Frédéric Austerlitz

Inferring coalescent trees from genomic data has become a major subject in population genetics, particularly with the recent advances in tree sequence reconstruction methods. However, it remains unclear how well these methods perform for imbalanced genealogies. Such imbalances can arise from processes such as cultural transmission of reproductive success (CTRS) or positive selection. Using simulated genomic data, we benchmarked three major software packages, SINGER, Relate, and tsinfer, by comparing the imbalance of reconstructed trees by these methods with that of the true simulated trees, for three indices that quantify this imbalance. The three methods performed well under scenarios yielding balanced trees. However, their accuracy declined as imbalance increased. Performances also varied with mutation rate, recombination rate, and sample size. This study opens possibilities for applying these methods to infer CTRS or positive selection in large-scale genomic datasets, using simulation-based inference such as approximate Bayesian computation.

The ever-increasing availability of high-throughput DNA sequences and the development of numerous computational methods have led to considerable advances in our understanding of the evolutionary and demographic history of populations. Several demographic inference methods have been developed to take advantage of these massive genomic data. Simulation-based approaches, such as approximate Bayesian computation (ABC), have proved particularly efficient for complex demographic models. However, taking full advantage of the comprehensive information contained in massive genomic data remains a challenge for demographic inference methods, which generally rely on partial information from these data. Using advanced computational methods, such as machine learning, is valuable for efficiently integrating more comprehensive information. Here, we showed how simulation-based supervised machine learning methods applied to an extensive range of summary statistics are effective in inferring demographic parameters for connected populations. We compared three machine learning (ML) methods: a neural network, the multilayer perceptron (MLP), and two ensemble methods, random forest (RF) and the gradient boosting system XGBoost (XGB), to infer demographic parameters from genomic data under a standard isolation with migration model and a secondary contact model with varying population sizes. We showed that MLP outperformed the other two methods and that, on the basis of permutation feature importance, its predictions involved a larger combination of summary statistics. Moreover, they outperformed all three tested ABC algorithms. Finally, we demonstrated how a method called SHAP, from the field of explainable artificial intelligence, can be used to shed light on the contribution of summary statistics within the ML models.

The ever-increasing availability of high-throughput DNA sequences and the development of numerous computational methods have led to considerable advances in our understanding of the evolutionary and demographic history of populations. Several demographic inference methods have been developed to take advantage of these massive genomic data. Simulation-based approaches, such as approximate Bayesian computation (ABC), have proved particularly efficient for complex demographic models. However, taking full advantage of the comprehensive information contained in massive genomic data remains a challenge for demographic inference methods, which generally rely on partial information from these data. Using advanced computational methods, such as machine learning, is valuable for efficiently integrating more comprehensive information. Here, we showed how simulation-based supervised machine learning methods applied to an extensive range of summary statistics are effective in inferring demographic parameters for connected populations. We compared three machine learning (ML) methods: a neural network, the multilayer perceptron (MLP), and two ensemble methods, random forest (RF) and the gradient boosting system XGBoost (XGB), to infer demographic parameters from genomic data under a standard isolation with migration model and a secondary contact model with varying population sizes. We showed that MLP outperformed the other two methods and that, on the basis of permutation feature importance, its predictions involved a larger combination of summary statistics. Moreover, they outperformed all three tested ABC algorithms. Finally, we demonstrated how a method called SHAP, from the field of explainable artificial intelligence, can be used to shed light on the contribution of summary statistics within the ML models.

Historical linguistics strongly benefited from recent methodological advances inspired by phylogenetics. Nevertheless, no available method uses contemporaneous within-population linguistic diversity to reconstruct the history of human populations. Here, we developed an approach inspired from population genetics to perform historical linguistic inferences from linguistic data sampled at the individual scale, within a population. We built four within-population demographic models of linguistic transmission over generations, each differing by the number of teachers involved during the language acquisition and the relative roles of the teachers. We then compared the simulated data obtained with these models with real contemporaneous linguistic data sampled from Tajik speakers from Central Asia, an area known for its large within-population linguistic diversity, using approximate Bayesian computation methods. Under this statistical framework, we were able to select the models that best explained the data, and infer the best-fitting parameters under the selected models. The selected model assumes that the lexicon of individuals is the result of a vertical transmission by two teachers, with a specific lexicon for each teacher. This demonstrates the feasibility of using contemporaneous within-population linguistic diversity to infer historical features of human cultural evolution.

Cultural Transmission of Reproductive Success (CTRS) has been observed in many human populations as well as other animals. It consists in a positive correlation of non-genetic origin between the progeny size of parents and children. This correlation can result from various factors, such as the social influence of parents on their children, the increase of children’s survival through allocare from uncle and aunts, or the transmission of resources. Here, we study the evolution of genomic diversity through time under CTRS. We show that CTRS has a double impact on population genetics: (1) effective population size decreases when CTRS starts, mimicking a population contraction, and increases back to its original value when CTRS stops; (2) coalescent trees topologies are distorted under CTRS, with higher imbalance and higher number of polytomies. Under long-lasting CTRS, effective population size stabilises but the distortion of tree topology remains, which yields U-shaped Site Frequency Spectra (SFS) under constant population size. We show that this CTRS’ impact yields a bias in SFS-based demographic inference. Considering that CTRS was detected in numerous human and animal populations worldwide, one should be cautious that inferring population past histories from genomic data can be biased by this cultural process.

Effective population size ( N _e ) is a key parameter in evolutionary and conservation studies. It represents the number of individuals of an ideal panmictic population that would have the same genetic drift as the observed population and can be used to establish management programmes. Several methods have been developed to estimate this parameter. Currently, for studies with one sample in time, the simplest methods are based on linkage disequilibrium. These methods rely on simple models, and biases have been shown when populations deviate from the assumptions made in those models. This occurs in particular when populations are age-structured or have overlapping generations. Recently, several methods have been developed to correct such biases. Here, we develop analytical equations to predict linkage disequilibrium within age groups, and use such results to infer cohort size from samples of newborn individuals. We can in turn use these equations to estimate N _e accurately for a variety of species. Furthermore, using publicly available data, we apply our method to the white-crowned sparrow ( Zonotrichia leucophrys ).

Backgrounds and aims – Previous studies showed that robusta coffee (Coffea canephora Pierre ex A.Froehner), one of the two cultivated coffee species worldwide, can be classified in two genetic groups: the Guinean group originating in Upper Guinea and the Congolese group in Lower Guinea and Congolia. Although C. canephora of the Guinean group is an important resource for genetic improvement of robusta coffee, its germplasm is under-represented in ex situ gene banks and its genetic diversity and population structure have not yet been investigated. Methods – To overcome the limitations of living collections, we explored old herbarium specimens collected in Guinea and Côte d’Ivoire and conserved at the Muséum National d’Histoire Naturelle, Paris. First, we reviewed the history of collection missions in both countries and how the C. canephora herbaria from the Muséum were assembled. Then, using 23 nuclear microsatellite markers, factorial and model-based Bayesian analyses, we investigated the genetic diversity of 126 specimens and 36 controls, analysed their distribution among the Congolese and Guinean groups, and estimated admixture proportions for each individual.Key results – For the first time, we detected population genetic structure within the Guinean group of C. canephora. The Guinean genotypes can be assigned to five sub-groups with distinct geographic distribution, especially in Guinea where two sub-groups (Maclaudii and Gamé) are characterized by a low level of admixture due to geographical isolation.Conclusions – We showed how combining a literature review and genetic data from old herbarium specimens can shed light on previous observations made by botanists and guide further actions to better preserve native coffee plants in forest remnants of West Africa.

Chronic mountain sickness (CMS) is a pathological condition resulting from chronic exposure to high-altitude hypoxia. While its prevalence is high in native Andeans (>10%), little is known about the genetic architecture of this disease. Here, we performed the largest genome-wide association study (GWAS) of CMS (166 CMS patients and 146 controls living at 4,380 m in Peru) to detect genetic variants associated with CMS. We highlighted four new candidate loci, including the first CMS-associated variant reaching GWAS statistical significance (rs7304081; P = 4.58 × 10⁻⁹). By looking at differentially expressed genes between CMS patients and controls around these four loci, we suggested AEBP2, CAST, and MCTP2 as candidate CMS causal genes. None of the candidate loci were under strong natural selection, consistent with the observation that CMS affects fitness mainly after the reproductive years. Overall, our results reveal new insights on the genetic architecture of CMS and do not provide evidence that CMS-associated variants are linked to a strong ongoing adaptation to high altitude.