We observed that the degree of tissue sharing of an eQTL is associated with several indicators of phenotypic impact. Tissue-shared eGenes are depleted from loss-of-function mutation-intolerant genes (as curated by ExAC41) (Fig. 5a), consistent with purifying selection removing large-effect regulatory variants that involve many tissues. Tissue-shared eGenes were also less likely than tissue-specific eGenes to be annotated disease genes (Fisher’s exact test, nominal P ≤10−6 for GWAS, Online Mendelian Inheritance in Man (OMIM), and loss-of-function-intolerant gene sets; Fig. 5a, Extended Data Fig. 14), highlighting the importance of broad tissue sampling for GWAS interpretation.
