Improving the performance of Bayesian phylogenetic inference under relaxed clock models.

Area of Science:

• Computational Biology
• Evolutionary Biology
• Bioinformatics

Background:

• Bayesian Markov chain Monte Carlo (MCMC) is widely used for phylogenetic inference.
• Increasingly large molecular sequence datasets necessitate improved computational efficiency in phylogenetic algorithms.

Purpose of the Study:

• To develop a novel algorithm for enhancing the computational efficiency of Bayesian phylogenetic inference.
• To specifically address models incorporating per-branch rate parameters.

Main Methods:

• A new proposal kernel for MCMC algorithms is presented, simultaneously adjusting evolutionary rates and divergence times while preserving genetic distances.
• The method operates on internal node divergence times and adjacent branch rates, with three strategies (Simple Distance, Small Pulley, Big Pulley) for root manipulation.
• The Big Pulley strategy allows for topological changes, enabling exploration of all possible rooted trees consistent with the unrooted tree.

Main Results:

• Implementation in BEAST2 software demonstrated improved performance.
• The proposed operator yields better phylogenetic estimates for a given MCMC chain length.
• Reduced running time is achieved for a specified level of accuracy.

Conclusions:

• The new operator significantly enhances the efficiency of Bayesian phylogenetic inference, particularly for large datasets.
• Effective samples per hour are improved, indicating more efficient mixing compared to existing operators in BEAST2.
• Improvements can be as substantial as a half order of magnitude for large-scale analyses.