Using GWAS top hits to inform priors in Bayesian fine-mapping association studies

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• The choice of prior distribution significantly impacts Bayesian fine-mapping results, particularly for single-nucleotide polymorphism (SNP) effect sizes.
• The commonly used Normal distribution prior is often selected for computational ease rather than empirical suitability.
• Prior studies highlight the sensitivity of causal SNP Bayes factors to the chosen prior distribution.

Purpose of the Study:

• To develop a data-driven method for selecting appropriate priors in Bayesian fine-mapping.
• To compare the suitability of Laplace versus Normal priors using existing genome-wide association study (GWAS) data.
• To enhance the accuracy of variant prioritization in genetic studies.

Main Methods:

• Utilizing effect sizes from GWAS top hits and estimates of undiscovered causal SNPs.
• Applying a methodology to select between Laplace and Normal priors and estimate their parameters.
• Quantifying the uncertainty associated with prior parameter estimation.

Main Results:

• Breast cancer GWAS data provide strong evidence favoring the Laplace prior over the Normal prior.
• The proposed methodology allows for more objective prior selection compared to current practices.
• The choice of prior has significant implications for prioritizing genetic variants.

Conclusions:

• The Laplace prior is demonstrably more suitable than the Normal prior for Bayesian fine-mapping in certain contexts, such as breast cancer genetics.
• This research provides a framework for deriving data-informed priors, leading to more reliable variant prioritization.
• Adoption of these methods can refine the identification of causal variants in complex diseases.