2026/01/03 by Tobias Dieselhorst, Johannes Berg · 1 voice
Biochemistry, Genetics and Molecular Biology · #Single-cell and spatial transcriptomics #Cancer Genomics and Diagnostics #Genomics and Phylogenetic Studies
paper · doi:10.1093/molbev/msag002
openalex publication_date 2026/01/03 · openalex created_date 2026/01/08 · openalex updated_date 2026/06/19
Understanding the evolutionary dynamics of cell populations requires models that link observed phylogenetic patterns to the underlying processes of cell division, death, and mutation. Classical phylodynamic inference methods-developed primarily for macroevolutionary settings-assume that mutations accrue in calendar time and often rely on a molecular clock. Here, we introduce a framework that ties mutations to discrete birth (division) events. In this setting, mutations accumulate via a compound Poisson process, capturing both visible and hidden cell divisions within the reconstructed phylogenetic tree. We present a computationally efficient dynamic programming algorithm to compute the likelihood based on tree topologies with associated mutations, integrating over latent variables such as branch durations and unobserved cell divisions. Our method is applicable to large-scale single-cell datasets, and we demonstrate its utility on simulated data and on single-cell phylogenies of hematopoietic stem cells.