Inference of past historical events using ABC and MCMC methods on population genomics data sets. Applications to human populations
Résumé
New computer-intensive estimation techniques such as Approximate Bayesian Computation (ABC) and Monte Carlo Markov chains (MCMC) allows inferring unknown parts of the history of species from contemporary population genetics and genomics data. I will illustrate these possibilities with several examples. In this context, we performed a study on worldwide human populations, in which by applying MCMC methods on a large set of populations with different lifestyles (farmers, herder and hunter-gatherers), we were able to show that these lifestyles strongly impacted the expansion patterns of these populations: farmers show strong expansion signals, herders weak expansion and hunter-gatherers no expansion at all. Moreover we showed that rapidly mutating markers like microsatellites allowed us to infer the recent Neolithic expansion, while slowly mutating markers like sequences allowed us to infer a more ancient Paleolithic expansion. The validity of this approach was verified through a simulation study.
More recently, we developed a parametric ABC method for whole-genome sequencing data. We studied which combinations of summary statistics allow best to infer the demographic processes occurring in the populations under study. We applied this method to human populations from the 1000 Genome project. Our first results show that we can infer contrasted results when comparing Eurasian and African populations.