A simple Bayesian mixture model with a hybrid procedure for genome-wide association studies

Area of Science:

• Genetics and Genomics
• Statistical Bioinformatics
• Computational Biology

Background:

• Genome-wide association studies (GWAS) often struggle with stringent error correction, leading to missed influential markers or difficulty in quantifying marker importance via P-values.
• Traditional significance testing in GWAS can result in overinterpretation; estimation procedures are preferred for quantifying the proportion of true associations.
• Existing methods may lack accuracy, power, or computational efficiency in identifying significant genetic associations.

Purpose of the Study:

• To develop and validate a Bayesian hierarchical mixture model for directly estimating the proportion of influential genetic markers in GWAS.
• To implement a marker selection procedure using Bayes factors (BF) to quantify the strength of evidence for marker-disease associations.
• To assess the performance of the proposed Bayesian method against existing approaches using simulations and real-world genetic data.

Main Methods:

• A Bayesian hierarchical mixture model was employed to estimate the proportion of influential markers directly.
• A standardized risk measure with unit variance was utilized, simplifying inference and accommodating data dependencies.
• Bayes factors (BF) were used to evaluate the strength of evidence supporting associations between genetic markers and diseases.

Main Results:

• The Bayesian mixture model effectively estimates the proportion of influential markers and accommodates data dependencies with few parameters.
• The Bayes factor magnitude accurately represents the strength of evidence for marker-disease associations.
• The proposed Bayesian procedure demonstrated superior accuracy, power, and computational efficiency compared to existing methods in simulations and real GWAS data.

Conclusions:

• The developed Bayesian hierarchical mixture model provides an accurate and efficient approach for marker selection in genome-wide association studies.
• This method overcomes limitations of traditional P-value based significance testing by directly estimating association proportions and utilizing Bayes factors.
• The R code for this Bayesian procedure is publicly available, facilitating its application in genetic association research.