ImOSM: intermittent evolution and robustness of phylogenetic methods.

Area of Science:

• Computational Biology
• Evolutionary Biology
• Bioinformatics

Background:

• Evaluating phylogenetic methods requires assessing robustness to violations of evolutionary models.
• Complete knowledge of evolutionary processes is often unavailable, necessitating robust inference techniques.
• Simulating specific model violations is crucial for rigorous testing of phylogenetic tools.

Purpose of the Study:

• Introduce ImOSM, a novel utility for embedding intermittent evolution as a model violation in sequence alignments.
• Assess the robustness of widely used phylogenetic inference methods (Maximum Likelihood, Maximum Parsimony, BIONJ) against simulated model violations.

Main Methods:

• Developed ImOSM to introduce intermittent evolution, characterized by random extra substitutions on tree branches.
• Simulated data with violations including rates across sites (RaS) heterogeneity and transition/transversion ratio changes.
• Tested Maximum Likelihood (ML), Maximum Parsimony (MP), and BIONJ phylogenetic methods on simulated datasets.

Main Results:

• Model violations, particularly RaS heterogeneity and ratio changes on external branches, impaired topological recovery for all tested methods on a four-taxon tree.
• On an eight-taxon tree, these violations led ML, MP, and BIONJ to infer distinct topologies, with ML and MP failing while BIONJ succeeded.
• Model homogeneity and goodness-of-fit tests demonstrated sufficient power to detect these simulated model violations.

Conclusions:

• Intermittent evolution poses a significant challenge to standard phylogenetic inference methods.
• The BIONJ method, utilizing ML-estimated parameters, showed greater robustness to the tested model violations.
• Model testing diagnostics are valuable for validating phylogenetic tree confidence and should be integrated into practical analyses.