Plesiomorphic character states cause systematic errors in molecular phylogenetic analyses: a simulation study.

Area of Science:

• Computational Biology
• Evolutionary Biology
• Bioinformatics

Background:

• Maximum likelihood (ML) methods are standard for phylogenetic inference.
• These methods often yield conflicting results.
• The role of character polarity in ML analyses is typically disregarded.

Purpose of the Study:

• To investigate sources of error in phylogenetic tree inference using ML.
• To analyze the impact of character polarity, specifically plesiomorphic states, on tree topology accuracy.
• To understand the mechanisms causing incorrect tree topologies.

Main Methods:

• Simulations of sequence evolution under various topologies.
• Inference of phylogenetic trees using a correct substitution model.
• Analysis of split-supporting site patterns, focusing on plesiomorphic states.
• Comparison of simulated data with known root sequences to identify plesiomorphic patterns.

Main Results:

• ML analyses can fail to recover correct tree topologies even with the correct model, particularly with branch-length heterogeneity.
• Plesiomorphic character states significantly contribute to errors in tree inference.
• Errors caused by plesiomorphic states on shorter branches are more misleading than chance similarities on longer branches.
• Removing sites with symplesiomorphies supporting false clades largely resolves systematic errors.

Conclusions:

• Character polarity is a relevant factor in ML phylogenetic analyses.
• The effects of plesiomorphic states have been neglected and can lead to widespread inaccuracies in published phylogenies.
• Future phylogenetic studies should account for character polarity to improve accuracy.