Inferring phylogenetic birth-death models from extant lineages through time
Birth-death processes have been used to study population growth, with broad-ranging biological applications such as identifying speciation and extinction rates, calibrating divergence times, and studying the dynamics of pathogens in infection trees. Recent theoretical work on phylogenetic birth-death models offer differing viewpoints on whether they can be estimated from lineages through time. Recently, Louca and Pennell (2020) demonstrated that time-varying birth and death rates are not identifiable from lineage-through-time data. This was a grave result, in view of thousands of published biological and computational studies that use this data. In this talk, I explain how identifiability can be restored, while re-focusing the discussion to what actually makes inference computationally challenging. This is based on joint work with Jonathan Terhorst (University of Michigan, Ann Arbor).