Multiple merger coalescent inference of effective population size

TL;DR

This paper introduces a Bayesian nonparametric approach to infer effective population size trajectories using multifurcating genealogies under the $ ext{Lambda}$-coalescent, improving modeling of complex ancestral processes.

Contribution

It develops a novel Bayesian nonparametric method for estimating population size and model parameters from multifurcating genealogies, extending beyond binary coalescent models.

Findings

Successfully applied to viral dynamics data

Analyzed Japanese sardine population changes

Demonstrated improved inference with multifurcating trees

Abstract

Variation in a sample of molecular sequence data informs about the past evolutionary history of the sample's population. Traditionally, Bayesian modeling coupled with the standard coalescent, is used to infer the sample's bifurcating genealogy and demographic and evolutionary parameters such as effective population size, and mutation rates. However, there are many situations where binary coalescent models do not accurately reflect the true underlying ancestral processes. Here, we propose a Bayesian nonparametric method for inferring effective population size trajectories from a multifurcating genealogy under the $\Lambda-$coalescent. In particular, we jointly estimate the effective population size and model parameters for the Beta-coalescent model, a special type of $\Lambda-$coalescent. Finally, we test our methods on simulations and apply them to study various viral dynamics as well as Japanese sardine population size changes over time. The code and vignettes can be found in the phylodyn package.