phyddle: software for exploring phylogenetic models with deep learning.

Area of Science:

• Computational Biology
• Evolutionary Biology
• Bioinformatics

Background:

• Phylogenetic models are crucial for understanding life's diversity and evolutionary history.
• Standard inference methods struggle with realistic models lacking tractable likelihood functions.
• Likelihood-free approaches offer a potential solution for complex phylogenetic modeling.

Purpose of the Study:

• Introduce phyddle, a novel software pipeline for phylogenetic modeling using likelihood-free deep learning.
• Enable integration of deep learning into evolutionary research workflows.
• Provide a flexible tool for analyzing phylogenetic trees.

Main Methods:

• Developed phyddle, a pipeline-based software utilizing deep learning for phylogenetic modeling.
• Implemented a five-step pipeline: Simulate, Format, Train, Estimate, and Plot.
• Compared deep learning-based inferences with traditional likelihood-based methods.

Main Results:

• Phyddle accurately performs phylogenetic inference tasks, including parameter estimation and model selection.
• The software successfully handles models lacking tractable likelihood functions.
• Benchmarks demonstrate comparable accuracy to likelihood-based methods for various evolutionary scenarios.

Conclusions:

• Phyddle offers a powerful and flexible alternative for phylogenetic modeling, especially for complex models.
• Deep learning approaches can effectively address limitations in traditional phylogenetic inference.
• Phyddle facilitates advanced evolutionary analyses and model exploration.