A Bayesian clustering approach for detecting gene-gene interactions in high-dimensional genotype data.

Area of Science:

• Genetics
• Statistical genetics
• Bioinformatics

Background:

• Identifying gene-gene interactions (epistasis) is crucial for understanding complex diseases.
• Bayesian methods offer a powerful framework for analyzing genetic association studies.
• Existing algorithms like BEAM have limitations in partitioning genetic markers.

Purpose of the Study:

• To introduce a novel Bayesian clustering algorithm, ABCDE, for detecting gene-gene interactions in case-control studies.
• To enhance the flexibility of epistasis detection by allowing variable numbers of marker partitions.
• To develop permutation tests for robust validation of identified SNP subsets.

Main Methods:

• Utilized Dirichlet process mixtures for modeling Single Nucleotide Polymorphism (SNP) marker partitions.
• Employed Gibbs weighted Chinese restaurant sampling to simulate posterior distributions.
• Developed permutation tests to validate disease associations of identified SNP subsets.

Main Results:

• The ABCDE algorithm demonstrates flexibility in evaluating any given partition, unlike BEAM's fixed three-group partition.
• Permutation tests provide robust validation, less sensitive to model specification and prior assumptions.
• Comparative analysis with BEAM using simulated and real schizophrenia SNP data was performed.

Conclusions:

• ABCDE offers a more flexible and robust Bayesian approach for detecting epistasis.
• The developed permutation tests improve the reliability of findings in genetic association studies.
• ABCDE shows promise for identifying complex genetic architectures underlying diseases.