In each brain region, mRNA probes within 500 kb of the chromosomal location of each SNP were incorporated into linear regression modeling using MACH2QTLv1.08. Estimates of the association between the allelic dose of each SNP as a predictor of proximal gene expression levels were generated. These linear regression models were adjusted for biological covariates of age at death and gender, the first 2 component vectors from multidimensional scaling, as well as methodological covariates including post-mortem interval (PMI), tissue bank and hybridization batch. SNPs with fewer than 3 minor homozygotes detected (based on either genotyped SNPs or maximum likelihood genotypes from imputation) were excluded from analyses. A consensus set of results was extracted from the frontal cortex, cerebellum and blood eQTL datasets with identical overlapping combinations of GWAS SNPs and proximal cis mRNA probes.
