Dimensionality reduction approach for many-objective epistasis analysis.

The article focuses on the development and evaluation of a many-objective multifactor dimensionality reduction method (MaODR) for identifying single-nucleotide polymorphism–single-nucleotide polymorphism interactions (SSIs) associated with multifactorial diseases. MaODR extends a previous multiobjective MDR (MOMDR) approach by integrating multiple objective functions derived from two-way contingency tables to improve SSI detection, especially for interactions with weak marginal effects. Using an objective function selection approach, the study identified the optimal combination of correct classification rate (CCR), likelihood ratio (LR), and normalized mutual information (NMI) measures, demonstrating superior detection success rates compared to existing algorithms in simulated genetic models and a coronary artery disease genome-wide association study from the Wellcome Trust Case Control Consortium. The results suggest that MaODR-CLN effectively reduces spurious variables and enhances the identification of disease-related genetic interactions without relying on prior inheritance models, supporting its utility in complex genetic analyses.