CHARACTERIZING THE INFLUENCE OF SAMPLING ASSUMPTIONS IN BAYESIAN PHYLOGENETICS AND PHYLODYNAMICS (Invited Presentation)

A range of models are available for inferring a phylogeny and the evolutionary rates which generated that phylogeny, and it is not obvious which models are preferable in which circumstances. In particular, Bayesian methods offer an increasing choice of constituent models for various evolutionary and sampling processes. Using two case studies, we demonstrate how model choice, particularly with respect to sampling assumptions, can influence the results of Bayesian phylogenetic and phylodynamic analyses. First, we infer the diversification trajectory of non-avian dinosaurs prior to their final demise at the Cretaceous-Paleogene boundary. We apply two different phylodynamic models to a selection of dinosaur supertrees, with uncertainty in branch lengths taken into account. The models produce different estimates of diversification rates in the latest Cretaceous, which we attribute to their different assumptions about sampling. Second, we demonstrate the utility of the recently-developed stratigraphic ranges model, which summarises multiple occurrences of each species into a range, explicitly representing information on the existence of a species through time. We use this model, newly implemented in the phylogenetics software BEAST2, to infer the phylogeny of penguins. Both molecular sequences for living penguins, and morphological characters for living and extinct penguins, are used. While the topology of the inferred phylogeny is largely consistent with previous analyses, changes in the posterior probabilities placed on clades, and in tip and node ages throughout the phylogeny, are evident. Overall, our results demonstrate the importance of sampling assumptions to the outcomes of phylogenetic and phylodynamic analyses.