To identify eQTLs, gene expression data from European-ancestry subjects (N = 467) were adjusted for known and hidden variables detected by surrogate variable analysis (SVA) conditional on diagnosis but excluding ancestry (Supplementary Fig. 2 and 4). Adjusted expression levels were then fit to imputed SNP genotypes, covarying for ancestry and diagnosis, using an additive linear model implemented in MatrixEQTL. The model identified 2,154,331 significant cis-eQTLs, (i.e., within 1 Mb of a gene) at a false discovery rate (FDR) ≤ 5%, for 13,137 (80%) of 16,423 genes. Many eQTLs for the same gene were highly correlated, due to linkage disequilibrium, and 32.8% of eQTL SNPs (“eSNPs”) predict expression of more than one gene. Cis-eSNPs were enriched within genic elements and non-coding RNAs, particularly within 100 kb of the transcription start and end sites12, and depleted in intergenic regions (Fig. 1A, B). As defined by GTEx13, an “eGene” is a gene with at least one significant eSNP after strict correction for multiple marker testing for that gene. There were 8,427 eGenes at FDR ≤ 5%, or 18 eGenes discovered per sample, consistent
