Detecting Polygenic Adaptation in Admixture Graphs

Abstract

An open question in human evolution is the importance of polygenic adaptation: adaptive changes in the mean of a multifactorial trait due to shifts in allele frequencies across many loci. In recent years, several methods have been developed to detect polygenic adaptation using loci identified in genome-wide association studies (GWAS). Though powerful, these methods suffer from limited interpretability: they can detect which sets of populations have evidence for polygenic adaptation, but are unable to reveal where in the history of multiple populations these processes occurred. To address this, we created a method to detect polygenic adaptation in an admixture graph, which is a representation of the historical divergences and admixture events relating different populations through time. We developed a Markov chain Monte Carlo (MCMC) algorithm to infer branch-specific parameters reflecting the strength of selection in each branch of a graph. Additionally, we developed a set of summary statistics that are fast to compute and can indicate which branches are most likely to have experienced polygenic adaptation. We show via simulations that this method -which we call PolyGraph -has good power to detect polygenic adaptation, and applied it to human population genomic data from around the world. We also provide evidence that variants associated with several traits, including height, educational attainment, and self-reported unibrow, have been influenced by polygenic adaptation in different populations during human evolution.

An evolutionary compass for detecting signals of polygenic selection and mutational bias

“…Inference of selection on complex traits is vulnerable to several possible confounders, including population stratification and pleiotropic selection on off-target phenotypes (Novembre and Barton 2018). Although theory suggests that stratification should be straightforward to detect and correct at high frequency variants in large samples (Patterson et al 2006), an uncorrected bias in the inferred β values due to population structure can make our test (as well most others, such as Berg and Coop 2014;Yang et al 2015;Field et al 2016;Berg et al 2017;Racimo et al 2018) anti-conservative. We used a series of experiments to show that population structure is unlikely to bias the majority of our results, including showing that the signal is robust to the exclusion of rare alleles and performing a replication study in an external cohort.…”

“…It should be noted that the model we used to fit these data assumed no more than one shift in the optimal phenotype value, whereas this quantity is likely to vary continuously with environmental conditions for some traits. Models that additionally account for sexual dimorphism (Stulp and Barrett 2016), higher dimensional trait spaces (Simons et al 2018), and evolutionary history of multiple populations (Berg and Coop 2014;Racimo et al 2018) may be required to better understand the generality of these results across human populations and traits.…”

“…We applied S β within our framework to GWAS summary data for BMI, height (Wood et al 2014), and educational attainment (Okbay et al 2016) to assess its power for detecting selection, given that previous studies have suggested that these phenotypes may be under selection (Turchin et al 2012;Berg and Coop 2014;Robinson et al 2015;Field et al 2016;Racimo et al 2018). We observe that effect sizes are correlated with frequency, and that S β is a non-monotonic function of frequency (Fig.…”

“…While these studies have argued that polygenic selection is likely to be an important determinant of variation in traits such as height and skin pigmentation (Berg and Coop 2014), important questions remain about the evolutionary mechanisms that drive complex trait variation. In particular, most previous studies of selection on human complex traits have focused either on polygenic adaptation (Turchin et al 2012;Berg and Coop 2014;Field et al 2016;Berg et al 2017;Edge and Coop 2018;Racimo et al 2018) or stabilizing/negative selection Zeng et al 2018;Simons et al 2018), and have not incorporated mutational bias (i.e., the propensity for new mutations to be preferentially trait-increasing or preferentially trait-decreasing). To detect polygenic adaptation, studies have relied on genotype data from multiple populations to probe the frequency and linkage properties of trait-associated alleles as compared to a null based on genome-wide single nucleotide polymorphisms (Turchin et al 2012;Berg and Coop 2014;Racimo et al 2018), or have used haplotype-based statistics and dense sequence data from a single population (Field et al 2016).…”

Something old, something borrowed: Admixture and adaptation in human evolution

“…This allows for the identification of loci with strong evidence for adaptive introgression or barriers to introgression. Another recent method -PolyGraph -can take as input a previously inferred history of population splits and admixture events (in the form of an admixture graph [52] ), and detect episodes of polygenic adaptation, that result in a systematic increase or decrease of the frequency of trait-associated alleles [53] . …”