eQTL analyses were carried out using FastQTL [18], using quantile normalized gene expression measures corrected for age, sex, RIN, sequencing batch, the first three genotype principal components (calculated in plink1.9, using LD pruned variants (r2 > 0.2 in 250 kb windows with a step size of 5 kb)), and 10 hidden confounders estimated through PEER [17]. FDR for each eQTL was calculated by first correcting P values for the number of SNPs tested per gene/transcript (within 1 Mb either side of the TSS) through estimation of a beta distribution using a minimum of 1000 permutations (maximum 10,000), and then correcting these P values for the number of genes/transcripts tested using Storey’s q value method [30]. Details of all steps of these analyses are available at www.github.com/hobrien/GENEX-FB2.
