Estimating population parameters using the structured serial coalescent with Bayesian MCMC inference when some demes...

Area of Science:

• Population genetics
• Computational biology
• Epidemiology

Background:

• Estimating population size is crucial in evolutionary and epidemiological studies.
• Unsampled or
• ghost
• demes
• can complicate accurate demographic inference.
• Previous models often assume complete sampling, which is unrealistic.

Purpose of the Study:

• To investigate the impact of unsampled demes on population size estimation.
• To assess the feasibility of accurate inference when using incorrect models.
• To extend these findings to scenarios with partial sampling of ghost demes at the final time point.

Main Methods:

• Structured serial coalescent simulations.
• Bayesian Markov Chain Monte Carlo (MCMC) methods.
• Analysis of simulated population genetic data with varying sampling schemes.

Main Results:

• Accurate population size and demographic parameter inference is achievable with ghost demes if the correct model is employed.
• Using an incorrect model leads to biased and misleading estimates.
• Including sequences from ghost demes at the last sampling time significantly reduces estimation bias.
• Accurate estimation of ghost deme parameters becomes possible with this extended sampling strategy.
• Migration rate estimates are reliable when migration values are low, even with ghost demes.

Conclusions:

• The structured serial coalescent model provides robust inference even with unobserved population structure.
• Careful model selection is critical to avoid biased demographic estimates.
• Partial sampling of previously unsampled demes at later time points can salvage accurate inference in complex scenarios, relevant to fields like HIV evolution.