We performed Mendelian randomization on the full set of 296 trans-eQTLs matched with a unique corresponding cis-eGene, measuring the causal impact of the cis-eGene on the trans-eGene, using the eVariant as the randomized instrumental variable (Supplementary Information 15). For trans-eQTLs with a cis-eGene, we observed strong evidence for regulation of the trans-eGene expression via the cis-eGene (Fig. 4a; P values ranging from P ≤ 3.0 × 10−5 to P ≤ 2.2 × 10−16). trans- eVariants with no cis-eGene may alter protein function, may reflect false negatives in the cis association test, or may arise from unmeasured regulatory mechanisms. Protein-coding loci were not enriched among our trans-eVariants (odds ratio 0.94; Fisher’s exact test, P ≤ 0.80), suggesting that modification of protein function is not the dominant mechanism for trans-eQTL effects.
