To provide computational benchmarks for the imputation methods used in this study, we simulated two reference panels: one containing 4800 haplotypes modeled on the ancestrally diverse reference dataset that is being produced by the 1000 Genomes Project, and another containing 1000 haplotypes modeled on the European component of the 1000 Genomes set. We restricted the panels to a shared set of 10,000 SNPs spanning 1.9 Mb of sequence, and we imputed from both panels into a simulated GWAS of 1000 European individuals. The GWAS samples were provided with genotypes at SNPs that mirror the properties of the Affymetrix 500k platform; there were 337 such SNPs in the simulated region.
