Sampling through time and phylodynamic inference with coalescent and birth-death models

Area of Science:

• Population Genetics
• Evolutionary Biology
• Computational Biology

Background:

• Population genetic models are crucial for inferring population size and growth rates from genetic data.
• The coalescent and birth-death-sampling models (BDM) are two prominent approaches for such inferences.
• Exponential population growth is often modeled using birth-death branching processes.

Purpose of the Study:

• To compare the performance of coalescent and BDM in estimating population size and birth rates under exponential growth.
• To investigate the impact of sampling strategies on the accuracy of these models.
• To develop improved coalescent estimators that incorporate sampling time information.

Main Methods:

• Comparative analysis of coalescent and BDM using simulated data from exponentially growing populations.
• Evaluation of model performance with genetic data sampled at single and multiple time points.
• Development of a novel coalescent estimator incorporating known sampling processes.

Main Results:

• Coalescent and BDM yield similar estimates for growth rates and sampled population fractions when data is sampled at a single time point, even in small populations.
• BDM estimators exhibit significant bias when the sampling process is misspecified for data sampled across multiple time points.
• The timing of samples significantly contributes to the statistical power of BDMs, independent of the genealogical tree structure.

Conclusions:

• While effective for single-time samples, BDMs require careful specification of sampling processes for multi-time samples to avoid bias.
• Incorporating sampling time information into coalescent models can enhance precision and potentially outperform standard coalescent approaches.
• The findings motivate the development of new coalescent methods that leverage known sampling histories for more accurate population genetic inferences.