Hierarchical Naive Bayes for genetic association studies

Area of Science:

• Genetics
• Bioinformatics
• Statistical Genetics

Background:

• Genome Wide Association Studies (GWAS) are crucial for understanding complex traits.
• Traditional single-SNP analysis fails to capture the multifactorial nature of genetic disorders.
• Linkage Disequilibrium (LD) among SNPs presents a challenge for standard GWAS methods.

Purpose of the Study:

• To introduce a Hierarchical Naïve Bayes (HNB) classification model.
• To address the challenge of analyzing SNPs in Linkage Disequilibrium (LD).
• To improve classification performance in genetic studies.

Main Methods:

• Implemented a Hierarchical Naïve Bayes model.
• Grouped SNPs in LD regions as replicates of a locus.
• Utilized latent variables to summarize information from correlated SNPs within blocks.

Main Results:

• The HNB model demonstrated superior classification performance compared to standard Naïve Bayes.
• Validation was performed on simulated datasets with varying numbers of SNPs.
• The approach was successfully applied to real genetic datasets for Type 1 and Type 2 Diabetes.

Conclusions:

• The Hierarchical Naïve Bayes model effectively handles classification with correlated SNPs.
• This method improves upon standard Naïve Bayes by incorporating within-locus variability.
• The HNB approach offers enhanced classification for complex genetic trait analysis.