Ancestry-based Methods for Characterizing the Evolutionary History of Admixed Populations

Admixture occurs when previously isolated populations come together to form a new population with genetic ancestry from those sources. Admixture is ubiquitous across the tree of life, including humans, and is often associated with migration and exposure to new environments and selective pressures. Admixed populations provide a unique opportunity to study adaptation on short timescales by introducing beneficial alleles at high frequency. However, admixed populations are often excluded from genomic studies due to lack of applicable methodology. Instead of relying on classical methods confounded by the process of admixture itself, we can detect changes in patterns of genetics ancestry that are informative about selection in admixed populations and at the short timescales often relevant for post-admixture selection. However, we lack theoretical expectations and methods to detect and characterize ancestry-based genomic signals indicative of post-admixture selection and adaptation. Common ancestry outlier approaches discard information about the surrounding genomic context and are prone to false positives due to drift and demography. Here, I present three studies which leverage patterns of genetic ancestry to investigate the evolutionary history of admixed populations. First, I develop a suite of ancestry-based summary statistics and computational methods to detect post-admixture adaptation, and demonstrate their application in a case study of human adaptation to malaria. In particular, these summary statistics incorporate patterns of ancestry beyond the site under selection, such as the length of contiguous ancestry tracts surrounding the locus, and are informative about the strength and timing of selection in admixed populations. I observe one of the strongest signals of recent selection in humans at the malaria protective Duffy-null allele, and show that this mode of strong single-locus selection over 20 generations has impacted genome-wide patterns of ancestry. Next, I move beyond summary statistics to develop a deep learning strategy for localizing regions of the genome under selection. This method takes images of chromosomes painted by ancestry as input to avoid the loss of information and bias that can occur when relying on user-defined summary statistics. I demonstrate this approach on simulated admixture scenarios and find that the method successfully localizes variants under selection 95% percent of the time, outperforms the common ancestry outlier approach, and is robust to demographic misspecification. Lastly, I present the first Illyrian genome sequences available from the Iron Age in a study of the ancestry and genetic relationships of five neonates buried in Korčula, Croatia. I find genetic support for classifying these individuals as Illyrian, and show that patterns of ancestry and genetic variation are consistent with their geographic location between Italy and the mainland Balkans. In the combined work presented here, I advance our ability to study the evolutionary history of admixed populations, which has implications for our understanding of phenotypic variation, disease risk, and conservation genetics across many study systems. Further, these methods tailored to the mosaic ancestry of admixed populations is a step towards expanding the diversity of populations, especially humans, who benefit from discoveries and advancement in genomic research.