Academic journal article Genetics

Fine-Scale Inference of Ancestry Segments without Prior Knowledge of Admixing Groups

Academic journal article Genetics

Fine-Scale Inference of Ancestry Segments without Prior Knowledge of Admixing Groups

Article excerpt

(ProQuest: ... denotes formulae omitted.)

ADMIXTURE occurs when reproductive isolation between groups allows genetic divergence via genetic drift and random mutation, followed by mixing of the diverged groups to form new populations. Such genetic admixture is near ubiquitous in observed human populations (Patterson etai. 2012; Loh et al. 2013; Hellenthal et al. 2014) and indeed other species including cattle (Upadhyay et al. 2017), bison (Musani et al 2006), and wolves (Pickrell and Pritchard 2012).

Genome-wide summaries can reveal not only complex relationships between modern populations but also details of their demographic histories (Pickrell and Pritchard 2012; Hellenthal et al. 2014; Peter 2016) while accurate inference of local ancestry can be used to correct for population structure in association testing (Diao and Chen 2012; Xu and Guan 2014), detect selection (Zhou et al 2016), and can be used for mapping disease loci (Zhang and Stram 2014).

Due to the process of recombination, contiguous chunks of admixed individuals' genomes are inherited intact from one mixing population or another. In the second generation following the initial admixture, chromosomes from distinct ancestral groups begin to recombine, and so the expected length of these chunks (in units of Morgans) will be 1 (by definition), and (neglecting crossover interference) chunk lengths can be modeled using an exponential distribution with rate parameter 1. In each subsequent generation, recombination further breaks down these chunks so that the chunk lengths (if they could be observed) are distributed according to an exponential distribution with rate parameter one less than the number of generations since admixture.

To fully characterize admixture for the above purposes, we need to infer: (1) Whether a group of individuals are admixed; (2) the component/mixing groups; (3) the timing of the admixture event(s); and (4) which segments of the admixed genome are inherited from each mixing group. Typically, we lack prior knowledge of each of these points and we do not have access to representative samples of the mixing groups, as these are often no longer present (without drift) in modern samples.

A wide variety of approaches to model admixture have been developed in recent years. STRUCTURE (Pritchard et al. 2000) clusters similar genomes together by fitting a mixture model using Gibbs sampling and STRUCTURE 2.0. Falush et al. (2003) extended this model to allow for admixed individuals using a Hidden Markov Model (HMM) that allowed for linkage along the genome. A drawback of these and similar approaches (e.g., Sohn et al. 2012) is that they do not attempt to fully model linkage disequilibrium (LD), because SNPs within each source population are assumed to be independent, meaning they are not maximally powerful for inferring ancestry segments, particularly for subtle admixture events.

Other approaches focus on dating/characterizing admixture events, without performing local ancestry estimation. In the ALDER model (Loh et al. 2013) the exponential decay of ancestry segments is estimated as a function of genetic distance, allowing dating of admixture events. GLOBETROTTER (Hellenthal et al. 2014) uses a related approach for dating events by leveraging haplotype data, accounting for LD, but also infers admixture proportions and properties of the ancestral mixing groups, by quantifying their relationships with modern observed populations, and can handle multi-way admixture. In common with other approaches (some discussed below), GLOBETROTTER incorporates LD between nearby SNPs by fitting a haplotype copying model (Lawson et al. 2012)closely related to the Hidden Markov model introduced by Li and Stephens (2003). Here, "target" chromosomes of interest are formed as a mosaic whereby they imperfectly "copy" segments of DNA from donor haplotypes, according to a HMM. See Gravel (2012) for a review of this and other local ancestry models. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.