In order to confirm these results in a different dataset, we replicated the study in European American subjects genotyped on two different platforms, Affymetrix 5.0 array and Illumina 550 K array. All subjects were controls from the National Institute of Mental Health Center for Collaborative Genetic Studies on Mental Disorders. We randomly divided about 400 individuals into two subgroups labeled “cases” and “controls” in a similar manner as above. “Cases” were genotyped by the Affymetrix 5.0 array and “controls” were genotyped by the Illumina 550 K array. In the replication, we also expanded our investigation to include those SNPs that were not genotyped in either array, but were imputed from their respective arrays. In fact, we had genotype data from both platforms. No genome wide significant SNPs were found. Therefore, if there were any significant SNPs in this simulation, they should be false positives. The result was similar with inflation of Type I error that is effectively filtered by IQS, whereas filtering by MAF and either the imputed information score or the variance ratio continue to have many false positive values (Table S2).
