Conversely, imputation strategies that are agnostic to population labels (as ours is) may become increasingly attractive as sequencing studies fill in the continuum of human genetic diversity. One benefit of such approaches is that they can capture unexpected allele sharing without needing to model the complexities of human demographic history, as we demonstrated by showing that African haplotypes can improve imputation accuracy in Europeans. Approaches like ours are also well-suited for imputation in recently admixed populations: methods that choose custom reference panels for different admixed individuals in different parts of the genome can increase accuracy by adapting to local ancestry changes, as previously suggested by Pasaniuc et al. (2010).
