To determine whether independent cis- regulatory signals exist for a given gene, we applied a stepwise association model as follows: For each probe that had a significant cis-eQTL at the 0.01 significance threshold, we regressed out of the expression levels the effect of the most-significant SNP, re-ran the SRC analysis on the rest of the significant cis-eQTL SNPs using the resulting expression residuals, and stored those SNPs with p-values more significant than the gene's permutation threshold. This was repeated separately for each probe until there were no SNPs from the initial significant eQTL list left to test (i.e. until none pass the permutation threshold after removing the effect of the most significant SNP at that step). At each iteration step, the most-significant SNP passing the permutation threshold is stored as an independent eQTL. We compared results obtained using the permutation thresholds based on the PCA corrected expression data to those obtained using permutation thresholds based on the residuals determined at each step of the stepwise model. We observed that there was no difference in the number of detected effects (not shown), so we used the thresholds based on the PCA residuals to evaluate p-values across steps.
