Compositionally Constrained Sites Drive Long-Branch Attraction.

Area of Science:

• Evolutionary biology
• Phylogenetics
• Bioinformatics

Background:

• Accurate evolutionary phylogenies are crucial but challenged by long branches and conflicting model estimates.
• Inadequate modeling of across-site compositional heterogeneity can cause erroneous phylogenetic topologies, particularly in deep evolutionary analyses.
• Existing models addressing compositional heterogeneity are often computationally intensive.

Purpose of the Study:

• To introduce and validate a new method, compositional constraint analysis, for investigating the impact of site-specific amino acid composition on phylogenetic inference.
• To develop and present the CAT-posterior mean site frequencies (CAT-PMSF) pipeline for diagnosing and resolving phylogenetic bias.
• To reassess animal phylogeny, specifically the placement of Porifera.

Main Methods:

• Compositional constraint analysis to assess site-specific amino acid diversity.
• Development and application of the CAT-posterior mean site frequencies (CAT-PMSF) pipeline based on the CAT model.
• Analysis of metazoan datasets to evaluate phylogenetic bias and tree topology.

Main Results:

• Compositionally constrained and unconstrained sites show conflicting signals under simplistic models, leading to long-branch attraction artifacts.
• The CAT-PMSF pipeline effectively diagnoses and resolves phylogenetic bias caused by inadequate modeling of compositional heterogeneity.
• CAT-PMSF demonstrates robustness against long-branch attraction across various alignments.

Conclusions:

• Compositional heterogeneity is a significant driver of long-branch attraction in phylogenomic datasets.
• The CAT-PMSF method provides a reliable approach to mitigate phylogenetic bias when CAT model convergence is uncertain.
• Analysis supports Porifera as the sister group to all other animals, refining animal phylogeny.