TL;DR
Estimating a population's effective size (Ne) is crucial for understanding its evolutionary trajectory, but traditional methods often require multiple samples over time, which is impractical for many species. This paper introduces OneSAMP, a web-based tool that uses Approximate Bayesian Computation (ABC) to estimate Ne from a single genetic sample, making this vital metric more accessible.
What Problem Does This Paper Address?
Understanding the effective population size (Ne) is fundamental in population genetics. Ne represents the size of an idealized population that would experience the same amount of genetic drift as the actual population. It's a key indicator of a population's vulnerability to genetic drift, inbreeding, and its ability to adapt. However, accurately estimating Ne has been a persistent challenge. The gold standard for Ne estimation often involves collecting genetic samples from the same population at two different time points, separated by at least one generation. This "two-sample" approach, while robust, is often infeasible, especially for long-lived species or populations that are difficult to access repeatedly. Single-sample Ne estimators exist, but they have historically suffered from imprecision and bias, limiting their widespread adoption. This paper addresses the need for a reliable and accessible method to estimate Ne using only a single genetic sample.
Methods
The core of OneSAMP's methodology lies in Approximate Bayesian Computation (ABC). Instead of directly calculating likelihoods (which can be computationally intractable for complex population genetic models), ABC works by simulating data under a given model and comparing these simulated datasets to the observed data.
Here's a breakdown of the process:
- Summary Statistics: OneSAMP first calculates eight specific summary statistics from the user-provided microsatellite genotype data. These statistics are chosen because population genetics theory and simulations suggest they are informative about Ne. Examples include allele diversity measures, heterozygosity, and measures of genetic differentiation.
- Simulation: The program then simulates 50,000 hypothetical populations. For each simulation, an effective population size (Ne) is drawn from a user-defined prior distribution (a range representing the likely values of Ne). Other parameters, like the initial genetic variation (theta) and the number of generations simulated, are also drawn from uniform distributions within specified ranges.
- Wright-Fisher Model: Each simulated population evolves for a specified number of generations using the Wright-Fisher model, a standard model of genetic drift.
- Comparison and Acceptance: Summary statistics are calculated for each simulated population. Simulated populations whose summary statistics are "close" to those of the observed data are accepted. Closeness is defined by a Euclidean distance threshold.
- Weighted Local Regression: To refine the estimate, the accepted simulated Ne values are used in a weighted local regression. This method adjusts the simulated Ne values based on how well their summary statistics match the observed data, effectively inferring the posterior distribution of Ne.
- Prior Information: The user must provide a prior range for Ne, which helps guide the simulations and reflects existing knowledge about the population.
What's novel here is the application of ABC to the single-sample Ne estimation problem, leveraging microsatellite data and a suite of informative summary statistics.
Datasets
The paper uses a published dataset of ibex (Capra ibex) from the Belledone Mountains in the French Alps. This population was re-introduced in 1983 and sampled after several generations of growth. The dataset consists of microsatellite genotypes from 26 individuals across 19 loci. The authors excluded two loci that deviated significantly from Hardy-Weinberg proportions or were monomorphic. This dataset serves as a benchmark to demonstrate the utility of OneSAMP.
Key Insights and Ideas
- Single-Sample Ne is Achievable: The most significant insight is that it's possible to obtain reasonably reliable estimates of effective population size from a single genetic sample, overcoming a major limitation of previous methods.
- ABC as a Powerful Tool: The paper highlights the effectiveness of Approximate Bayesian Computation for tackling complex population genetics problems where traditional likelihood-based methods are difficult to implement.
- Importance of Summary Statistics: The careful selection of informative summary statistics is crucial for the success of the ABC approach.
- Prior Sensitivity: While the results are generally robust, the paper shows that the choice of prior distribution for Ne can influence the final estimate, underscoring the importance of informed prior selection.
Key Results
- Belledone Ibex Estimate: OneSAMP estimated the effective population size of the Belledone ibex population to be 19.06, with a 95% credible interval of 16.59–24.10. This estimate is consistent with the known history of the population.
- Robustness to Priors: The estimates remained consistent even when the prior range for Ne was varied (e.g., 2-400, 4-1000), suggesting the method is not overly sensitive to the exact prior specification, provided it's reasonably informative.
- Web-Based Accessibility: The program is available as a web-based tool, making it accessible to researchers without specialized computational expertise.
Limitations and Future Work
- Assumptions: OneSAMP assumes that all loci are unlinked and neutral. Violations of these assumptions could affect the accuracy of the estimates.
- Data Requirements: The method requires at least two polymorphic loci and assumes the input data is in GENEPOP format.
- Bias in Small Samples: While the paper focuses on a single-sample estimator, the authors acknowledge that one-sample estimators, in general, can still be imprecise and biased, especially with very small sample sizes or limited genetic variation.
- Future Work: The authors suggest that further testing with more extensive simulations and real-world datasets will continue to validate and refine the approach. The web-based nature of OneSAMP also facilitates ongoing development and user feedback.
Summary updated