Bayesian variable selection using partially observed categorical prior information in fine-mapping association...

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Bayesian fine-mapping methods integrate functional genomic data to improve association studies.
• Existing methods struggle with partially observed functional genomic information.

Purpose of the Study:

• To develop a novel Bayesian approach for fine-mapping case-control association studies that accommodates partially observed functional genomic data.
• To improve the localization of causal single-nucleotide polymorphisms (SNPs) by leveraging functional significance (FS) scores.

Main Methods:

• Utilized functional significance (FS) scores derived from multiple bioinformatics sources to inform effect size prior distributions.
• Employed finite mixtures of elicited priors to handle missing FS scores by partitioning SNPs into FS score groups.
• Applied prior scale mixtures of normals for differential shrinkage of effect sizes and marginal posterior probability intervals for SNP selection.

Main Results:

• The proposed method demonstrated improved localization of causal SNPs compared to existing multi-SNP fine-mapping approaches in simulation studies.
• Successfully applied the method to fine-map a region around the CASP8 gene using iCOGS consortium breast cancer SNP data.

Conclusions:

• The developed Bayesian approach effectively incorporates functional genomic information, including partially observed data, for enhanced SNP fine-mapping.
• This method offers a robust alternative for identifying causal variants in genetic association studies, particularly in complex diseases.