Estimating the probability of continuous character transitions
on phylogenetic trees

Inferring the evolutionary history of species traits is a basic problem in evolutionary biology and an issue of practical concern for ecologists, developmental biologists, and others testing biological hypotheses using phylogenetic data. Methods of analyzing categorical data in a phylogenetic context have addressed the problem of localizing character transitions to branches. Locating transitions in continuous characters have received less focus, in part because many methods are based on a constant variance (Brownian motion) model, in which the rate of character evolution is modeled as shifting gradually on the tree if at all. Evaluation of Ornstein-Uhlenbeck models that allow shifts in character equilibria or Brownian motion models on subtrees are straightforward methods of modeling transitions in continuous characters. Assessed in an information theoretic framework, these models are useful means of estimating the probability that continuous trait change on a selected phylogenetic branch violates the expectations of a constant variance model. In collaboration with Alexander Platt (University of Southern California), I am evaluating the relative effectiveness of existing methods for identifying shifts in continuous traits and comparing these methods with a Bayesian association mapping method developed by Platt and colleagues that directly estimates the probability and magnitude of quantitative trait change on all branches of a phylogeny. In combination, these techniques provide a useful complement to existing methods for investigating the evolution of continuous characters.

Project links