with nearly 630 distinct complex traits. In total, 5195 “independent” SNPs were identified after LD pruning at r2 ≥ 0.8 and counting SNPs only once (14). Of these, 308 (~6%) are in strong LD(r2 ≥ 0.8), with a “best eQTL per gene” (at FDR < 0.05) from either the single-tissue or multitissue eQTL discovery analysis (table S12) in at least one of the nine tissues tested. For two-thirds of these cases (211 SNPs), no putative deleterious coding variants (nonsynonymous or splice variants) in the target gene product lie in LD (r2 ≥ 0.8) with the GWAS SNP; this finding suggests that regulatory effects may underlie the causal mechanism, although additional work is needed to prove causality. GWAS SNPs in LD with an eQTL show a factor of 2 higher representation in coding regions relative to all GWAS SNPs (11% versus 4.6%; table S13). Notably, about one-third of the eQTLs in LD with GWAS SNPs were detected only with methods that leverage the multitissue nature of GTEx data. Increasing both sample sizes and the range of tissues will likely increase the number of detected GWAS-eQTL loci.
