Mutation-selection models of coding sequence evolution with site-heterogeneous amino acid fitness profiles

Abstract

Modeling the interplay between mutation and selection at the molecular level is key to evolutionary studies. To this end, codonbased evolutionary models have been proposed as pertinent means of studying long-range evolutionary patterns and are widely used. However, these approaches have not yet consolidated results from amino acid level phylogenetic studies showing that selection acting on proteins displays strong site-specific effects, which translate into heterogeneous amino acid propensities across the columns of alignments; related codon-level studies have instead focused on either modeling a single selective context for all codon columns, or a separate selective context for each codon column, with the former strategy deemed too simplistic and the latter deemed overparameterized. Here, we integrate recent developments in nonparametric statistical approaches to propose a probabilistic model that accounts for the heterogeneity of amino acid fitness profiles across the coding positions of a gene. We apply the model to a dozen real proteincoding gene alignments and find it to produce biologically plausible inferences, for instance, as pertaining to site-specific amino acid constraints, as well as distributions of scaled selection coefficients. In their account of mutational features as well as the heterogeneous regimes of selection at the amino acid level, the modeling approaches studied here can form a backdrop for several extensions, accounting for other selective features, for variable population size, or for subtleties of mutational features, all with parameterizations couched within population-genetic theory.codon substitution | Dirichlet process | phylogeny | selection coefficients F ollowing the seminal works of Muse and Gaut (1) and Goldman and Yang (2), most early applications of codon-based evolutionary models were focused on evaluations of selective effects operating at different positions along a gene or at different time points along the phylogeny (see refs. 3, 4 for reviews). Many of these approaches have modeled selective effects using a parameter representing the nonsynonymous/synonymous rate ratio. However, this may not be ideal, in particular because it amounts to ignoring differences between different pairs of possible amino acid replacements resulting from nonsynonymous point mutations. In recent years, questions regarding selective effects have diversified, such as in the work of Yang and Nielsen (5), who propose a test for selection on codon usage. This test is based on models that invoke a multidimensional specification of scaled selection coefficients, based on either 20 or 61 (under the universal genetic code) scaled fitness parameters-adding 19 or 60 degrees of freedom to the underlying codon substitution model-in contrast with the more conventional use of the single nonsynonymous/synonymous rate ratio parameter, viewing all nonsynonymous events as equivalent (e.g., see ref. 6). By assigning scaled fitness parameters to each of the 20 amino acids, or to the 61 sense codons, Yang and Nie...

Contingency and entrenchment in protein evolution under purifying selection

“…This approach accounts for possible dependencies among sites, whereas most models of protein sequence evolution along a phylogeny assume that sites evolve independently (51)(52)(53)(54)(55). Such phylogenetic models necessarily disregard any possible epistatic interactions between sites.…”

“…Indeed, although some more sophisticated models have been proposed (e.g., refs. [45][46][47][48][49][50], all commonly used phylogenetic models of long-term protein evolution assume that epistasis is absent so that sites evolve independently (51)(52)(53)(54)(55)(56).…”

Extensively Parameterized Mutation–Selection Models Reliably Capture Site-Specific Selective Constraint

“…Previous quantitative comparisons of swMutSel and pbMutSel inferences have focused nearly exclusively on asking how well they recapitulate the gene-wide distribution of S, or similarly the gene-wide proportions of deleterious and beneficial substitutions (Rodrigue et al 2010;Tamuri et al 2012Tamuri et al , 2014Rodrigue 2013;Rodrigue and Lartillot 2014). In spite of these efforts, however, there remains no conclusive evidence supporting either swMutSel or pbMutSel as the more reliable inference approach.…”

“…Findings from previous mutation-selection model studies would suggest that natural simulations would favor the swMutSel platform, which is known to estimate large proportions of deleterious changes, and conversely DMS simulations would favor the pbMutSel platform, which tends to infer strictly unimodal S distributions (Rodrigue et al 2010;Tamuri et al 2012Tamuri et al , 2014Rodrigue 2013). Therefore, the different features across our simulation sets allowed us to directly contrast how each mutation-selection inference platform behaves on data with realistic levels of evolutionary heterogeneity, without biasing results towards one particular implementation.…”

“…The first implementation, known as swMutSel, estimates site-specific fitness parameters as fixed-effect variables through a maximum penalized-likelihood (MPL) approach (Tamuri et al 2012(Tamuri et al , 2014. The second implementation, available in the PhyloBayes software package, instead employs a Dirichlet Process (DP) Bayesian framework and models site-specific fitness parameters as random effects (Rodrigue et al 2010;Rodrigue and Lartillot 2014). For simplicity, we will refer to the latter implementation as "pbMutSel" throughout this article.…”