Adaptive Preferential Sampling in Phylodynamics With an Application to SARS-CoV-2.

Area of Science:

• Population Genetics
• Epidemiology
• Computational Biology

Background:

• Longitudinal molecular data from evolving viruses offer insights into disease spread, complementing traditional surveillance.
• Coalescent theory models viral genealogy, assuming coalescent event rates are inversely proportional to effective population size (N(t)).
• Sampling bias, where sample collection depends on N(t), can lead to inaccurate N(t) estimations if not jointly modeled with coalescent processes.

Purpose of the Study:

• To develop a novel approach for jointly modeling coalescent and sampling processes to improve N(t) estimation.
• To address potential bias in N(t) estimation caused by time-varying sampling dependencies.
• To provide an efficient and flexible framework for analyzing viral population dynamics.

Main Methods:

• Modeled the sampling process as an inhomogeneous Poisson process with a rate dependent on N(t) and a time-varying coefficient.
• Employed Markov random field priors to make minimal assumptions about the functional forms of N(t) and the sampling coefficient.
• Developed efficient algorithms for Bayesian inference and assessed model performance through simulations and real-world data.

Main Results:

• The proposed model demonstrated improved performance compared to alternative methods in simulation studies.
• Application to SARS-CoV-2 sequences from California counties showed the model's utility in real-world epidemiological surveillance.
• The methodology is implemented in the R package 'adapref' for broader accessibility.

Conclusions:

• Jointly modeling coalescent and time-varying sampling processes provides more accurate estimates of effective population size (N(t)).
• This approach enhances the reliability of molecular data for understanding viral evolution and disease dynamics.
• The adapref R package facilitates the application of these advanced methods in public health and research.