Results of “converge first then cluster” (approach 1) and “cluster first then converge (approach 2) analytic strategies applied to substance dependence vulnerability datasets described herein.

Columns list the numbers of SNPs that display abuser vs control differences with nominal p<0.05 (nominally positive) and lie in clusters, the maximal distance between nominally positive SNPs that is considered to indicate clustering, the numbers of clustered, nominally positive SNPs in African American samples, the numbers of clustered, nominally positive SNPs in European American samples, the numbers of “convergent” SNPs that display nominally positive results in both samples, the fraction of “convergent” SNPs that are likely to be true positives, on average, based on comparison with randomly chosen SNPs that are selected for similar convergence analyses (data not shown), the numbers of genes identified by the clusters of nominally positive SNPs and the overlap between the genes in each set and the 104 genes identified by the preplanned criteria used for primary analysis, using approach 1 with 3 SNP and 25kb intervals (boldfaced). The primary comparison set from approach (2) is also listed in boldface, based on the similar fraction of true positives anticipated using these criteria. We summarize these data in Fig. 2.