genes with a significant cis-association had an association in at least two of the populations, and 2% of genes in all eight populations. If we conservatively exclude those probes known to overlap common SNPs, we reduce the number of non-redundant genes exhibiting a cis-association to 2,900, but observe that 957 (33%) of the remaining genes had a significant cis-association in at least two of the populations, and 54 (2%) in all eight populations. Thus, probes with underlying SNPs are not contributing significantly to our estimates of across-population sharing (i.e., replication) of eQTLs. At higher stringency (permutation threshold 0.001), we note that none of the significant associations involve probes with underlying SNPs, and 48% of genes with a significant cis-association had an association in at least two of the populations, and 2% of genes in all eight populations (Table 1).
