Inhomogeneous continuous-time Markov chains to infer flexible time-varying evolutionary rates

Area of Science:

• Evolutionary Biology
• Computational Biology
• Virology

Background:

• Estimating evolutionary rates from molecular data is crucial for understanding evolution and infectious diseases.
• Current methods may struggle to accurately capture temporal variations in evolutionary rates.

Purpose of the Study:

• Introduce a flexible Bayesian phylogenetic inference framework to model time-varying evolutionary rates.
• Develop a computationally efficient method for phylogenetic analysis with changing rates.

Main Methods:

• Modeled sequence evolution using inhomogeneous continuous-time Markov chains (ICTMCs).
• Introduced the 'polyepoch clock model' by parameterizing the rate function as piecewise constant.
• Employed Gaussian Markov random field priors and Hamiltonian Monte Carlo sampling for efficiency.

Main Results:

• The polyepoch clock model accurately recovers true timescales and evolutionary rates in simulations.
• Demonstrated computational efficiency through scalable gradient evaluation and Hamiltonian Monte Carlo sampling.

Conclusions:

• The polyepoch clock model offers a flexible and efficient approach for phylogenetic inference with time-varying rates.
• Applied the model to estimate evolutionary rates for West Nile virus, Dengue virus, influenza A/H3N2, and SARS-CoV-2 spread.