Going back to the roots: Evaluating Bayesian phylogeographic models with discrete trait uncertainty

Area of Science:

• Evolutionary biology
• Epidemiology
• Computational biology

Background:

• Phylogeography analyzes virus sequences with trait data for public health surveillance.
• Identifying the geographic origin of outbreaks (root state) is critical for intervention.

Purpose of the Study:

• Investigate how discrete state space and sequence dataset size influence root state classification accuracy in phylogeographic models.
• Evaluate the utility of Kullback-Leibler (KL) divergence as a performance metric.

Main Methods:

• Performed phylogeographic inference on simulated DNA datasets.
• Varied the number of sequences and the number of discrete trait values.
• Used logistic regression to model root state classification accuracy and KL divergence.

Main Results:

• Phylogeographic models showed optimal performance at intermediate sequence dataset sizes.
• KL divergence increased with both discrete state space and dataset size.
• KL divergence was not supported as a predictor of model accuracy.

Conclusions:

• Relying solely on KL divergence may inflate support for models with finer discretization and larger datasets.
• Results are important for public health practitioners using phylogeographic models for infectious disease surveillance.