Empar: EM-based algorithm for parameter estimation of Markov models on trees

TL;DR

Empar introduces an EM-based algorithm for estimating parameters and branch lengths in nonhomogeneous Markov models on trees, improving phylogenetic inference accuracy for complex evolutionary data.

Contribution

This work presents the first open-source tool capable of handling nonhomogeneous models, extending phylogenetic analysis beyond traditional homogeneous assumptions.

Findings

High accuracy in parameter estimation and branch length recovery.

Effective performance across various simulated data settings.

First tool to handle nonhomogeneous evolutionary models.

Abstract

The goal of branch length estimation in phylogenetic inference is to estimate the divergence time between a set of sequences based on compositional differences between them. A number of software is currently available facilitating branch lengths estimation for homogeneous and stationary evolutionary models. Homogeneity of the evolutionary process imposes fixed rates of evolution throughout the tree. In complex data problems this assumption is likely to put the results of the analyses in question. In this work we propose an algorithm for parameter and branch lengths inference in the discrete-time Markov processes on trees. This broad class of nonhomogeneous models comprises the general Markov model and all its submodels, including both stationary and nonstationary models. Here, we adapted the well-known Expectation-Maximization algorithm and present a detailed performance study of this approach for a selection of nonhomogeneous evolutionary models. We conducted an extensive performance assessment on multiple sequence alignments simulated under a variety of settings. We demonstrated high accuracy of the tool in parameter estimation and branch lengths recovery, proving the method to be a valuable tool for phylogenetic inference in real life problems. $\empar$ is an open-source C++ implementation of the methods introduced in this paper and is the first tool designed to handle nonhomogeneous data.