Parameterizing the genetic architecture under stabilizing selection

Area of Science:

• Statistical Genetics
• Evolutionary Biology
• Quantitative Genetics

Background:

• Genetic variants with large effects are often at low frequencies, suggesting stabilizing selection.
• The phenomenological alpha-model describes this but lacks a mechanistic basis.
• A direct population-genetic interpretation for the alpha-model is missing.

Purpose of the Study:

• To derive a new model for the frequency dependence of genetic effect sizes based on evolutionary theory.
• To provide a mechanistic and population-genetic interpretation for observed relationships between effect size and allele frequency.
• To integrate evolutionary principles with standard mixed-model methodology for genetic prediction.

Main Methods:

• Developed a linear mixed model derived from evolutionary theory.
• Incorporated interpretable evolutionary quantities: mutational variance, selection intensity, and trait coupling.
• Utilized restricted maximum likelihood (REML) to estimate two identifiable variance components.
• Linked fitness-landscape models with mixed-model methodology for inference and prediction.

Main Results:

• The new model naturally yields frequency dependence of effect sizes.
• The model allows estimation of key evolutionary parameters.
• Forward simulations demonstrated accurate recovery of trait variance.
• The proposed model generally improved genetic prediction accuracy compared to the alpha-model.

Conclusions:

• The derived model offers a mechanistic alternative to the alpha-model.
• It provides a population-genetic interpretation for effect size distributions.
• The framework enhances genetic prediction by integrating evolutionary insights.