A simple polytomy resolver for dated phylogenies

Abstract

Summary1. Unresolved nodes in phylogenetic trees (polytomies) have long been recognized for their influences on specific phylogenetic metrics such as topological imbalance measures, diversification rate analysis and measures of phylogenetic diversity. However, no rigorously tested, biologically appropriate method has been proposed for overcoming the effects of this phylogenetic uncertainty.2. Here, we present a simple approach to polytomy resolution, using biologically relevant models of diversification. Using the powerful and highly customizable phylogenetic inference and analysis software beast and r, we present a semi‐automated ‘polytomy resolver’ capable of providing a distribution of tree topologies and branch lengths under specified biological models.3. Utilizing both simulated and empirical data sets, we explore the effects and characteristics of this approach on two widely used phylogenetic tree statistics, Pybus’ gamma (γ) and Colless’ normalized tree imbalance (Ic). Using simulated pure birth trees, we find no evidence of bias in either estimate using our resolver. Applying our approach to a recently published Cetacean phylogeny, we observed the expected small positive bias in γ and decrease in Ic.4. We further test the effect of polytomy resolution on diversification rate analysis using the Cetacean phylogeny. We demonstrate that using a birth–death model to resolve the Cetacean tree with 20%, 40% and 60% of random nodes collapsed to polytomies gave qualitatively similar patterns regarding the tempo and mode of diversification as the same analyses on the original, fully resolved phylogeny.5. Finally, we applied the birth–death polytomy resolution approach to a large (>5000 tips), but unresolved, supertree of extant mammals. We report a distribution of fully resolved model‐based trees, which should be useful for many future analysis of the mammalian supertree.