Propagating uncertainty about molecular evolution models and prior distributions to phylogenetic trees

Area of Science:

• Evolutionary biology
• Computational phylogenetics
• Bioinformatics

Background:

• Phylogenetic tree construction relies on assumptions regarding substitution models, rate distributions, molecular clocks, and prior distributions.
• These assumptions influence results like clade probabilities and divergence time estimates, potentially leading to overconfidence.
• Existing methods like Bayesian model averaging may oversimplify uncertainty by favoring a single model.

Purpose of the Study:

• To demonstrate a method for propagating uncertainty in phylogenetic inference.
• To address the issue of overly confident conclusions due to unaddressed assumption uncertainty.
• To develop a more comprehensive approach than Bayesian model averaging for phylogenetic analysis.

Main Methods:

• A novel model combination method is proposed to integrate various assumptions and their prior distributions.
• The method accounts for uncertainty across a range of potential models and parameters.
• Nucleotide sequence data is used to illustrate the practical application of the model combination technique.

Main Results:

• The proposed model combination method incorporates a broader range of uncertainty compared to standard approaches.
• It provides a more realistic assessment of confidence in phylogenetic results.
• The method avoids the tendency of Bayesian model averaging to converge on a single, potentially inaccurate, model.

Conclusions:

• Propagating uncertainty in phylogenetic model assumptions is essential for robust evolutionary inference.
• The presented model combination method offers a more comprehensive way to handle assumption uncertainty.
• This approach leads to more reliable clade probabilities and divergence time estimates in molecular phylogenetics.