Title: Detecting and correcting bias in phylogenetic tree inference
Authors: Megan Owen - Lehman College CUNY (United States) [presenting]
Abstract: Phylogenetic tree inference is the problem of reconstructing the phylogenetic tree, which represents the evolutionary history of a set of organisms, from some genetic data, like DNA. Tree inference methods, like maximum likelihood and Bayesian MCMC, are known to exhibit bias, meaning they can produce trees with a tendency towards certain shapes or edge lengths. We propose a comprehensive method for detecting bias in tree shape and/or edge lengths using the Billera-Holmes-Vogtmann (BHV) tree space framework. The BHV tree space is a non-positively curved (or CAT(0)) geometric space containing all possible trees for a given set of leaves. The method employs a logarithm map to the tangent space, yielding a Euclidean space in which to perform analysis. We show that different tree inference methods have different biases and suggest methods for correcting them.