BEST-FIT MAXIMUM-LIKELIHOOD MODELS FOR PHYLOGENETIC INFERENCE: EMPIRICAL TESTS WITH KNOWN PHYLOGENIES

Area of Science:

• Evolutionary Biology
• Bioinformatics
• Computational Phylogenetics

Background:

• Selecting appropriate DNA sequence evolution models is a significant challenge in phylogenetic reconstruction.
• Model choice is critical, particularly when substantial variation exists among phylogenetic branch lengths.

Purpose of the Study:

• To evaluate how nested models of DNA sequence evolution perform in reconstructing known phylogenies.
• To assess the impact of varying terminal branch lengths on model selection and phylogenetic accuracy.
• To identify optimal models for overcoming long-branch attraction under diverse evolutionary scenarios.

Main Methods:

• Utilized experimentally generated phylogenies of bacteriophage T7 with manipulated terminal branch lengths.
• Determined best-fit evolutionary models by progressively adding parameters to simpler models.
• Compared the phylogenetic performance of different models, including discrete maximum-likelihood and minimum-evolution-distance methods.

Main Results:

• Model selection was sometimes influenced by the sequence of parameter addition.
• Model performance showed minimal differences with branch length ratios of 1:3 or less.
• Extreme branch length variation revealed significant model differences, with best-fit models excelling at mitigating long-branch attraction.
• Correcting for among-site rate variation proved vital for overcoming long-branch attraction.

Conclusions:

• The choice of evolutionary model significantly impacts phylogenetic reconstruction accuracy, especially with high branch length variation.
• Best-fit models, particularly those accounting for among-site rate variation, are essential for robust phylogenetic inference and avoiding long-branch attraction.
• Findings are supported by simulation studies and preliminary analyses of mitochondrial and nuclear sequences.