Evaluating the Nuclear Reaction Optimization (NRO) Algorithm for Gene Selection in Cancer Classification.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Cancer classification from microarray data is challenging due to high dimensionality.
• Effective gene selection is crucial for accurate cancer subtyping and analysis.
• Advanced optimization techniques are needed to handle complex genomic datasets.

Purpose of the Study:

• To introduce and evaluate Nuclear Reaction Optimization (NRO) for gene selection in cancer microarray datasets.
• To assess NRO's performance as a standalone method without prior dimensionality reduction.
• To compare NRO's effectiveness against other optimization algorithms for cancer classification.

Main Methods:

• Nuclear Reaction Optimization (NRO), inspired by nuclear processes, was applied to six cancer microarray datasets.
• Support Vector Machine (SVM) and k-Nearest Neighbors (k-NN) classifiers were used to evaluate selected gene subsets.
• Leave-One-Out Cross-Validation (LOOCV) was employed for robust classification accuracy assessment.

Main Results:

• NRO demonstrated high classification accuracy, especially when combined with SVM.
• The algorithm showed competitive performance against state-of-the-art methods like HHO, ABC, PSO, and FFA on select datasets.
• The number of selected genes was relatively high due to the lack of additional feature reduction steps.

Conclusions:

• Nuclear Reaction Optimization (NRO) is a viable approach for gene selection in cancer classification.
• Further enhancements can be achieved by integrating NRO with hybrid models and feature reduction techniques.
• Future research should focus on optimizing gene subset size for improved efficiency and accuracy.