mRNA transcripts that are well detected within one or more brain regions but not all four tissue regions. The number of SNPs, CpG sites and mRNA transcripts tested for each brain tissue can be found in Table 1. With these data for each tissue we then performed linear regression of allele dosage against each measure using CpG methylation or expression of mRNA as the dependent variable and genotype as the independent variable. This regression analysis was performed with Plink [17] to correlate allele dosage with the quantitative trait. We corrected for number of tests per trait by permutation, computing a genome-wide empirical p-value for each of the ∼1.6 million SNPs tested against each individual trait. To correct for the number of traits tested within each brain region by assay type an FDR threshold for significance was determined based on the empirical p-values from the proceeding step. This yielded a necessarily conservative threshold for significance [18]. Prior to analysis each trait was adjusted using available biological and methodological covariates in an attempt to reduce the influence of systematic confounding effects (Text S1). Post hoc we annotated significant QTLs as cis if the SNP lay within 1MB of either the CpG methylation
