EmpPrior: using outside empirical data to inform branch-length priors for Bayesian phylogenetics

Area of Science:

• Phylogenetics
• Computational Biology
• Evolutionary Biology

Background:

• Branch-length parameters are crucial in phylogenetic models but default priors can be problematic.
• Existing Bayesian phylogenetic software may use unintentionally informative or uninformative priors for branch lengths.
• Biologists often do not leverage external data to inform branch-length priors for specific analyses.

Purpose of the Study:

• To develop a software package, EmpPrior, for incorporating relevant outside information into branch-length prior specification.
• To facilitate the use of external datasets for more accurate phylogenetic analyses.
• To improve the accuracy and precision of branch-length and tree-length estimates.

Main Methods:

• Developed EmpPrior software with two components: EmpPrior-search (Java) for querying TreeBASE and EmpPrior-fit (R) for parameterizing distributions.
• EmpPrior-search identifies similar datasets based on taxonomic and genetic sampling.
• EmpPrior-fit uses identified data to inform branch-length priors for focal analyses.

Main Results:

• EmpPrior efficiently queries TreeBASE to find relevant external data.
• The software successfully informs branch-length priors using outside information.
• An example analysis demonstrated improved tree-length estimates from a focal dataset using EmpPrior.

Conclusions:

• EmpPrior is a user-friendly and fast tool that enhances branch-length estimates.
• The software improves both the accuracy and precision of phylogenetic analyses.
• The EmpPrior strategy can be extended to other prior parameterization challenges in phylogenetics.