More broadly, we believe that any imputation algorithm should strive to incorporate as much of the available reference information as possible while remaining easy to use. For example, in Scenario B it is desirable to simultaneously impute the SNPs in the expanded panel (to improve accuracy) and the SNPs represented only in the HapMap (to maintain genomic coverage). IMPUTE v2 provides an integrated framework for handling this kind of problem: it is flexible enough to handle numerous variations of Scenarios A and B, yet it remains tractable by focusing computational effort on the parts of the dataset that are most informative.
