GNR: Genetic-Embedded Nuclear Reaction Optimization with F-Score Filter for Gene Selection in Cancer Classification.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Cancer classification from gene expression profiles is challenging due to high-dimensional microarray data.
• Effective gene selection is vital for accurate, efficient, and interpretable cancer diagnostics.

Purpose of the Study:

• Introduce Genetic-Embedded Nuclear Reaction Optimization (GNR), a hybrid metaheuristic for enhanced gene selection.
• Improve cancer classification accuracy and reduce model complexity in precision oncology.

Main Methods:

• Implemented a two-stage approach: F-score filtering followed by GNR optimization.
• Embedded a genetic uniform crossover into the Nuclear Reaction Optimization (NRO) algorithm's fusion phase.
• Utilized Support Vector Machines (SVM) and Leave-One-Out Cross-Validation (LOOCV) for performance evaluation on six cancer datasets.

Main Results:

• GNR consistently outperformed original NRO and other hybrid algorithms across all tested datasets.
• Achieved 100% classification accuracy with significantly smaller gene subsets.
• Demonstrated enhanced local exploitation and preserved global search capabilities.

Conclusions:

• The genetic-embedded fusion strategy in GNR effectively improves gene selection for cancer classification.
• GNR offers a robust and interpretable method for identifying informative gene subsets in precision oncology.
• This approach advances the field of bioinformatics by providing a powerful tool for analyzing complex genomic data.