Comparative evaluation of a new effective population size estimator based on approximate bayesian computation

Area of Science:

• Population genetics
• Evolutionary biology
• Conservation biology

Background:

• Effective population size (N(e)) is crucial for understanding evolutionary dynamics and informing conservation strategies.
• Accurate estimation of N(e) is challenging, with existing methods having limitations in bias and performance.
• There is a need for robust N(e) estimators that perform well across various population and sampling scenarios.

Purpose of the Study:

• To introduce and evaluate a novel estimator for effective population size (N(e)) called "SummStat."
• To assess the performance of the SummStat estimator against existing methods using simulations.
• To highlight the flexibility and strengths of the SummStat estimator in inferring N(e).

Main Methods:

• Developed the "SummStat" estimator utilizing summary statistics within an approximate Bayesian computation (ABC) framework.
• Conducted simulations of a Wright-Fisher population model with varying parameters (sample size, loci, generations, N(e)).
• Compared SummStat performance against two likelihood-based and one moment-based N(e) estimator, analyzing bias and RMSE.

Main Results:

• The SummStat estimator demonstrated the lowest bias in 32 out of 36 tested parameter combinations.
• Relative Mean Square Error (RMSE) for SummStat was generally intermediate, but significantly reduced with 3+ generations between samples for small N(e).
• SummStat confidence intervals were more conservative and more likely to contain the true N(e) compared to likelihood-based methods.

Conclusions:

• The SummStat estimator provides a less biased and more reliable method for estimating effective population size (N(e)).
• Its flexible structure allows integration of diverse population genetic summary statistics.
• SummStat shows particular promise for accurate N(e) inference in scenarios with longer time intervals between samples.