for only a single study: for example, although the OZALC-NAG study show a significant excess of pathways (p-value = 0.0346) with p-values lower than 0.005 when we analyzed the SNPs with p-values lower than 0.005, the SAGE study was not significant (p-value = 0.141) for these same cutoff values (Table 2). Because of this lack of consistency, we did not consider the categories of genes identified by these specific values for the thresholds. In contrast, there was a replicated significant excess of categories of genes for both studies when the SNPs with p-values <0.001 were analyzed. This occurred independently of the threshold for categories considered (i.e., 0.005, 0.01 and 0.05) (Table 2); thus we considered all of the terms and pathways with a category-specific p-value that satisfied the most relaxed constraint for category cutoff (i.e., = 0.05), which includes all of the pathways satisfying the more stringent constraints. When we analyzed the SNPs with p-values <0.05 we observed a significant excess of enriched categories of genes with p-values <0.005 and 0.01 consistently in both OZALC-NAG and SAGE studies (Table 2). Applying the same criteria described above, we analyzed the categories with category-specific p-values <0.01. Our replication strategy was to evaluate
