Bayesian Inference of Natural Selection from Allele Frequency Time Series.

Area of Science:

• Population genetics
• Paleogenomics
• Evolutionary biology

Background:

• Ancient DNA technology allows direct allele frequency ascertainment in ancestral populations.
• Allele frequency time series offer a more powerful approach to detect natural selection than methods using modern DNA.
• Detecting natural selection in non-equilibrium populations remains a challenge.

Purpose of the Study:

• To develop a Bayesian method for inferring diploid selection parameters and allele age using allele frequency time series.
• To introduce a novel path augmentation approach for analyzing allele frequency dynamics.
• To assess the method's power and potential biases in real-world data.

Main Methods:

• Developed a Bayesian statistical framework to analyze allele frequency time series data.
• Implemented a novel path augmentation approach using Markov chain Monte Carlo (MCMC) to integrate over allele frequency trajectories.
• Validated the method using simulations and applied it to horse coat color genetic data.

Main Results:

• The Bayesian method demonstrates good power for estimating selection coefficients and allele age.
• Simulations confirm the effectiveness of the path augmentation approach.
• Analysis of horse coat color data revealed significant biases when demographic history is ignored.

Conclusions:

• Directly observed allele frequency time series from ancient DNA provide a robust method for inferring evolutionary parameters.
• The developed Bayesian approach accurately estimates selection and allele age, even in complex demographic scenarios.
• Accounting for demographic history is crucial for reliable inference of natural selection from genetic data.