Optimizing accuracy and dimensionality: a swarm intelligence strategy for robust cancer genomics classification

Area of Science:

• Bioinformatics
• Computational Biology
• Machine Learning in Oncology

Background:

• High-dimensional gene expression data presents challenges in cancer classification, including data redundancy, noise, and overfitting.
• Effective feature selection is crucial for improving the accuracy and interpretability of cancer classification models.

Purpose of the Study:

• To develop and evaluate a novel hybrid framework combining the Dung Beetle Optimizer (DBO) for gene feature selection and Support Vector Machines (SVM) for cancer classification.
• To address the limitations of high-dimensional data in cancer classification by reducing noise and redundancy.

Main Methods:

• A hybrid DBO-SVM framework was proposed, utilizing DBO's nature-inspired optimization for selecting informative gene subsets.
• Support Vector Machines with Radial Basis Function (RBF) kernels were employed for classification on the selected features.
• The framework was validated on diverse, publicly available cancer gene expression datasets.

Main Results:

• The DBO-SVM framework achieved high accuracy rates: 97.4-98.0% on binary classification and 84-88% on multiclass classification tasks.
• The model demonstrated balanced Precision, Recall, and F1-scores, indicating robust performance across different classes.
• Significant reduction in computational cost and improved biological interpretability were observed.

Conclusions:

• The proposed DBO-SVM hybrid model effectively enhances cancer classification accuracy and efficiency using high-dimensional gene expression data.
• This approach shows strong potential as a reliable tool for precision medicine and biomedical data analysis.
• The integration of DBO for feature selection offers a promising strategy for complex biological data challenges.