“…For each subset, we created SNP weights that combine expression data from males and females, based on the assumption that cis-heritable expression is mostly shared across sexes 81 , and on the fact that the combined sample provides additional power to detect genetic features with cis-heritable expression. We used limma 3.42.0 82 to adjust the expression data for the institution of sample origin, case-control status, RNA integrity number, sex, post-mortem interval, age (determined in bins: #1 = 17−29 years, #2 = 30−49 years, #3 = 50−69 years, #4 = 70−89 years, #5 = 90+ years), the first ten population covariates estimated through a principal component analysis performed in PLINK 1.9 58 , and surrogate variables calculated using sva 3.34.0 83 , following previous work 53,84 . The number of surrogate variables was determined as a function of sample size (N), as suggested by GTEx (i.e., 30 for sample sizes between 150 and 250, and 60 for sample sizes above 350).…”