Lateral transfer in Stochastic Dollo models

TL;DR

This paper introduces a Bayesian method to account for lateral transfer in phylogenetic models, improving the fit of evolutionary trait data, especially in linguistic studies, by addressing model misspecification caused by trait exchange.

Contribution

It develops a Bayesian approach for controlling lateral transfer effects in Stochastic Dollo models, handling intractable likelihoods through differential equations.

Findings

Improved model fit on Polynesian language data

Demonstrates the impact of lateral transfer on phylogenetic inference

Provides a new framework for modeling trait exchange in evolution

Abstract

Lateral transfer, a process whereby species exchange evolutionary traits through non-ancestral relationships, is a frequent source of model misspecification in phylogenetic inference. Lateral transfer obscures the phylogenetic signal in the data as the histories of affected traits are mosaics of the overall phylogeny. We control for the effect of lateral transfer in a Stochastic Dollo model and a Bayesian setting. Our likelihood is highly intractable as the parameters are the solution of a sequence of large systems of differential equations representing the expected evolution of traits along a tree. We illustrate our method on a data set of lexical traits in Eastern Polynesian languages and obtain an improved fit over the corresponding model without lateral transfer.