Abstract: The enormous and still growing amount of genetic data now available provides more detailed insights into the evolution of many organisms. In particular, we now have hundreds of thousands of human and microbial genomes available. The data show much greater flexibility of bacterial genomes than previously thought. On the other hand, bacteria are mostly haploid, meaning that they do not recombine their genomes as frequently as humans and other diploids.
Taken together, this makes their evolution both extremely flexible and highly dependent on the underlying clonal relationship of the ancestors. I will show why dependence on this ancestral tree or graph is essential when considering microbial evolution. In particular, we will consider a tree-indexed Markov chain that models the evolution of microbial immunity memory arrays.
In addition, I will explain why statistical learning tools for population genetics face unique challenges and how this relates to coalescent trees. Machine learning techniques can process large amounts of genetic data, but their black-box nature makes it difficult to gain biological insights. I will highlight recent advances in machine learning for population genetics and bridge the gap between optimal model-based mutation rate estimators and neural networks.

us02web.zoom.us/j/4082603129