The project aims to develop approximate Bayesian computation (ABC) and deep learning methods to infer the demographic history of populations, from current and ancient DNA sequences. We will study, in particular, the case of populations that separated at a given moment in the past, continuing or not to exchange migrants afterwards and going through demographic events such as expansions or contractions. The objective will be to determine which are the most efficient methods and, for the ABC methods, the most efficient summary statistics. The methods will then be applied to a set of human populations with different lifestyles: farmers, herders and hunter-gatherers. These populations coexist in various part of the World. For instance, farmers and hunter-gatherers live near each other in Equatorial Africa, and farmers and herders coexist in Central Asia (these populations are well studied in UMR7206). The methods that we will develop will allow inferring the joint demographic history of these neighbouring populations. We aim in particular to determine the timing of their splitting, and to assess whether they have been under expansion and/or contraction events since then, and if they have exchanged some migrants. The history of splitting of populations from different continents will also be studied. Finally, we will also investigate the contribution of ancient DNA samples to the quality of estimates, a question that has been little studied so far.
PhD student: Arnaud QUELIN
PhD supervisors: Frédéric Austerlitz, Flora Jay