We simulated 1,000 sets of 1,416 SNPs (to match the NHGRI GWAS Catalog SNP list), selected to overlap each of the predefined seven functional categories. In each set, variants were randomly selected from a given functional category. In order to mimic a typical GWAS approach, we then identified a tagging SNP that was in LD and was available on a commercial genotyping array (Illumina Human Omni2.5) for each functional variant. In total, 5,569,657 of the available SNPs were tagged (r2 ≥ 0.5) by 1,218,618 common (MAF ≥ 5%) variants on the Illumina Human Omni 2.5 array. If multiple SNPs were in LD with the functional variant, we selected the best tag with the greatest r2. Finally, we required SNPs in the final set to be more than 100 kb apart from each other to ensure independence. As an alternative to simulating a sequencing-based study, we also simulated 200 sets of 1,416 causal SNPs for each of the predefined functional categories by selecting tagging SNPs with strong LD (r2 ≥ 0.8) from the EUR subset of the 1000 Genomes Project data. If multiple SNPs were in LD with the functional variant, we selected the best tag with the greatest r2, and we selected the functional SNP itself when no tag SNP could be selected according to these criteria.