A new hybrid algorithm for three-stage gene selection based on whale optimization.

Area of Science:

• Biomedical Data Mining
• Bioinformatics
• Computational Biology

Background:

• High-dimensional gene expression data presents challenges in biomedical data mining due to a larger gene dimension than sample size.
• Effective feature selection is crucial for identifying relevant genes correlated with phenotype, ensuring accurate downstream analysis.

Purpose of the Study:

• To develop and evaluate a novel three-stage hybrid feature gene selection method.
• To enhance the accuracy and efficiency of analyzing high-dimensional gene expression data in biomedical research.

Main Methods:

• A three-stage hybrid approach combining a variance filter, extremely randomized trees, and the whale optimization algorithm.
• Initial dimension reduction using a variance filter, followed by further reduction with extremely randomized trees.
• Optimal feature gene subset selection using the whale optimization algorithm.

Main Results:

• The proposed method demonstrated significant advantages across various evaluation metrics.
• Comparative analysis against other advanced feature selection algorithms showed superior performance.
• Successful evaluation on seven published gene expression profile datasets using three different classifiers.

Conclusions:

• The presented three-stage hybrid feature gene selection method is effective for high-dimensional biomedical data.
• This approach offers significant improvements in selecting optimal gene subsets for accurate phenotypic correlation.
• The method shows promise for advancing biomedical data mining and analysis.