Testing adequacy for DNA substitution models.

Area of Science:

• Phylogenetics
• Computational Biology
• Evolutionary Biology

Background:

• Model adequacy testing is crucial for accurate phylogenetic inference.
• Mis-specified DNA substitution models can lead to inconsistent results, particularly with maximum likelihood methods.
• Current model adequacy testing methods in phylogenetics are underdeveloped.

Purpose of the Study:

• To develop a simple, general, powerful, and robust model adequacy test for phylogenetic inference.
• To address the limitations of existing methods for assessing DNA substitution model fit.
• To improve the reliability of phylogenetic analyses by ensuring appropriate model selection.

Main Methods:

• Development of a novel goodness-of-fit test based on Pearson's test.
• Utilizing binning of site patterns for test construction.
• Validation through extensive simulations and real data analysis.

Main Results:

• The proposed test demonstrates high power in rejecting inadequate models across various binning strategies.
• Type I error rates are well-controlled, ensuring reliable test performance.
• The test successfully rejected complex models (e.g., GTR+I+Γ) that simpler models failed to reject, and outperformed the Goldman-Cox test in some cases.

Conclusions:

• Model adequacy testing, alongside bootstrapping, is essential for validating phylogenetic results.
• The new goodness-of-fit test provides a simple and powerful tool for regular model checking.
• Researchers should be cautious about conclusions drawn from analyses using inadequate models; results can now be flagged using this new test.