Inferring genome-wide correlations of mutation fitness effects between populations

Abstract

The effect of a mutation on fitness may differ between populations, depending on environmental and genetic context. Experimental studies have shown that such differences exist, but little is known about the broad patterns of such differences or the factors that drive them. To quantify genome-wide patterns of differences in mutation fitness effects, we extended the concept of a distribution of fitness effects (DFE) to a joint DFE between populations. To infer the joint DFE, we fit parametric models that included demographic history to genomic data summarized by the joint allele frequency spectrum. Using simulations, we showed that our approach is statistically powerful and robust to many forms of model misspecification. We then applied our approach to populations of Drosophila melanogaster, wild tomatoes, and humans. We found that mutation fitness effects are overall least correlated between populations in tomatoes and most correlated in humans, corresponding to overall genetic differentiation. In D. melanogaster and tomatoes, mutations in genes involved in immunity and stress response showed the lowest correlation of fitness effects, consistent with environmental influence. In D. melanogaster and humans, deleterious mutations showed a lower correlation of fitness effects than tolerated mutations, hinting at the complexity of the joint DFE. Together, our results show that the joint DFE can be reliably inferred and that it offers extensive insight into the genetics of population divergence. 2 mutation's effect on fitness, population genetics theory can predict a great deal; for example, how likely 3 the mutation is to be lost from or fix in the population. But population genetics theory cannot predict 4 how likely a new mutation is to have a given effect on fitness. It is known that typically the majority of 5 mutations are deleterious (reduce fitness) or nearly neutral (negligible effect on fitness), so only a small 6 minority are adaptive (increase fitness). But these three categories encompass a continuum of fitness effects. 7This continuum is quantified by the distribution of fitness effects (DFE) among new mutations (Eyre-Walker 8

Haplotype-based inference of the distribution of fitness effects

“…2013; Lynch 2016). Our results argue in favor of conserved selective coefficients over time in humans, in line with recent results (Fortier et al . 2019).…”

“…Two other extensions have been taken to model the correlation between the fitness effects of multiple nonsynonymous alleles at a particular position (Ragsdale et al . 2016) and to calculate the joint DFE between pairs of populations (Fortier et al . 2019).…”

A community-maintained standard library of population genetic models

“…While population genetics has always used statistical methods to make inferences from data, the degree of sophistication of the questions, models, data, and computational approaches used have all increased over the past two decades. Currently, there exist a myriad of computational methods that can infer the histories of populations ( Gutenkunst et al, 2009 ; Li and Durbin, 2011 ; Excoffier et al, 2013 ; Schiffels and Durbin, 2014 ; Terhorst et al, 2017 ; Ragsdale and Gravel, 2019 ), the distribution of fitness effects ( Boyko et al, 2008 ; Kim et al, 2017 ; Tataru et al, 2017 ; Fortier et al, 2019 ; Huang and Siepel, 2019 ; Vecchyo et al, 2019 ), recombination rates ( McVean et al, 2004 ; Chan et al, 2012 ; Lin et al, 2013 ; Adrion et al, 2020 ; V Barroso et al, 2019 ), and the extent of positive selection in genome sequence data ( Kim and Stephan, 2002 ; Eyre-Walker and Keightley, 2009 ; Alachiotis et al, 2012 ; Garud et al, 2015 ; DeGiorgio et al, 2016 ; Kern and Schrider, 2018 ; Sugden et al, 2018 ). While these methods have undoubtedly increased our understanding of genetic and evolutionary processes, very little has been done to systematically benchmark the quality of these inferences or their robustness to deviations from their underlying assumptions.…”

Inferring the Demographic History of Inbred Species From Genome-Wide SNP Frequency Data

“…2017; Tataru et al . 2017; Fortier et al . 2019), allowing researchers to model patterns of selection while simultaneously controlling for demography (Williamson et al .…”