GTRpmix: A Linked General Time-Reversible Model for Profile Mixture Models.

Area of Science:

• Computational Biology
• Molecular Evolution
• Phylogenetics

Background:

• Profile mixture models account for site-specific biochemical constraints in protein evolution.
• Existing phylogenetic methods often use a single empirical exchangeability matrix (e.g., LG), which may not be optimal for mixture models.

Purpose of the Study:

• To develop and evaluate a new model (GTRpmix) for estimating a common exchangeability matrix within profile mixture models.
• To introduce novel exchangeability matrices (ELM, EAL) optimized for specific phylogenetic analyses.

Main Methods:

• Maximum likelihood estimation of a common exchangeability matrix under profile mixture models.
• Development of the GTRpmix model and associated exchangeability matrices (ELM, EAL).
• Evaluation of model fit and topological accuracy using empirical datasets.

Main Results:

• Exchangeability matrices estimated under profile mixture models differ significantly from the LG matrix.
• The GTRpmix model and the new ELM/EAL matrices show improved model fit and topological accuracy.
• IQ-TREE2 (v2.3.1+) supports the estimation of linked exchangeabilities under profile mixture models.

Conclusions:

• Profile mixture models with optimized exchangeability matrices provide a more accurate representation of protein evolution.
• The ELM and EAL matrices offer improved phylogenetic inference for eukaryotes and eukaryotes/archaea, respectively.
• The GTRpmix framework facilitates more robust phylogenetic analyses by accounting for complex evolutionary processes.