Moreover, it should be noted that we applied an additive genetic model, which implies a gene-dose effect, but also reveals recessive or dominant associations at lower significance. Importantly, these three types of associations represent the biologically most meaningful possibilities. In contrast, the eQTL association analysis by Schadt et al.16 was based on Kruskal–Wallis tests, that is, a codominant model, which also reveals extreme deviations from additivity, that is, overdominant associations in which the two homozygous groups show similar expressions but the heterozygotes differ significantly. This (unknown) fraction of the eQTL associations reported by Schadt et al.16 must therefore be expected to be not reproducible by our study.
