The next step was to thin the SNPs down to a set that would be present on a typical genotyping chip; we used the Affymetrix Genome-wide Human SNP Array 6.0 (Affymetrix, Santa Clara, CA) for this purpose. Examining only SNPs on this chip, for each we applied the trend test and calculated a P-value. We then took SNP with the smallest P-value, which we refer to as the hit SNP, and checked whether it showed a P-value less than 1×10−6. If this occurred we then modeled a replication study at this SNP using an additional 2,000 cases and 2,000 controls, and required a P-value less than 0.01. In what follows we only considered those simulations where the hit SNP on the genotyping chip met both criteria, as these model the ascertainment implicit in reported GWAS associations.
