Leaping through Tree Space: Continuous Phylogenetic Inference for Rooted and Unrooted Trees

Area of Science:

• Evolutionary Biology
• Computational Biology
• Genomics

Background:

• Phylogenetics is crucial for understanding life's history and disease spread.
• Inferring accurate phylogenetic trees from vast datasets remains a significant computational challenge.
• Current methods can be susceptible to local minima, limiting exploration of the complete tree space.

Purpose of the Study:

• To develop a novel method for phylogenetic tree exploration and inference in a continuous space.
• To overcome limitations of existing methods in navigating complex tree spaces and avoiding local optima.
• To enhance the accuracy and efficiency of phylogenetic inference, particularly for data-deficient scenarios.

Main Methods:

• Introduced a continuous relaxation of tree space enabling gradient computation.
• Implemented tree exploration and inference within this continuous space.
• Utilized automatic differentiation for optimization.

Main Results:

• The continuous space approach allows for significant leaps across tree space, improving exploration.
• Outperformed existing methods in unrooted tree inference and accurately inferred trees and roots in ultrametric simulations.
• Demonstrated effectiveness on empirical data with limited genetic information, successfully resolving major vertebrate lineages.

Conclusions:

• The novel continuous space method offers a powerful new approach for phylogenetic inference.
• It is particularly effective for challenging, data-deficient phylogenetic questions.
• This method represents a significant advancement for exploring evolutionary relationships.