The imputation accuracy of common variants does not appear to be substantially affected by GWAS QC steps. Our data demonstrate that there is little difference in imputation accuracy observed in unQCed GWAS data when compared with QCed GWAS data. Furthermore, the implementation of additional QC steps (eg, filtering out variants with MAF<0.05 and <0.10) does not considerably improve overall imputation accuracy. Missing variants and directly typed variants that fail pre-imputation QC checks are imputed and these data are used for downstream analyses. Post-imputation QC successfully eliminates a good proportion of inaccurately imputed SNPs. Specifically, by applying a very stringent post-imputation QC threshold, a smaller set of variants with more accurately predicted genotypes remain. The IMPUTE-info threshold of <0.8 and MAF ≤5% criterion successfully filtered out the majority of poorly imputed SNPs. However, the application of these strict filters in GWAS data could result in many SNPs being excluded from the data, and thus potential true association signals could be missed. Some of the inaccurately imputed variants were due to poor clustering properties. It is plausible that the handful of variants
