysis but not for the PC-adjusted analysis. Among the top 3000 eQTLs, the conservativeFDR estimate in the original analysis was ∼30% compared with 1% for the PC-adjusted analysis.
