cis-eQTL replication rates in all data sets were considerably greater than the applied type 1 error rates (Supplementary Table 3), despite the reduced power to replicate implied either by the reduced sample size (Heinzen et al.14) or the different expression array (NABEC and Zeller et al.25) or different tissue (Zeller et al.25). Region-specific replication patterns were also compelling (see below). We found these replication rates to compare favorably to other published replication rates for eQTL data sets26,27. For example, the published replication rate for two cis-eQTL data sets using the same array platform on a shared single cell type (monocytes) was 63.8% (refs. 25,26). In our data set, the replication rate using the same array platform in the same brain region was 48.8%, despite the challenges arising from the use of post-mortem tissue containing a mixture of cell types. We note that this does not mean that the remaining 51.2% are all false positives. The false positive rate in the set of non-overlapping cis-eQTL signals depends on various unknown factors, which include the power to detect cis-eQTL signals of different
