For each population, we employed a linear regression model as described above to test for association between SNP variation and expression variation. We confined the trans analysis to those SNP-gene combinations where the genomic distance between probe midpoint and SNP was greater than 1Mb (or where probe and SNP were on different chromosomes). Significance was evaluated through 10,000 permutations as described above. We identified 43, 37, 38, 23 genes in CEU, CHB, JPT and YRI, respectively, with significant trans- associations (0.001 permutation threshold). In total, 108 genes show significant trans- association in at least one population (16 genes or 15% show a significant trans- association in at least 2 populations and 5 in all four populations). We also performed analysis in pooled populations as described above and detected 44, 52 and 39 genes for the 4-, 3- and 2-population pools respectively. Overall there seems to be low power to detect trans effects in these cell lines and sample sizes (Table 4), as indicated by the low number of discovered genes and consequently the high FDR.
