Evaluating the Adequacy of Molecular Clock Models Using Posterior Predictive Simulations.

Area of Science:

• Evolutionary biology
• Phylogenetics
• Computational biology

Background:

• Molecular clock models are essential for estimating evolutionary rates and timescales from nucleotide sequences.
• Current model selection relies on statistical fit, which can be unreliable if all models are inadequate.
• Assessing absolute model performance is critical for robust phylogenetic inference.

Purpose of the Study:

• To introduce and evaluate a novel method for assessing the absolute performance of molecular clock models.
• To determine if posterior predictive simulations can reliably detect inadequate clock models.
• To compare the proposed method with existing statistics for model adequacy testing.

Main Methods:

• Utilized posterior predictive simulations to assess molecular clock model adequacy.
• Tested the method with simulated data to evaluate sensitivity to branch length estimation bias.
• Compared the power of the multinomial test statistic against the proposed simulation method.

Main Results:

• The posterior predictive simulation method is sensitive to bias in branch length estimates caused by underparameterized clock models.
• The multinomial test statistic demonstrated low power in assessing molecular clock model adequacy.
• Empirical data from coronaviruses, simian immunodeficiency virus, killer whales, and marine turtles were used for illustration.

Conclusions:

• Posterior predictive simulations offer a valuable approach for evaluating molecular clock model adequacy.
• Model adequacy assessment should be routinely combined with traditional model selection in phylogenetic studies.
• This integrated approach can guide the selection of a broader range of appropriate clock models for phylogenetic analysis.