Title: Markov Genealogy Processes: Exact Phylodynamic Likelihood via Structured Filtering
Speaker: Qianying Lin, Assistant Professor, Division of Biostatistics, College of Public Health, The Ohio State University
Abstract: Genealogies inferred from pathogen genomes contain information about transmission and within-host progression, but likelihood-based inference has largely been limited to special cases such as coalescent and linear birth–death models. In this talk, I present a general construction showing how a continuous-time Markov population process induces a genealogy-valued stochastic process, a Markov genealogy process, and how this yields exact expressions for the likelihood of an observed genealogy. The key result is a model-determined nonlinear filtering equation that propagates likelihood forward in time and supports statistically efficient Monte Carlo inference. This framework generalizes existing likelihood formulas as special cases and removes the need to reverse transition densities, enabling exact phylodynamic likelihood computation for a broad class of epidemiological compartmental models.