Imputation-Based Local Ancestry Inference in Admixed Populations

Accurate inference of local ancestry from whole-genome genetic variation data is critical for understanding the history of admixed human populations and detecting SNPs associated with disease via admixture mapping. Although several existing methods achieve high accuracy when inferring local ancestry for individuals resulting from the admixture of genetically distant ancestral populations (e.g., AfricanAmericans), ancestry inference in the case when ancestral populations are closely related remains challenging. Surprisingly, methods based on the analysis of allele frequencies at unlinked SNP loci currently outperform methods based on haplotype analysis, despite the latter methods seemingly receiving more detailed information about the genetic makeup of ancestral populations. In this paper we propose a novel method for imputation-based local ancestry inference that exploits ancestral haplotype information more effectively than previous haplotype-based methods. Our method uses the ancestral haplotypes to impute genotypes at all typed SNP loci (temporarily marking each SNP genotype as missing) under each possible local ancestry. We then assign to each locus the local ancestry that yields the highest imputation accuracy, as estimated within a neighborhood of the locus. Experiments on simulated data show that imputation-based ancestry assignment is competitive with best existing methods in the case of distant ancestral populations, and yields a significant improvement for closely related ancestral populations. Further demonstrating the synergy between imputation and ancestry inference, we also give results showing that the accuracy of untyped SNP genotype imputation in admixed individuals improves significantly when using estimates of local ancestry. The open source C++ code of our method, released under the GNU General Public Licence, is available for download at http://dna.engr.uconn.edu/software/GEDI-ADMX/.