There are two potential limitations to the current study that should be considered when interpreting its results. First, we obtained genome-wide genotype data for African American study subjects from Illumina’s iControlDB, and there could be unrecognized problems in the data collection. Use of publicly available genotype data for GWAS is becoming more common as a resourceful approach to testing analytic methods (as done in our study), among other uses. Genetic studies using controls from public sources require stringent QC procedures, as demonstrated by our population structure analyses showing that several of the publicly available African American study subjects were ancestral outliers, as previously reported [36]. Second, the imputation performance patterns were deduced from chromosome 22. We compared imputation performance metrics between the largest and smallest autosomes (chromosomes 1 and 22) using the YRI reference population, and we found that imputation quality and accuracy were somewhat better on the larger chromosome 1 but not different in character from the chromosome 22 results. Other chromosomes likely have higher performance metric values than those presented here for chromosome 22, but we do not expect the genome-wide imputation patterns to differ greatly.
