Model Selection and Parameter Inference in Phylogenetics Using Nested Sampling.

Area of Science:

• Computational Biology
• Evolutionary Biology
• Bayesian Statistics

Background:

• Bayesian inference in phylogenetics often requires significant computational resources for model selection and parameter estimation.
• Estimating marginal likelihood for model comparison and posterior distributions for parameter inference are computationally intensive challenges.
• Current methods for marginal likelihood estimation in phylogenetics can be impractical due to tuning parameter dependence and lack of uncertainty quantification.

Purpose of the Study:

• Introduce Nested Sampling (NS), a Bayesian computation algorithm, to the field of phylogenetics.
• Evaluate the performance of NS in phylogenetic inference scenarios.
• Compare NS with established methods for phylogenetic analysis.

Main Methods:

• Nested Sampling (NS) algorithm applied to phylogenetic inference.
• Analysis of NS performance across various evolutionary scenarios.
• Comparative study against existing phylogenetic inference techniques.

Main Results:

• Nested Sampling (NS) provides a unified approach to estimate marginal likelihoods and sample from posterior distributions simultaneously.
• NS demonstrates competitive performance in phylogenetic inference tasks.
• The algorithm offers a practical alternative to existing methods, addressing limitations in parameter tuning and uncertainty estimation.

Conclusions:

• Nested Sampling (NS) is a practical and computationally efficient algorithm for phylogenetic inference.
• NS is a competitive and attractive method for estimating marginal likelihoods and posterior distributions in phylogenetics.
• An open-source implementation of NS for BEAST 2 is available.