New Statistical Criteria Detect Phylogenetic Bias Caused by Compositional Heterogeneity

Area of Science:

• Evolutionary biology
• Phylogenetics
• Genomics

Background:

• Statistical phylogenetic analyses of DNA sequences commonly assume stationary base composition through time and across lineages.
• Violations of this assumption are frequent, but their impact on phylogenetic inference remains unclear.

Purpose of the Study:

• To investigate the impact of compositional heterogeneity on phylogenetic estimates.
• To develop and evaluate methods for assessing model adequacy in phylogenetic analyses.

Main Methods:

• Employed a method for assessing model adequacy, including a detailed simulation study.
• Proposed new diagnostic criteria and guidelines for their application.
• Applied these criteria to genome-scale data from 40 bird species.

Main Results:

• Common frequentist criteria for model adequacy are highly conservative, often rejecting models without clear phylogenetic bias.
• Loci with severely non-homogeneous base composition were found to be uncommon in the studied bird genome-scale data.
• The proposed new criteria offer improved assessment of model adequacy.

Conclusions:

• Compositional heterogeneity, while a violation of common assumptions, may not always mislead phylogenetic inference.
• Well-informed diagnostic statistics are crucial for testing model adequacy in phylogenomic analyses.
• The developed criteria provide a more reliable approach to evaluating evolutionary models.