Inferring Phenotypic Trait Evolution on Large Trees with Many Incomplete Measurements

• Activity Number: 376 - SBSS Paper Competition Winners (Part 2)
• Type: Topic-Contributed
• Date/Time: Thursday, August 12, 2021 : 12:00 PM to 1:50 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Gabriel Hassler* and Max Tolkoff and William L. Allen and Ho Lam Si Tung and Philippe Lemey and Marc Suchard
• Companies: UCLA and UCLA and University of Swansea and Dalhousie University and Rega Institute, KU Leuven and University of California Los Angeles
• Keywords: phylogenetics; missing data; Bayesian; multivariate Brownian diffusion; comparative methods; BEAST

Abstract

Evolutionary biologists often study covariation between multiple biological traits sampled across numerous related taxa. When studying these relationships, one must account for the shared evolutionary history of the taxa to avoid spurious inference. An additional challenge arises as obtaining complete data becomes increasingly difficult with increasing taxa. This generally requires data imputation or integration, and existing control techniques typically scale poorly as the number of taxa grows. We develop an inference regime that integrates out missing measurements analytically and scales linearly with the number of taxa using a post-order traversal algorithm under a multivariate Brownian diffusion model (MBD) of trait evolution. We also exploit this technique to extend the MBD model to account for sampling error or nonheritable variance. We find computational efficiency increases that top two orders of magnitude over current best practices in three real-world data sets. While we focus on the utility of this approach to phylogenetics, it generalizes to solve long-standing challenges to inference under matrix-normal and multivariate normal likelihoods with missing data at scale.