Using the Coloc method54,55, we extracted eQTL signals of genes within 200 kb distance to significant GWAS variants (P<5 x 10−8) using effect sizes (β-values) and standard errors from eQTL and GWAS as input. Four partially hierarchical hypotheses55 were tested: H0, no association; H1, GWAS association only; H2, eQTL association only; H3, both GWAS and eQTL association but no co-localization; H4, both GWAS and eQTL association and co-localization, considering posterior probability for H4 (PPH4) >0.8 as strong evidence from both GWAS, eQTL and co-localization. In addition, we conducted the eQTL and GWAS CAusal Variants Identification in Associated Regions (eCAVIAR) approach56, in which, both eQTL and GWAS were fine-mapped, and the product of posterior probability (CLPP) was calculated, prioritizing genes with at least a single variant with CLPP>=0.01.
