Résumé
A classical approach in population genetic inference is to assume that genome wide patterns of genetic diversity are overwhelmingly determined by demography and that selection has only local effects along the genome. This allows to make inferences of neutral processes from genome-wide data and make inferences on selection from regions with outlier patterns. However, over the last years, the limits of this approach are becoming more and more evident. It has been shown that background selection and recurrent selective sweeps can have pervasive effects in the genome, biasing demographic inference. In addition, study of selection based on detection of outliers is very restrictive, as it only focus on loci with major effects. In this work, we propose to use Approximate Bayesian Computation (ABC) for the joint inference of demography and selection from a population sampled several times at different generations. Using ABC we were able to classify populations either as quasi-neutral or under strong linked selection. We were able to estimate census and effective population sizes, proportion of loci under strong selection (loci with Ns > 1) and genetic load (sensu substitution load). These results demonstrate the potential of ABC to address the joint inference of demography and selection and gain new insight on the processes under specific mechanistic models.