EvoLunch Seminar: Curro Campuzano (University of Antwerp)

Population genomic studies suggest that speciation in the presence of gene flow is common across many taxa and that divergent selection plays an important role in postzygotic reproductive isolation between incipient species. One can quantify the effect of divergently selected alleles on neutral gene flow by means of an effective migration rate (m_e) that varies along the genome. Conversely, inference of $m_e$ along the genome from observed neutral variation allows one to map barrier loci and obtain insight into the genetic architecture of reproductive isolation. Existing approaches to estimate $m_e$ along the genome from population genomic data are, however, restricted to consider independent genomic windows and ignore the correlations across physically linked windows. The motivation of my research on ARGs is twofold. First, is it possible for a given $m_e$ profile (or, better, genetic architecture of divergent selection) to simulate ARGs (and hence genetic data) backwards in time? This would allow one to perform more appropriate absolute-goodness-of-fit test and improve existing ABC methods. Second, can we use ARGs to do inference of $m_e$ while accounting for the correlations in genealogies along the genome? This may allow one to obtain improved statistical inference and properly calibrated uncertainty intervals. I have explored these questions using a sequentially Markovian coalescent approximation and have some preliminary results.