Finding high posterior density phylogenies by systematically extending a directed acyclic graph.

Area of Science:

• Computational Biology
• Phylogenetics
• Statistical Modeling

Background:

• Bayesian phylogenetics commonly uses Markov chain Monte Carlo (MCMC) for posterior distribution estimation.
• MCMC methods explore tree space via random walks, which can be inefficient for large datasets.

Purpose of the Study:

• To develop novel methods for efficiently searching the posterior distribution of phylogenetic trees.
• To overcome limitations of random walk methods in exploring vast tree spaces.

Main Methods:

• Utilized subsplit directed acyclic graphs (sDAGs) to represent multiple trees simultaneously.
• Developed and evaluated two methods for introducing, ranking, and selecting local rearrangements on sDAGs.
• Compared sDAG-based approaches with traditional MCMC random walk methods.

Main Results:

• One proposed sDAG method successfully identified high posterior density trees across various datasets.
• A simpler strategy of aggregating trees into an sDAG proved computationally faster.
• The sDAG aggregation strategy yielded a higher fraction of probable trees compared to other methods.

Conclusions:

• Subsplit directed acyclic graphs offer a more efficient framework for Bayesian phylogenetic inference.
• sDAG aggregation presents a computationally advantageous and effective approach for exploring tree space.
• This work advances methods for identifying probable trees in complex phylogenetic analyses.