Hybrid Gene Selection Algorithm for Cancer Classification Using Nuclear Reaction Optimization (NRO).

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Microarray gene expression data present challenges in cancer classification due to high dimensionality and small sample sizes.
• Effective gene selection is crucial for accurate and interpretable cancer classification models.

Purpose of the Study:

• To develop and evaluate a hybrid gene selection method combining filter-based reduction and metaheuristic optimization for cancer classification.
• To assess the performance of the proposed F-score-based Nuclear Reaction Optimization (F-NRO) method on diverse cancer datasets.

Main Methods:

• A hybrid approach integrating the F-score statistical filter for initial gene ranking and reduction.
• Utilizing Nuclear Reaction Optimization (NRO) as a metaheuristic optimizer to refine gene subset selection.
• Evaluating the F-NRO method using Support Vector Machines (SVMs) and Leave-One-Out Cross-Validation (LOOCV) on six cancer microarray datasets.

Main Results:

• The F-NRO method achieved high cancer classification accuracy across multiple datasets.
• Perfect classification accuracy was obtained on five out of the six evaluated cancer datasets.
• F-NRO demonstrated the ability to identify compact and informative gene subsets for classification.

Conclusions:

• The F-score-based Nuclear Reaction Optimization (F-NRO) method is an effective and interpretable solution for gene selection in cancer classification.
• This hybrid approach addresses the challenges of high dimensionality in microarray data for improved diagnostic tools.
• F-NRO shows significant potential for advancing cancer research and clinical applications through precise gene identification.