Our large compendium of brain xQTLs can also be leveraged to accelerate gene discovery by boosting statistical power in GWAS. The simplest way of using our xQTL SNP list would be to restrict association analysis to our xQTL SNPs. However, such a strategy would miss other relevant SNPs that are not in our list (or were not tested in the cis xQTL analysis). Thus, we opted to use a weighted Bonferroni procedure44, which permits all SNPs to be analyzed but weights their p-values by their potential phenotypic relevance. We refer to this approach as an “xQTL-weighted GWAS”. Provided that the weights are non-negative and average to one, strong control on family-wise error rate is guaranteed44. We employed a binary weighting scheme, where p-values of xQTL SNPs were divided by w1 and all other SNPs were divided by w0 with s = w1/w0 > 1 (see Supplementary Information for s selection). Consistent with the standard GWAS convention, significance was declared at p < 5x10−8. To not over-count the number of significant hits due to correlations between SNPs, we applied PLINK1.945 on
