Résumé
One of population genetics oldest challenges is the inferrence of effective population size (Ne), a central parameter that impacts the genetic diversity and resilience of a population. Population and conservation genetics are keenly interested in its estimation. The constant improvement of sequencing methods and the quickly growing amount of available genetic data have motivated the development of methods capable of screening vast quantities of whole-genome data to estimate past Ne trajectories for thousands of generations. Such algorithms include Sequentially Markovian Coalescent (SMC) approaches such as the PSMC (Li and Durbin, 2011) and MSMC (Schiffels and Durbin, 2014), but also methods based on Approximate Bayesian Computation (ABC, Tavaré et al. 1997). ABC methods use summary statistics to infer Ne, which allow them to take into account whole-genome data such as Single Nucleotide Polymorphism (SNP). The SNPs' abundance (how often a SNP is present in a population), their allele frequency distribution and linkage disequilibrium (how allele frequencies at two SNPs are correlated) are good indicators of past population size. However, all those methods use modern DNA samples for their calculations. Ancient DNA, or DNA extracted from archeological samples, is a potential way to improve past Ne estimates by giving us a glimpse in the population' past genetic diversity. To test this hypothesis, we updated PopSizeABC (Boitard et al. 2016), an ABC method using whole-genome SNP data to infer past Ne trajectories, so that it can take into consideration DNA data from several time points. Using simulated modern and ancient DNA under various population' histories, we show that the addition of new summary statistics accounting for allele frequency variations across generations reduces prediction error. In the future, we plan to apply PopSizeABC to empirical data, using the European Eel (Anguilla anguilla) as a test species. The European Eel, being a species supposed to be panmictic, is an especially interesting model as both ancient and modern samples can be assumed to come from the same population.