A Surrogate Function for One-Dimensional Phylogenetic Likelihoods.

Area of Science:

• Computational Biology
• Evolutionary Biology
• Bioinformatics

Background:

• Phylogenetic analyses face increasing computational demands due to larger datasets and complex substitution models.
• Approximating likelihood functions is crucial for efficient branch length optimization and Bayesian inference.

Purpose of the Study:

• To develop an efficient approximation for 1D phylogenetic likelihood functions parameterized by single branch lengths.
• To create a versatile surrogate function applicable to diverse phylogenetic models and tree structures.

Main Methods:

• Developed a four-parameter surrogate function inspired by the binary symmetric model.
• Validated the surrogate's ability to approximate likelihoods across various branch lengths.
• Integrated the surrogate as a proposal mechanism within Bayesian sampling frameworks.

Main Results:

• The four-parameter surrogate effectively approximates 1D likelihood functions.
• The method demonstrates applicability across a wide range of phylogenetic models and tree topologies.
• The surrogate functions as an efficient proposal mechanism for Bayesian phylogenetic analyses.

Conclusions:

• The developed surrogate function offers a computationally efficient alternative for approximating phylogenetic likelihoods.
• This approximation method is valuable for optimizing branch lengths and enhancing Bayesian sampling in phylogenetics.
• The open-source C library implementation facilitates its integration into existing phylogenetic software, improving performance.