Séminaire interne - Antoine Aragon (LPENS)

Abstract:

Cellular lineage tracing is important for understanding evolutionary processes in contexts ranging from developmental biology to cancer progression. Classical phylogenetic reconstruction methods, such as those based on Bayesian inference, infer time-calibrated trees representing the phylogenies of multiple sequence alignments. These methods rely on the presence of significant genetic diversity in the data, either due to large genomic mutation rates or to sufficiently long divergence times between individuals, and face limitations in cases where mutations are rare or sparsely distributed. We introduce a probabilistic framework to infer cellular growth, degradation, and mutation rates without requiring full phylogenetic tree reconstruction. Instead, we leverage easily derived allelic trees which are built using mutational distances between observed individuals. We show on synthetic data that our inference scheme performs well in the low mutation regime where the branch lengths of the allelic tree are not a good proxy for the time elapsed along a branch. Our method and analysis focus on neutral evolutionary processes, however the structure of the scheme makes it possible to include selection in the future, extending its applicability.