Context-dependent codon partition models provide significant increases in model fit in atpB and rbcL protein-coding...

Area of Science:

• Molecular Evolution
• Bioinformatics
• Computational Biology

Background:

• Accurate modeling of protein-coding sequence evolution is computationally intensive with standard codon models.
• Codon partition models offer a computationally efficient alternative but assume independent evolution of codon positions.
• Empirical evidence suggests context-dependent substitution patterns at four-fold degenerate sites, challenging the independence assumption.

Purpose of the Study:

• To introduce and evaluate context-dependent codon partition models for sequence evolution.
• To test the hypothesis that four-fold degenerate site evolution depends on flanking base composition.
• To compare the performance of context-dependent models against existing evolutionary models.

Main Methods:

• Developed context-dependent codon partition models incorporating flanking base effects at four-fold degenerate sites.
• Applied and compared independent models, codon models, codon partition models, and context-dependent codon partition models.
• Utilized atpB and rbcL gene datasets from land plants for model evaluation.
• Employed Bayes factors and thermodynamic integration for model fit assessment.

Main Results:

• Context-dependent codon partition models significantly improved model fit for both atpB and rbcL gene datasets.
• A specific context-dependent model provided the largest increase in model fit compared to independent models.
• These context-dependent models approach the performance of computationally expensive codon models.

Conclusions:

• Context-dependent codon partition models are computationally viable and accurate alternatives to traditional codon models.
• The evolutionary patterns in atpB and rbcL genes are distinct, suggesting gene-specific modeling is advisable.
• Incorporating context-dependency enhances the biological realism of sequence evolution models.