An efficient Bayesian inference framework for coalescent-based nonparametric phylodynamics.

Area of Science:

• Computational Biology
• Epidemiology
• Evolutionary Biology

Background:

• Phylodynamics reconstructs population size dynamics from genetic samples.
• This is crucial for tracking infectious disease spread, like influenza.
• Current methods struggle with large, real-time surveillance datasets.

Purpose of the Study:

• To develop a computationally efficient Bayesian inference framework for phylodynamics.
• To address the limitations of existing methods when analyzing large datasets.
• To improve the speed and accuracy of population size dynamics estimation.

Main Methods:

• Developed a Bayesian inference framework using Hamiltonian Monte Carlo (HMC) for coalescent models.
• Implemented a strategy of splitting the Hamiltonian function to further enhance computational efficiency.
• Applied the method to simulated and real-world infectious disease datasets.

Main Results:

• The HMC-based framework provides accurate estimates of population size dynamics.
• The proposed method is substantially faster than alternative Bayesian inference techniques (e.g., elliptical slice sampler, Metropolis-adjusted Langevin algorithm).
• The framework demonstrates improved efficiency, particularly for large datasets.

Conclusions:

• The new computationally efficient Bayesian phylodynamics framework enables faster and accurate inference of population dynamics.
• This method is well-suited for analyzing large-scale genetic data from infectious disease surveillance.
• The R package 'phylodyn' and associated code are publicly available for use.