A recent simulation study focused on imputation accuracy for BEAGLE, fastPHASE, IMPUTE, MACH, and PLINK using 250 kb exemplary regions and the HapMap phase II CEU reference panel [Pei et al., 2008]. Another recent study examined both imputation accuracy and efficacy for BEAGLE, IMPUTE, MACH, and PLINK for genome-wide imputation on a sample of German individuals, also using the HapMap phase II CEU reference panel [Nothnagel et al., 2009]. Two groups have investigated imputation for an admixed study sample; one study was based solely on IMPUTE [Zhao et al., 2008] and the other study was based solely on fastPHASE [Pemberton et al., 2008]. Here, we performed a two-way comparison of the major classes of imputation algorithms and the selection of reference panel(s) specifically for an admixed (African American) study sample.
