Comparison of models for nucleotide substitution used in maximum-likelihood phylogenetic estimation

Area of Science:

• Evolutionary biology
• Phylogenetics
• Computational biology

Background:

• Evolutionary models are crucial for inferring phylogenetic trees from sequence data.
• Assumptions within these models, such as base frequencies and rate variation, can impact accuracy.
• Evaluating model adequacy is essential for reliable evolutionary inference.

Purpose of the Study:

• To assess the impact of nucleotide substitution model assumptions on evolutionary tree estimation.
• To evaluate the robustness of maximum-likelihood tree topology estimates.
• To determine the effects of model complexity on branch length estimation and tree reliability.

Main Methods:

• Analysis of real sequence data.
• Evaluation of nucleotide substitution patterns (equilibrium base frequencies, transition/transversion ratio).
• Assessment of substitution rate variation across sites.

Main Results:

• Maximum-likelihood tree topology estimates are robust to violations of common model assumptions.
• Simpler, less realistic models can lead to misleading evaluations of tree reliability.
• Ignoring rate variation across sites causes more severe underestimation of branch lengths than ignoring substitution patterns.

Conclusions:

• While tree topology may be reliably estimated, branch lengths are sensitive to model assumptions, particularly rate variation.
• Using overly simplistic models for tree reliability assessment can be problematic.
• Accurate phylogenetic tree construction requires careful consideration of model realism and its impact on parameter estimation.