Improvement of distance-based phylogenetic methods by a local maximum likelihood approach using triplets

Area of Science:

• Phylogenetics
• Computational Biology
• Evolutionary Genetics

Background:

• Estimating evolutionary distances is crucial for reconstructing phylogenetic trees.
• Existing methods can struggle with accuracy, especially for long-distance estimations.
• Distance-based phylogenetic methods rely on accurate distance matrices.

Purpose of the Study:

• To introduce a new method, TripleML, for more precise evolutionary distance estimation.
• To improve the topological accuracy of distance-based phylogenetic inference.
• To assess the performance of TripleML in conjunction with common phylogenetic algorithms.

Main Methods:

• Developed TripleML, a method utilizing a three-leaf tree for distance estimation.
• Branch lengths are optimized via maximum likelihood.
• Integrated TripleML with Neighbor-Joining-like (NJ-like) algorithms for distance matrix computation and refinement during tree building.

Main Results:

• TripleML significantly enhances the topological accuracy of phylogenetic trees generated by NJ, BioNJ, and Weighbor.
• The method maintains reasonable computation times, outperforming existing tools.
• Simulations showed an 11% reduction in wrongly inferred branches when combining NJ with TripleML.

Conclusions:

• TripleML offers a substantial improvement in evolutionary distance estimation accuracy.
• The method enhances the performance of popular distance-based phylogenetic algorithms.
• TripleML provides a computationally efficient approach for accurate phylogenetic inference.