Annotating a GWAS SNP with an eQTL can help to highlight candidate causal genes within a locus (i.e., the eQTL target gene). We found that proximity-based and eQTL-based gene assignments for GWAS SNPs were often discordant (47). A surprising proportion of trait-associated SNPs in LD (r2 ≥ 0.80) with a GTEx eQTL showed disagreement between the strongest eQTL-derived target gene and the genes that were physically proximal to the GWAS SNP (table S14). Of 190 GWAS loci (P < 5 × 10−8) where the lead SNP is an eQTL from the single-tissue analysis (FDR < 0.05) with only a single target gene, in 65 cases (~34%) this eQTL target gene differs from any of the genes that were closest to the SNPs in LD. These results were also observed when we restricted the target genes to protein-coding genes, when we pruned the GWAS SNPs for each trait examined (r2 ≥ 0.80), and when we used the eQTLs identified from the multitissue joint eQTL analysis.
