In our gene-wise assessment in the AA sample, we observed a different set of results. We cannot be sure the degree to which these differences result from chance effects, or reflect differences in the underlying genetic architecture of risk for AD in these two ancestries (as has recently been suggested for schizophrenia (Purcell et al. 2009)). The most significant gene in the AA sample in our study was POMC, which had been previously tested in a family and population-based case-control study but had not produced significant association (Zhang et al. 2009); our study yielded a permuted P=0.0008. Of interest are results from two genes, DKK2 and ZNF699, which had previously shown significant association with AD symptom counts (Kalsi et al. 2010) and AD (Riley et al. 2006) in a severely affected clinical sample of Irish ancestry.
