Project Summary Recent studies have shown that population mixture (or `admixture') is pervasive throughout human evolution and has played a major role in shaping human genetic and phenotypic variation. Despite the ubiquity and importance of population mixture, we still lack adequate methods to characterize the impact of admixture on a genomic scale and leverage this information for effective gene mapping. Addressing these topics is the central focus of research in my lab. In this proposal, our goal is to develop new methods to reconstruct fine-scale genomic ancestry in admixed groups and leverage this information to identify novel disease and adaptive mutations and genes. The application of these methods to large genomic surveys will help to discover novel disease and adaptive variants. The first step in characterizing the genomic impact of admixture is to infer the ancestry of each chromosomal segment, referred to as local ancestry. Towards this goal, we are developing new methods for local ancestry inference using machine-learning approaches that are ideally suited for classification problems and computationally tractable for large datasets. Our preliminary results show that our method is highly accurate and applicable across a range of demographic models. With reliable local ancestry inference, we will be well placed to study the impact of admixture on disease architecture and evolution of complex traits. We propose to use Admixture Mapping, a method to identify disease associations by leveraging ancestry differences across the genome, between cases and controls or among cases alone. By applying Admixture Mapping to complex admixed groups like South Asians and Latinxs, we aim to discover new population- specific disease associations and advance our understanding of disease architecture. Further, we will develop a novel method to leverage the demographic history of admixed groups to identify adaptive variants. By applying the method to study selection at various timescales in human evolution, we will uncover candidate genes and pathways related to adaptive gene flow and characterize its role in shaping human genetic variation. Finally, we will build reference-free ancestral genomes by recovering chromosomal segments of our lost ancestors hidden in admixed genomes. We will use these genomes to reconstruct the demographic history of our ancestors, as well as understand the fitness effects of population mixtures and the phenotypic legacy of our extinct ancestors. The successful completion of the proposed project will provide new statistical tools to leverage patterns of admixture to perform effective disease mapping and evolutionary inference in diverse, admixed groups. Application of these methods to large-scale genomic datasets will provide insights into the genetic, evolutionary, and functional impact of admixture during human evolution. Algorithms proposed here will be implemented in freely available software for use by other researchers.