Markov-Modulated Continuous-Time Markov Chains to Identify Site- and Branch-Specific Evolutionary Variation in BEAST

Area of Science:

• Evolutionary biology
• Computational biology
• Genomics

Background:

• Phylogenetic inference relies on Markov models of character substitution.
• Standard models often assume homogeneous substitution rates over time and across sites.
• Site-specific rate variation over time (heterotachy) is crucial but frequently overlooked in molecular evolution.

Purpose of the Study:

• To introduce and implement Markov-modulated models (MMMs) to capture time-varying substitution processes at individual sites in phylogenetics.
• To integrate MMMs into the BEAST software package for flexible phylogenetic analysis.
• To evaluate the impact of MMMs on phylogenetic tree estimation and model fit.

Main Methods:

• Developed a general Markov-modulated model (MMM) framework within BEAST.
• Extended covarion-like models to allow site-specific substitution processes to vary across lineages.
• Utilized the BEAGLE library for efficient computation of phylogenetic inference with MMMs.
• Applied MMMs to bacterial, viral, and plastid genome evolution data.

Main Results:

• MMMs significantly impact phylogenetic tree estimation.
• MMMs provide a substantially improved model fit compared to standard substitution models.
• Marginal likelihood estimation accurately identifies the true generative model, without favoring overly complex MMMs.
• The implementation leverages BEAGLE for computational efficiency.

Conclusions:

• Markov-modulated models are essential for accurately modeling molecular evolution by accounting for temporal heterogeneity in substitution rates.
• The implemented MMM framework in BEAST offers a flexible and powerful tool for phylogenetic inference.
• MMMs enhance phylogenetic accuracy and model fit, particularly for complex evolutionary scenarios.