adaPop: Bayesian inference of dependent population dynamics in coalescent models.

Area of Science:

• Evolutionary biology
• Statistical genetics
• Computational biology

Background:

• The coalescent framework analyzes population dynamics using ancestral relationships from sequence data.
• Understanding dependence between distinct populations is crucial for biomedical applications like infectious disease, cell development, and tumorgenesis research.
• Advances in sequencing technology enable detailed analysis of complex population histories.

Purpose of the Study:

• To present adaPop, a probabilistic model for estimating past population dynamics of dependent populations.
• To quantify the degree of dependence between evolving populations.
• To track time-varying associations between populations with minimal assumptions.

Main Methods:

• Developed a probabilistic model (adaPop) using Markov random field priors.
• Incorporated nonparametric estimators and extensions for integrating multiple data sources.
• Designed fast and scalable inference algorithms for computational efficiency.

Main Results:

• adaPop successfully estimates past population dynamics and quantifies inter-population dependence.
• The model effectively tracks time-varying associations between populations.
• Demonstrated utility in analyzing evolutionary histories of SARS-CoV-2 variants using simulated and real-world data.

Conclusions:

• adaPop provides a robust framework for inferring the evolutionary dynamics of dependent populations.
• The model's flexibility and scalability make it applicable to diverse biological datasets.
• This approach enhances our understanding of complex evolutionary processes in various biomedical contexts.