Invariant Versus Classical Quartet Inference When Evolution is Heterogeneous Across Sites and Lineages

Area of Science:

• Computational Biology
• Phylogenetics
• Molecular Evolution

Background:

• Classical phylogenetic methods often fail due to oversimplified models.
• Accurate inference of evolutionary relationships is crucial for understanding biodiversity and disease.

Purpose of the Study:

• To propose a novel quartet reconstruction method using phylogenetic invariants.
• To develop a method compatible with the most general Markov model of nucleotide substitution.
• To handle data from mixtures on the same topology.

Main Methods:

• Utilized phylogenetic invariants for quartet reconstruction.
• Developed a system of weights for quartet-based methods.
• Evaluated performance on real and simulated 4-taxon data under various conditions (time-homogeneous/nonhomogeneous, rate heterogeneity, branch lengths).
• Compared against Neighbor-Joining, Maximum Likelihood, and Maximum Parsimony.

Main Results:

• The proposed method demonstrates accuracy and robustness.
• It performs comparably to Maximum Likelihood when model assumptions align.
• It outperforms other methods when substitution models are inappropriate.
• It shows superior performance even when some assumptions are violated, given sufficient data length.

Conclusions:

• The proposed phylogenetic reconstruction method offers a robust and accurate alternative to classical approaches.
• It is particularly advantageous when dealing with complex evolutionary models and data heterogeneity.
• This method enhances the reliability of inferring evolutionary topologies.