Société Française de Bioinformatique

Mots-Clés

populations genomics pedigree demographic inference livestock species ARG deep learning

Description

Demographic inference is a central tool for reconstructing the genetic history of a population. Recently, several new approaches have brought this tool into a new era, that of the exhaustive use of whole-genome sequence. These approaches can be divided into two categories: demographic inference based on ancestral recombination graphs (ARGs) and deep learning based inference. However, without any other source of information, it is difficult to assess the actual degree of accuracy of these methods. In the case of livestock species, we usually have access to an additional information, rather unique and valuable, the pedigree over several generations. This set of relationships makes it possible to compare inferred demographic histories with the exact, albeit incomplete, history of the population.
In this thesis project, we aim to use the pedigree information on the one hand, to assess the results obtained with the new approaches of demographic inference, and, on the other hand, to refine methods for ARG estimation. Therefore, we propose a research program divided into three main tasks: (i) to appropriate new approaches (ARG, deep learning) in demographic inference on goat and sheep datasets, (ii) to compare these approaches together, in particular to evaluate their inference of the effective size of a population in the light of genealogical information, and (iii) to integrate this information into the inference of ARGs.
You will be welcomed in the CHAMADE team (“CHAracterization and MAnagement of Diversity”) of the GenPhySE research unit (https://genphyse.inrae.fr/), located in the Occitanie-Toulouse research centre (31320, Castanet-Tolosan). The CHAMADE team is part of the “Diversity and Selection” scientific division of the research unit. The team is interested in methodological issues in the field of population genomics, genetic evaluation of livestock species and quantitative and evolutionary genetics. On deep learning approaches for demographic inference, collaboration is also planned with the BioInfo team from the Laboratoire Interdisciplinaire des Sciences du Numérique (LISN, Paris-Saclay University).