Low-parameter phylogenetic inference under the general markov model.

Area of Science:

• Phylogenetics
• Computational Biology
• Evolutionary Biology

Background:

• Sumner et al. introduced squangles, Markov invariants for phylogenetic quartets, consistent with the general Markov (GM) model.
• The GM model is non-stationary, suggesting squangles may outperform standard methods with changing base composition.
• However, the GM model's assumption of constant rates across sites may limit squangle performance with rate variation.

Purpose of the Study:

• Implement squangles in a least-squares framework for weighted quartet estimation.
• Evaluate squangle robustness to violated rate-across-sites constancy using simulated and real data.
• Suggest modifications to enhance squangle performance, particularly with invariant sites.

Main Methods:

• Developed a least-squares implementation of squangles to generate weighted quartets.
• Assessed robustness by analyzing performance on simulated and empirical datasets with varying rates across sites.
• Proposed and tested a modification to improve squangle accuracy in the presence of invariant sites.

Main Results:

• Weighted quartets from squangles can be integrated into supertree and supernetwork analyses.
• Quantitatively demonstrated the robustness of squangles to deviations from the constant rates-across-sites assumption.
• Identified a modification that enhances squangle performance when invariant sites are present.

Conclusions:

• Squangles offer a novel phylogenetic estimation method complementary to existing approaches.
• The method performs well even when evolutionary rates vary across sites.
• The proposed modification further improves squangle utility in complex evolutionary scenarios.