Our method is well-suited to this kind of dataset: in addition to its ability to accurately impute rare SNPs, which will constitute most of the new variants in the 1,000 Genomes data, IMPUTE v2 expends relatively little computational effort on haploid imputation steps. This means that, for a given SNP chip typed in a given study sample, doubling the number of untyped variants in a phased reference panel will increase the computational burden of imputation by a factor of less than two. By contrast, other imputation methods (such as IMPUTE v1, BEAGLE, and fastPHASE) would slow down by a factor of at least two.
