One caveat to these studies is that direct comparison of datasets derived from separate tissue types with differential ascertainment methods is difficult. Specifically, the brain samples were taken from deceased subjects whereas blood samples were drawn in life. However, post-mortem interval has been shown not to be a major confound within brain expression data (Gibbs et al., 2010, Trabzuni et al., 2011) and we corrected for this and other known methodological variables in the statistical model. However, because the samples used here were from different individuals, we cannot exclude that we are detecting rare alleles and/or genetic variants on a background of common SNPs. As demonstrated by power analysis, the current dataset is not powered to directly detect rare alleles but has good power to detect relatively large eQTL effect sizes. Therefore, this analysis performs best for loci that are tagged by common variants and where the effect of the minor allele on expression is relatively large. It is also important to note that in the present study, we limited our analysis to transcripts within a relatively narrow (0.5 MB)
