Structured Genome-Wide Association Studies with Bayesian Hierarchical Variable Selection.

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Genome-wide association studies (GWAS) are crucial for identifying genetic variants linked to traits.
• Current methods face challenges in handling the high dimensionality of genetic data and incorporating biological information effectively.
• Selecting important genetic information requires robust strategies for both individual variants and biologically relevant sets.

Purpose of the Study:

• To propose a novel Bayesian framework for hierarchical variable selection at both SNP-set and SNP levels.
• To address the limitation of existing posterior updating schemes in handling ultrahigh-dimensional genetic data.
• To enhance selection power and biological interpretability in genetic association studies.

Main Methods:

• Developed a novel Bayesian framework for hierarchical variable selection.
• Proposed a new sampling scheme to accommodate ultrahigh-dimensional genetic data.
• Constructed an auxiliary SNP-set level model to guide posterior inference for the hierarchical model.

Main Results:

• The proposed method demonstrated computational efficiency.
• Achieved substantially better performance in both SNP-set and SNP selection compared to competing approaches.
• Identified biologically meaningful genetic factors associated with neuroimaging phenotypes in Alzheimer's Disease Neuroimaging Initiative (ADNI) data.

Conclusions:

• The novel Bayesian framework offers an effective approach for variable selection in GWAS.
• The method enhances the identification of genetic factors for complex traits and diseases.
• The framework is generalizable to various biomedical studies requiring genetic analysis.