Class Balanced Multifactor Dimensionality Reduction to Detect Gene-Gene Interactions.

Area of Science:

• Genetics
• Biostatistics
• Computational Biology

Background:

• Detecting gene-gene interactions is crucial for understanding disease susceptibility.
• Existing methods for analyzing single-nucleotide polymorphism (SNP) data can be limited by small sample sizes in case-control studies.

Purpose of the Study:

• To propose a novel Balance approach for Multifactor Dimensionality Reduction (BMDR) method.
• To enhance the accuracy of prediction error rate estimates, especially in small sample sizes.
• To improve the detection of gene-gene interactions in genetic association studies.

Main Methods:

• The proposed BMDR method evaluates the average prediction error rates over k-fold cross-validation to select the best model.
• It operates without cross-validation consistency selection.
• Performance was evaluated using simulated datasets with various epistatic models, heritability, and minor allele frequencies.

Main Results:

• BMDR successfully identified gene-gene interactions, demonstrating superior performance on datasets with small sample sizes.
• The method was validated on a large dataset from the Wellcome Trust Case Control Consortium.
• BMDR effectively detected significant gene-gene interactions in real-world data.

Conclusions:

• The BMDR method offers an effective approach for detecting gene-gene interactions, particularly when dealing with limited sample sizes.
• This method improves the accuracy of prediction error rate estimation in genetic studies.
• BMDR provides a valuable tool for dissecting complex disease susceptibility.