Sampling phylogenetic tree space with the generalized Gibbs sampler.

Area of Science:

• Computational Biology
• Evolutionary Biology
• Statistical Modeling

Background:

• A recent critique questioned the originality of an algorithm for phylogenetic inference presented in a 2005 paper.
• The criticism focused on the algorithm for economizing parsimony calculations on trees undergoing subtree pruning and regrafting (SPR) rearrangements.

Discussion:

• The author acknowledges the criticism regarding the algorithm's originality but clarifies the paper's primary focus was not the algorithm itself, nor parsimony.
• The core contribution is a novel Markov chain Monte Carlo (MCMC) sampling strategy within a tree-space state.
• This strategy employs a Gibbs-like approach, drawing from conditional distributions over tree sets, a method previously unused for sampling tree-space.

Key Insights:

• The paper introduces a novel Gibbs-like MCMC strategy for sampling phylogenetic tree-space.
• This approach offers a potentially advantageous alternative to commonly used Metropolis-like strategies in phylogenetic inference.
• The author apologizes for incorrectly claiming originality for the SPR-based algorithm but asserts the paper's continued value.

Outlook:

• The author advocates for the integration of this Gibbs-like MCMC technique into phylogenetic MCMC samplers.
• Potential applications include sampling phylogenies for biological invasions and agent-based Bayesian ecological modeling.
• This novel sampling strategy holds promise for advancing phylogenetic and ecological modeling research.