Model choice for phylogeographic inference using a large set of models

Area of Science:

• Evolutionary biology
• Population genetics
• Computational biology

Background:

• Model-based phylogeographic analyses are crucial for understanding population dynamics like gene flow and expansion.
• Approximate Bayesian Computation (ABC) is a flexible model-based method for inferring these processes.

Purpose of the Study:

• To investigate methods for objectively identifying suitable models in Approximate Bayesian Computation (ABC) for phylogeographic inference.
• To compare subjective versus objective model selection strategies using empirical data.

Main Methods:

• Conducted an ABC analysis on Plethodon idahoensis data using an initial set of five models.
• Developed an objective method by simulating prior distributions for 143 models and identifying a subset exceeding a chance probability threshold.
• Calculated relative posterior probabilities for the selected models.

Main Results:

• The initial subjective analysis favored a simple population isolation model (PP=0.70).
• The objective model selection identified a more complex model (divergence with population expansion and gene flow) as the best fit for the P. idahoensis data.
• This highlights a contrast between subjective and objective model selection outcomes.

Conclusions:

• The objective determination of models for inclusion in ABC analyses is critical for robust phylogeographic inference.
• The choice of models significantly impacts the inferred evolutionary processes.
• This study provides a framework for improving the reliability of model-based phylogeographic studies.