For cis eQTLs (primary and secondary), to adjust for multiple testing, we permuted genotype sample labels in each cis window 10,000 times, keeping everything else in the model constant. To derive an empirical P-value distribution, the test statistic of the most significant variant in each permutation round was stored. A region-wise adjusted cis P-value was derived from the proportion of permuted test statistics that were larger than the most significant observed test statistic in the region. These threshold P-values were further adjusted for genome-wide analysis using the Benjamini-Hochberg (BH) correction. A gene was considered an eGene if its final BH-adjusted P-value was less than 0.05.
