Sujet : Development of statistical methods for the reconstruction of routes of biological invasion: inference of complex evolutionary scenario from genomic data using admixture graphs.
Dates : 1er mars 2025 – 31 août 2026
Responsables CBGP : M. Gautier & A. Estoup
Admixture graphs (AG) describe the demographic history of a set of populations as a directed acyclic graph that represents population splits and merges. They are particularly useful in studying biological invasions, as they can model the recent introduction history of individuals from native and invasive population samples. AGs have great, as yet unexplored, potential for selecting a set of (most probable) invasion scenarios from large-scale population polymorphism data (obtained from the entire genome of the organisms under study). AGs can be inferred using statistical methods that employ simplified models of evolution based on allele frequency covariances between population samples. The selected graphs can then be exploited by more sophisticated methods, that use complex models and likelihood free inference techniques for model choice, parameter estimation, or goodness of fit. The inference of AGs is an active field of research, with recent methods based on maximum likelihood or Bayesian approaches. Because these methods need to explore the huge space of possible graphs, they are subject to a number of algorithmic and mathematical challenges.
The aim of the post-doctoral project is to study the behaviour of these methods on simulated and real datasets (mainly full-genome polymorphism data for a large number of population samples from two invasive insect species), and, based on the results, to propose new improved statistical methods. Possible lines of research could include
