We performed set-based permutation using an a priori list of 48 genes selected from four previous GWAS based gene-finding efforts in AD (Table 2). Although none of these genes achieved experiment-wise correction for multiple testing, gene-wise evaluation produced a number of genes yielding significance under the null (Table 3). We note that not all of them produced results; Table 3 shows the genes which achieved significance, or not, under the null. We considered first the full sample (EA and AA) and then each sub-sample separately. In the full sample, 20 genes yielded results, of which 2 survived permutation under the null hypothesis to produce p<0.05 and a total of 7 at p<0.1. Results in the individual samples showed statistical trends (p<0.1) for 5/19 genes in the EA samples and 7/23 for the AA sample. Overall, the most significant result was observed for ADH1C (alcohol dehydrogenase 1C) which produced a permuted value of p=0.015 in the combined sample. Considering the two samples separately, the most significant result was ADH1C in the EA sample (p=0.003) and POMC (proopiomelanocortin) in the AA sample
