Analysis of genome-wide association data by large-scale Bayesian logistic regression

Area of Science:

• Genetics
• Statistical genetics
• Bioinformatics

Background:

• Genome-wide association (GWA) studies generate vast amounts of data, posing challenges for traditional single-locus analysis due to severe multiple comparison adjustments.
• Multivariate logistic regression struggles with high-dimensional data, particularly when sample size is small relative to the number of single-nucleotide polymorphisms (SNPs) or when SNPs are highly correlated.

Purpose of the Study:

• To develop a robust variable selection method for GWA data that addresses the limitations of traditional approaches.
• To accommodate large-scale GWA data while controlling for collinearity and overfitting in high-dimensional predictor spaces.

Main Methods:

• Proposed a variable selection procedure using Bayesian logistic regression to analyze a large number of SNPs simultaneously.
• Explored the relationship between Bayesian regression with specific priors and L1/L2 penalized logistic regression.
• Applied the developed methods to analyze Genetic Analysis Workshop 16 North American Rheumatoid Arthritis Consortium data.

Main Results:

• The Bayesian logistic regression approach effectively selects important SNPs, reducing the severity of multiple comparison and collinearity issues.
• Simulation studies demonstrated the method's ability to correctly identify disease-contributing SNPs.
• Successful application to real-world rheumatoid arthritis data.

Conclusions:

• Bayesian logistic regression offers a powerful tool for variable selection in high-dimensional GWA studies.
• This method enhances the accuracy and efficiency of identifying genetic associations for complex diseases.
• The approach effectively mitigates common statistical challenges in genetic association analysis.