A Bayesian Hierarchical Model for Relating Multiple SNPs within Multiple Genes to Disease Risk

Area of Science:

• Genetics
• Biostatistics
• Computational Biology

Background:

• Studying complex diseases requires analyzing associations among multiple genes within biological pathways.
• Existing methods often struggle to incorporate multiple single nucleotide polymorphisms (SNPs) per gene and external prior information effectively.

Purpose of the Study:

• To extend Bayesian hierarchical modeling for joint analysis of multiple SNPs within genes.
• To integrate external prior information at both SNP and gene levels.
• To perform variable selection at the SNP level and Bayesian shrinkage at the gene level.

Main Methods:

• Developed an extended Bayesian hierarchical model incorporating latent indicator variables for SNP-level variable selection.
• Utilized Bayesian shrinkage at the gene level, informed by external data.
• Employed Markov chain Monte Carlo (MCMC) methods for model fitting.
• Applied the method to analyze 504 SNPs in 38 candidate genes related to DNA damage response.

Main Results:

• Simulation studies demonstrated the model's ability to identify truly causal SNPs and genes.
• The recovery of causal variants was dependent on their frequency and effect size.
• The method was successfully applied to real-world data from the WECARE study.

Conclusions:

• The proposed Bayesian approach offers a robust strategy for dissecting complex genetic associations in disease pathways.
• The model effectively handles multiple SNPs per gene and leverages external information for improved inference.
• This method advances the analysis of genetic data in large-scale epidemiological studies.