Genotype data16 were generated from 2,093 individuals of European-descent. Of these individuals, gene expression (RNA-seq)(n=494), DNA methylation17 (450K Illumina array)(n=468), and histone modification data (H3K9Ac ChIP-seq)(n=433) were derived from post-mortem frozen samples of a single cortical region, the dorsolateral prefrontal cortex (DLPFC) (Figure 1A). 411 individuals have all four data types. Demographics of the analyzed individuals are summarized in Table S1. Although some of these data have been previously published with respect to analysis of aging brain phenotypes (see Table S2), here we report genome-wide xQTL analyses for these datasets for the first time. Genotype imputation was performed using BEAGLE 3.3.218 with the 1000 genome reference panel19, yielding 7,321,515 SNPs for analysis. For the molecular phenotype data, 13,484 expressed genes, 420,103 methylation sites, and 26,384 acetylation peaks remained after quality control (QC) analyses (Figure S1–S3). The effects of known and hidden confounding factors were removed from the molecular phenotype data using linear regression (Supplementary Information). Consistent with previous studies, we observed that accounting for hidden confounding factors greatly enhances the statistical power of cis eQTL detection20, and we confirm that this observation holds true for cis mQTL and cis haQTL detection (Figure S4).
