Adaptive preferential sampling in phylodynamics.

Area of Science:

• Epidemiology
• Computational Biology
• Genetics

Background:

• Longitudinal molecular data from viruses offer insights into disease spread, complementing traditional case count surveillance.
• Coalescent theory models viral genealogy, assuming event rates are inversely proportional to effective population size ($N_{e}(t)$).
• Jointly modeling coalescent and sampling processes can improve $N_{e}(t)$ estimation, but misspecification leads to bias.

Approach:

• Introduces a novel approach modeling the sampling process as an inhomogeneous Poisson process.
• Utilizes Markov random field priors for minimal assumptions on the functional shapes of $N_{e}(t)$ and time-varying coefficients.
• Develops scalable algorithms for inference and evaluates model performance against alternatives.

Key Points:

• Accurate estimation of effective population size ($N_{e}(t)$) is crucial for understanding viral evolution and spread.
• The proposed method accounts for time-varying dependencies between sampling rates and genetic diversity.
• Demonstrates improved model performance in simulation studies.

Conclusions:

• The developed methodology provides a more robust framework for estimating effective population size ($N_{e}(t)$) in rapidly evolving pathogens.
• Applied to SARS-CoV-2 data, this approach enhances our ability to track viral dynamics.
• The methodology is implemented in the R package 'adapref' for broader accessibility.