present in populations more closely related to African Americans. To efficiently remove the occurrence of low frequency SNPs that are most likely monomorphic in African Americans, investigators should consider filtering SNPs based on their MAF in subpopulations of interest. We found that postimputation filtering based on MAF in 1000 Genomes YRI+CEU+ASW reference subjects alleviated the problem with reduced overall imputation quality when including more distantly related reference subjects. We also found that the alternative approach of applying an r2 threshold (e.g., the widely applied r2≥0.3 threshold) was less favorable because it reduced but did not eliminate the occurrence of monomorphic SNPs in African Americans. Further, setting a stringent r2 threshold has been shown empirically to reduce statistical power, especially in regions of low linkage disequilibrium [35]. Our findings likely extend to other admixed populations. Future work is needed to evaluate the impact of using the 1000 Genomes ALL reference panel on low frequency SNPs for European-derived and other populations.
