To investigate reproducibility of our eQTL results, we performed a detailed comparison with the Seattle study,16 which differed in several important aspects including technology platform, origin and number of samples analyzed and available medical information. Surprisingly, only 200 of all 1262 associations (16%) were also found in the Seattle study when Bonferroni's adjustment was applied to both studies (Figure 2). A more detailed analysis showed that 30 of the 200 replicated associations were exact validations, that is, the same SNPs were found in both studies to be associated with the same trait, whereas 170 SNPs were validated by LD, that is, they were found to be tight linkage (r2>0.8) to SNPs of the Seattle study. The remainder of the 1062 SNPs represented independent associations only found in our study (Supplementary Table S2). Remarkably, none of the trans-associations in either study was replicated by the other (Figure 2). Reanalysis of the data after omitting most of the covariates except those considered in the former study revealed 1313 associations, whereas still only ∼20% of all associations were found in both studies (data not shown).
