Biases in Maximum Likelihood and Parsimony: A Simulation Approach to a 10-Taxon Case

Area of Science:

• Phylogenetics
• Computational Biology
• Evolutionary Biology

Background:

• Phylogenetic inference aims to reconstruct evolutionary history.
• Maximum likelihood and parsimony are common methods for inferring phylogenies.
• Both methods can be susceptible to biases, particularly with complex evolutionary scenarios.

Purpose of the Study:

• To investigate biases in maximum likelihood and parsimony phylogenetic methods.
• To evaluate method performance under varying conditions of branch length disparity and data size.
• To examine the impact of taxonomic sampling on phylogenetic inference.

Main Methods:

• A simulation study was conducted using a 10-taxon case.
• Simulations involved topologies with coexisting long and short branches.
• Branch lengths, data size, and taxonomic sampling schemes were systematically varied.

Main Results:

• Parsimony methods confirmed to exhibit significant long-branch attraction bias.
• Maximum likelihood performance was sensitive to extreme branch length disparities.
• Likelihood method retrieved topologies consistent with both long-branch attraction and repulsion.

Conclusions:

• Parsimony is strongly affected by long-branch attraction.
• Maximum likelihood is also vulnerable to branch length heterogeneity, potentially leading to inaccurate tree reconstruction.
• Careful consideration of method biases is crucial for accurate phylogenetic inference, especially with challenging datasets.