Phylodynamic Inference with Kernel ABC and Its Application to HIV Epidemiology

Area of Science:

• Computational biology
• Epidemiology
• Virology

Background:

• Phylogenetic tree shape reflects virus adaptation and host transmission dynamics.
• Phylodynamic inference estimates epidemiological parameters from viral phylogenies.
• Quantifying tree similarity is crucial but challenging for phylodynamic inference.

Purpose of the Study:

• To develop and validate a novel distance measure for phylodynamic inference.
• To integrate this measure into an Approximate Bayesian Computation (ABC) framework.
• To apply the new method to analyze viral evolution and outbreaks.

Main Methods:

• Utilized a subset tree kernel function from computational linguistics as a novel distance measure.
• Incorporated the kernel-based distance into an Approximate Bayesian Computation (ABC) framework (kernel-ABC).
• Validated kernel-ABC using simulated data and applied it to a real HIV outbreak.

Main Results:

• The kernel-based distance measure significantly improved tree shape classification.
• Kernel-ABC demonstrated higher accuracy in estimating virus transmission parameters compared to existing software.
• The method was successfully applied to study a recombinant HIV outbreak in China.

Conclusions:

• The kernel-ABC framework provides a powerful and versatile tool for phylodynamic inference.
• This approach bypasses the need for analytical model likelihoods, enabling broader model fitting.
• Kernel-ABC enhances our ability to understand viral evolution and transmission dynamics from sequence data.