Biology subjects

Latrille, T.

Publications and source records attributed to Latrille, T..

14 recordsLinked to original sources

bioRxiv2026

Triangulating evidence to detect signatures of stabilizing selection acting on protein levels in humans

Stabilizing selection is thought to maintain quantitative traits near evolutionary optima, yet the extent to which it constrains human molecular phenotypes remains largely unknown. Here, we investigate signatures of stabilizing selection on the circulating levels of 2'923 plasma proteins in up to 50'000 UK Biobank participants by integrating complementary evidence across evolutionary timescales. We combine genetic signatures of selection accumulated over generations with associations between protein abundance and contemporary fitness-related outcomes. Genetic evidence was assessed using the frequency-dependent architecture of common variants with LDpred2 and a newly developed likelihood-based framework leveraging rare loss-of-function variant burdens. Contemporary evidence was obtained through quadratic associations between protein abundance and three fitness proxies: reproductive success, somatic health, and parental longevity. Across all approaches, negative selection coefficients predominated. Among significant estimates, 100% of common-variant and 92.4% of rare-variant estimates were consistent with stabilizing selection, compared with 99.8% for somatic health, 97.7% for parental longevity, and 89.3% for reproductive success. Despite capturing distinct aspects of selection, the approaches showed partial convergence. Rare-variant-based estimates correlated with health-score-derived coefficients (rho = 0.36, p = 5.2 x 10^-6) and independent measures of evolutionary constraint, including protein heritability, loss-of-function intolerance, and cross-species transcript evolutionary rates. Together, our findings suggest that stabilizing selection is widespread across the human plasma proteome and highlight the value of integrating genetic architecture with fitness-related phenotypes to understand evolutionary constraints on molecular variation.

evolutionary biology↗

bioRxiv2026

Heterogeneous evolution of effective recombination landscapes across percomorph fishes

Recombination shapes allelic combinations across the genome and plays a central role in adaptation and speciation. While recombination rates are known to vary and evolve, how recombination landscapes diverge between populations and which processes drive this divergence remains poorly understood. Here, we use the Population Recombination Divergence Index (PRDI) to quantify divergence in effective recombination landscapes between seven pairs of populations and species from four percomorph fish lineages (cichlids, sticklebacks, killifish, and clownfish). While accounting for stochasticity and confounders of linkage disequilibrium (LD) -based recombination rate estimates, we find that recombination landscapes diverged substantially in all comparisons. This divergence generally increases with genetic differentiation and divergence time, but the amount of PRDI accumulated per unit of genetic divergence (FST), net nucleotide divergence (da), or divergence time varies markedly among population and species pairs. Comparisons across systems suggest that PRDI accumulation tends to be faster (per unit of FST, da, or divergence time) in pairs with higher inferred gene flow. We also introduce a local version of PRDI (_l) to better characterize genomic heterogeneity in effective recombination divergence. We find that higher inferred gene flow between populations tends to be associated with stronger intragenomic correlation between local FST and _l. Finally, we propose a conceptual framework for investigating how divergent selection and gene flow may interact to shape effective recombination landscape divergence. Our work highlights the heterogeneous dynamics of effective recombination landscape evolution, and motivates further theoretical and empirical efforts to integrate recombination landscape evolution into studies of adaptation and speciation.

evolutionary biology↗

bioRxiv2026

DTX3L-PARP9 co-evolve as a single adaptive unit linking ubiquitination and ADP-ribosylation in antiviral immunity

Functional protein complexes are central to innate immunity, by integrating and diversifying signaling pathways of immune responses. Understanding how such complexes evolve is therefore key to elucidate the molecular basis of immune adaptation. Using a comparative genomics framework, we show that the DTX3L-PARP complex, which integrates ubiquitin-dependent and ADP-ribosylation pathways to regulate antiviral immunity, is an ancient and rapidly evolving functional module in vertebrates. We demonstrate that this complex has undergone recurrent adaptive evolution, with strong signatures of co-evolution between its components, consistent with early functional coupling and coordinated evolution as a single adaptive unit rather than independent proteins. Adaptive residues are enriched at protein-protein interaction surfaces within the RING-DTC domain of DTX3L, while PARP9 and PARP14 concentrate adaptive changes in regions involved in complex assembly and catalytic activities, consistent with hallmarks of pathogen-driven selection. These evolutionary signals remain detectable at short evolutionary timescales, including within species and sub-species, indicating ongoing adaptive evolution of the complex. These patterns support a model in which pathogens recurrently target the DTX3L-PARP axis to disrupt ubiquitin-ADP-ribose signaling, either by simultaneously interfering with catalytic domains or by destabilizing the complex interfaces. Altogether, our findings reveal the DTX3L-PARP9-PARP14 complex as a co-evolving adaptive module shaped by persistent host-virus arms races, highlighting how interconnected post-translational modification systems evolve in concert to sustain vertebrate antiviral immunity.

evolutionary biology↗

bioRxiv2025

RegEvol: detection of directional selection in regulatory sequences through phenotypic predictions and phenotype-to-fitness functions

Regulatory DNA controls when and where genes are expressed, making it a key driver of phenotypic evolution. Yet detecting selection in non-coding regions remains difficult, as most approaches rely on sequence conservation or changes in substitution rate rather than molecular effects. RegEvol bridges this gap by linking machine learning-based predictions of transcription factor binding to explicit evolutionary models. It uses the distribution of predicted mutational effects to infer fitness functions under different evolutionary scenarios including random drift, stabilising selection, and directional selection. Through maximum-likelihood estimation, it identifies the regime that best explains observed changes along a lineage from an ancestral sequence. When substitution numbers are limited, such as along short evolutionary branches, likelihood differences can be aggregated across sets of regulatory elements to increase statistical power. RegEvol corrects biases that affected previous tests based on machine learning of transcription factor binding, while remaining conservative across different levels of divergence. Applied to over 3 million Drosophila melanogaster regulatory regions, we identify 5.1% under directional selection, enriched near reproductive and immune genes. Applying the aggregation strategy to human CTCF binding across tissues reveals enrichment of directional signals in nervous and male reproductive systems. The framework is readily applicable to experimentally detected regulatory elements with alignable ancestral sequences and is flexible to future advances in understanding regulatory function, providing a powerful basis for investigating adaptation in non-coding regions.

evolutionary biology↗

bioRxiv2025

Phylogram instead of chronogram when assessing the neutral evolution of a trait

Comparative analyses of trait evolution aim to uncover the different evolutionary forces shaping phenotypic diversity among species. This is typically done by fitting various evolutionary models to observed trait changes along a species tree. For instance, under neutral evolution, trait values are modelled as changing randomly along the branches of the tree. In contrast, for a trait under selection, species are typically assumed to track an optimal trait value, which itself may shift along the tree. Here, rather than relying on alternative models to discriminate among evolutionary scenarios, we focus on the underlying species tree, and specifically the units in which its branch lengths are measured. Species trees are usually represented as chronograms, with branch lengths proportional to time. The rationale is that time correlates with trait changes through its connection to the number of generations. However, since the generation time of species can also vary along the phylogenetic tree, we argue that chronograms introduces biases. In contrast, phylograms with branch lengths in units of sequence divergence will account for the effect of changing generation time. In this study, we develop a method to test whether a phylogram provides a better fit than a chronogram for modelling trait evolution. We show using simulations that, for a trait evolving neutrally, the fit of a random evolution model has more support on a phylogram than on a chronogram. However, comparing models and testing different scenarios of selection using a phylogram leads to incorrect predictions. Given these results, we argue that we should use phylograms instead of chronograms when assessing the neutral evolution of a trait. Nevertheless, we support the fact that we should generally continue to use chronograms to model selection acting on a quantitative trait. Lay summaryWhen we look at different species, we can see that some traits are different between them. For example, the brain size of different primate and human lineages varies. The question we ask is: does the time that passed explain the differences in their traits? Or are genetic differences between species better at explaining the differences? The standard in evolutionary biology is to use time to explain differences in traits. However, if the trait is not under natural selection and instead is changing due to accumulating mutations, the genetic differences between species instead of time should best explain the differences in traits. As a result, we argue that using both genetic differences and species divergence time simultaneously provides a better understanding of how traits evolve across species.

evolutionary biology↗

bioRxiv2025

Genetic basis of the evolution of vertical bars in clownfishes

Clownfish exhibit striking color patterns, characterized primarily by the presence of zero to three vertical white bars, along with three main colors: orange, white, and black. The common ancestor of clownfish likely possessed three vertical bars, with several instances of gains and losses occurring throughout clownfish evolutionary history over the past 10 million years. However, the evolutionary genomic mechanisms underlying the gain or loss of vertical bars remain unknown. In this study, we tested whether vertical bar transitions across the clownfish phylogeny were associated with changes in non-synonymous to synonymous substitution rates (dN/dS values). Our analyses identified pigmentation-related genes that underwent changes in selective pressure, including gch2, oca2, and vps11, which are linked to melanophores, iridophores, and visual function. Additionally, pmel, a key melanogenesis gene, was found under positive selection, suggesting its role in shaping bar patterning. These results provide new insights into the genomic basis of coloration in clownfish, highlighting how selection and genetic variation influence phenotypic evolution.

evolutionary biology↗

bioRxiv2024

Widespread selection relaxation in aquatic mammals

While mammals are predominantly terrestrial, several lineages within them have independently transitioned to aquatic environments and exhibit a great variety of evolutionary changes, which ultimately lead to irreversible transitions to aquatic lifestyle. Such changes should also be detectable at the molecular level and, for instance, olfactory genes have been found to evolve under reduced selection and functionality in whales. However, the presence of a molecular signature related to aquatic transitions across other genes remains unknown as is the degree to which this affects only fully aquatic groups or also semiaquatic ones. Here, we use a Bayesian framework to investigate differences in the strength of purifying selection among terrestrial, semiaquatic and fully aquatic mammals using a set of 1000 orthologous genes, while accounting for shared ancestry, and controlling for body mass and effective population size. We found that selection relaxation linked to aquatic transitions is not only occurring in olfactory genes, but also detected across 20% of the other genes analysed here. This consistent trend of genes under weaker selection is inferred in both semi- and fully aquatic mammals although the pattern is stronger in the latter. While most of the genes analysed here likely remain functional in all mammals, their evolution is inferred to be neutral in a substantial number of aquatic species, consistently with genes that might be losing their functionality. The inferred widespread relaxation of selection among genes is consistent with a macroevolutionary scenario where secondary transitions to aquatic environment eventually become irreversible.

evolutionary biology↗

bioRxiv2023

Detecting diversifying selection for a trait from within and between-species genotypes and phenotypes

To quantify selection acting on a trait, methods have been developed using either within or between-species variation. However, methods using within-species variation do not integrate the changes at the macro-evolutionary scale. Conversely, current methods using between-species variation usually discard within-species variation, thus not accounting for processes at the micro-evolutionary scale. The main goal of this study is to define a neutrality index for a quantitative trait, by combining within- and between-species variation. This neutrality index integrates nucleotide polymorphism and divergence for normalizing trait variation. As such, it does not require estimation of population size nor of time of speciation for normalization. Our index can be used to seek deviation from the null model of neutral evolution, and test for diversifying selection. Applied to brain mass and body mass at the mammalian scale, we show that brain mass is under diversifying selection. Finally, we show that our test is not sensitive to the assumption that population sizes, mutation rates and generation time are constant across the phylogeny, and automatically adjust for it.

evolutionary biology↗

bioRxiv2023

Improved estimation of molecular evolution coupling stochastic simulations and deep learning

Models have always been central to inferring molecular evolution and to reconstructing phylogenetic trees. Their use typically involves the development of a mechanistic framework reflecting our understanding of the underlying biological processes, such as nucleotide substitutions, and the estimation of model parameters by maximum likelihood or Bayesian inference. However, deriving and optimizing the likelihood of the data is not always possible under complex evolutionary scenarios or tractable for large datasets, often leading to unrealistic simplifying assumptions in the fitted models. To overcome this issue, we couple stochastic simulations of genome evolution with a new supervised deep learning model to infer key parameters of molecular evolution. Our model is designed to directly analyze multiple sequence alignments and estimate per-site evolutionary rates and divergence, without requiring a known phylogenetic tree. The accuracy of our predictions matches that of likelihood-based phylogenetic inference, when rate heterogeneity follows a simple gamma distribution, but it strongly exceeds it under more complex patterns of rate variation, such as codon models. Our approach is highly scalable and can be efficiently applied to genomic data, as we show on a dataset of 26 million nucleotides from the clownfish clade. Our simulations also show that the per-site rates obtained by deep learning increase the likelihood of the true tree and could therefore lead to more accurate phylogenetic inference. We propose that future advancements in phylogenetic analysis will benefit from a semi-supervised learning approach that combines deep-learning estimation of substitution rates, which allows for more flexible models of rate variation, and probabilistic inference of the phylogenetic tree, which guarantees interpretability and a rigorous assessments of statistical support.

evolutionary biology↗

bioRxiv2023

Mammalian protein-coding genes exhibit widespread beneficial mutations that are not adaptive

Mutations can be beneficial by bringing innovation to their bearer, allowing them to adapt to environmental change. These mutations are typically unpredictable since they respond to an unforeseen change in the environment. However, mutations can also be beneficial because they are simply restoring a state of higher fitness that was lost due to genetic drift in a stable environment. In contrast to adaptive mutations, these beneficial non-adaptive mutations can be predicted if the underlying fitness landscape is stable and known. The contribution of such non-adaptive mutations to molecular evolution has been widely neglected mainly because their detection is very challenging. We have here reconstructed protein-coding-gene fitness landscapes shared between mammals, using mutation-selection models and a multi-species alignments across 87 mammals. These fitness landscapes have allowed us to predict the fitness effect of polymorphisms found in 28 mammalian populations. Using methods that quantify selection at the population level, we have confirmed that beneficial non-adaptive mutations are indeed positively selected in extant populations. Our work confirms that deleterious substitutions are accumulating in mammals and are being reverted, generating a balance in which genomes are damaged and restored simultaneously at different loci. We observe that beneficial non-adaptive mutations represent between 15% and 45% of all beneficial mutations in 24 of 28 populations analyzed, suggesting that a substantial part of ongoing positive selection is not driven solely by adaptation to environmental change in mammals. Significance statementThe extent to which adaptation to changing environments is shaping genomes is a central question in molecular evolution. To quantify the rate of adaptation, population geneticists have typically used signatures of positive selection. However, mutations restoring an ancestral state of higher fitness lost by genetic drift are also positively selected, but they do not respond to a change in the environment. In this study, we have managed to distinguish beneficial mutations that are due to changing environments and those that are restoring pre-existing functions in mammals. We show that a substantial proportion of beneficial mutations cannot be interpreted as adaptive.

evolutionary biology↗

bioRxiv2022

Genes and sites under adaptation at the phylogenetic scale also exhibit adaptation at the population-genetic scale

Adaptation in protein-coding sequences can be detected from multiple sequence alignments across species, or alternatively by leveraging polymorphism data inside a population. Across species, quantification of the adaptive rate relies on phylogenetic codon models, classically formulated in terms of the ratio of non-synonymous over synonymous substitution rates. Evidence of an accelerated non-synonymous substitution rate is considered a signature of pervasive adaptation. However, because of the background of purifying selection, these models are potentially limited in their sensitivity. Recent developments have led to more sophisticated mutation-selection codon models aimed at making a more detailed quantitative assessment of the interplay between mutation, purifying and positive selection. In this study, we conducted a large-scale exome-wide analysis of placental mammals with mutation-selection models, assessing their performance at detecting proteins and sites under adaptation. Importantly, mutation-selection codon models are based on a population-genetic formalism and thus are directly comparable to McDonald & Kreitman tests at the population level to quantify adaptation. Taking advantage of this relationship between phylogenetic and population genetics, we integrated divergence and polymorphism data across the entire exome for 29 populations across 7 genera, and showed that proteins and sites detected to be under adaptation at the phylogenetic scale are also under adaptation at the population-genetic scale. Altogether, our exome-wide analysis shows that phylogenetic mutation-selection codon models and population-genetic test of adaptation can be reconciled and are congruent, paving the way for integrative models and analyses across individuals and populations. Significance StatementDetecting genes under adaptation represents a key step in the decoding of genomes. Several methods have been proposed, focussing either on the short time scale (population genetics, e.g. human populations), or on the long time scale (phylogenetics, e.g. across mammals). However, the accuracy of these methods is still under debate, and it is still unclear whether the signatures of adaptation are congruent across evolutionary scales. In this study, using novel phylogenetic methods and gathering genome data across and within species, we show that the signatures of adaptation at the phylogenetic and population-genetic scales can be reconciled. While providing a mutual confirmation of the two approaches, our work paves the way for further methodological integration between micro- and macro-evolutionary genomics.

evolutionary biology↗

bioRxiv2021

An improved codon modeling approach for accurate estimation of the mutation bias

Nucleotide composition in protein-coding sequences is the result of the equilibrium between mutation and selection. In particular, the nucleotide composition differs between the three coding positions, with the third position showing more extreme composition than the first and the second positions. Yet, phylogenetic codon models do not correctly capture this phenomenon and instead predict that the nucleotide composition should be the same for all 3 positions of the codons. Alternatively, some models allow for different nucleotide rates at the three positions, a problematic approach since the mutation process should in principle be blind to the coding structure and homogeneous across coding positions. Practically, this misconception could have important consequences in modelling the impact of GC-biased gene conversion (gBGC) on the evolution of protein-coding sequences, a factor which requires mutation and fixation biases to be carefully disentangled. Conceptually, the problem comes from the fact that phylogenetic codon models cannot correctly capture the fixation bias acting against the mutational pressure at the mutation-selection equilibrium. To address this problem, we present an improved codon modeling approach where the fixation rate is not seen as a scalar anymore, but as a tensor unfolding along multiple directions, which gives an accurate representation of how mutation and selection oppose each other at equilibrium. Thanks to this, this modelling approach yields a reliable estimate of the mutational process, while disentangling fixation probabilities in different directions.

evolutionary biology↗

bioRxiv2021

Inferring long-term effective population size with Mutation-Selection models

Mutation-selection phylogenetic codon models are grounded on population genetics first principles and represent a principled approach for investigating the intricate interplay between mutation, selection and drift. In their current form, mutation-selection codon models are entirely characterized by the collection of site-specific amino-acid fitness profiles. However, thus far, they have relied on the assumption of a constant genetic drift, translating into a unique effective population size (Ne) across the phylogeny, clearly an unreasonable hypothesis. This assumption can be alleviated by introducing variation in Ne between lineages. In addition to Ne, the mutation rate () is susceptible to vary between lineages, and both should co-vary with life-history traits (LHTs). This suggests that the model should more globally account for the joint evolutionary process followed by all of these lineage-specific variables (Ne, , and LHTs). In this direction, we introduce an extended mutation-selection model jointly reconstructing in a Bayesian Monte Carlo framework the fitness landscape across sites and long-term trends in Ne, and LHTs along the phylogeny, from an alignment of DNA coding sequences and a matrix of observed LHTs in extant species. The model was tested against simulated data and applied to empirical data in mammals, isopods and primates. The reconstructed history of Ne in these groups appears to correlate with LHTs or ecological variables in a way that suggests that the reconstruction is reasonable, at least in its global trends. On the other hand, the range of variation in Ne inferred across species is surprisingly narrow. This last point suggests that some of the assumptions of the model, in particular concerning the assumed absence of epistatic interactions between sites, are potentially problematic.

evolutionary biology↗

bioRxiv2021

A theoretical approach for quantifying the impact of changes in effective population size and expression level on the rate of coding sequence evolution

Molecular sequences are shaped by selection, where the strength of selection relative to drift is determined by effective population size (Ne). Populations with high Ne are expected to undergo stronger purifying selection, and consequently to show a lower substitution rate for selected mutations relative to the substitution rate for neutral mutations ({omega}). However, computational models based on biophysics of protein stability have suggested that{omega} can also be independent of Ne, a result proven under general conditions. Together, the response of{omega} to changes in Ne depends on the specific mapping from sequence to fitness. Importantly, an increase in protein expression level has been found empirically to result in decrease of{omega} , an observation predicted by theoretical models assuming selection for protein stability. Here, we derive a theoretical approximation for the response of{omega} to changes in Ne and expression level, under an explicit genotype-phenotype-fitness map. The method is generally valid for additive traits and log-concave fitness functions. We applied these results to protein undergoing selection for their conformational stability and corroborate out findings with simulations under more complex models. We predict a weak response of{omega} to changes in either Ne or expression level, which are interchangeable. Based on empirical data, we propose that fitness based on the conformational stability may not be a sufficient mechanism to explain the empirically observed variation in{omega} across species. Other aspects of protein biophysics might be explored, such as protein-protein interactions, which can lead to a stronger response of{omega} to changes in Ne.

evolutionary biology↗