To resolve multiple associations due to haplotype effects as one signal, we calculated the linkage disequilibrium between markers and clustered them such that the markers in a cluster had a pairwise linkage disequilibrium of r2 > 0.5 with each other. Then for a given transcript ID-cluster pairing, we chose the marker with the lowest P value for association across all relevant tests as the sentinel marker. On average, there were 13.7 markers per sentinel marker among the significant results, and we refer to the resulting significant sentinelized marker-expression ID-tissue combinations as ‘subsignals’. Finally, we refer to the grouping of subsignals into significant sentinelized marker-transcript ID combinations as eQTL ‘signals’.
