Improving the coverage of credible sets in Bayesian genetic fine-mapping

Area of Science:

• Genetics
• Statistical Genetics
• Genomic Medicine

Background:

• Genome-Wide Association Studies (GWAS) identify disease-associated loci.
• Bayesian fine-mapping aims to pinpoint causal variants within these loci using credible sets.
• Current credible set coverage probabilities are often over-conservative due to biases in variant selection.

Purpose of the Study:

• To demonstrate the over-conservative nature of standard credible set coverage probabilities.
• To develop a method for re-estimating credible set coverage.
• To improve the resolution of genetic fine-mapping studies for identifying causal variants.

Main Methods:

• Utilized simulations to assess coverage probabilities in fine-mapping scenarios.
• Developed an 'adjusted coverage estimate' method using rapid simulations and SNP correlation structure.
• Extended this to create 'adjusted credible sets' by minimizing variant inclusion while meeting coverage targets.
• Applied the method to a type 1 diabetes fine-mapping study.

Main Results:

• Demonstrated that coverage probabilities are typically over-conservative in fine-mapping.
• The adjusted coverage estimate method was developed and validated.
• In a type 1 diabetes study, the method reduced the number of candidate variants in 27 out of 39 regions.
• The method requires only GWAS summary statistics and estimated SNP correlations.

Conclusions:

• The proposed method provides more accurate credible set coverage estimates.
• Adjusted credible sets enhance the efficiency of fine-mapping studies.
• This approach facilitates more cost-effective identification of causal variants, genes, and pathways implicated in human diseases.