EBST

EBST applies a modified Multi Objective Imperialist Competitive Algorithm (MOICA) to identify microRNA (miRNA) biomarkers for ovarian cancer, aiming to improve early detection and diagnostic accuracy.

Key Features:

• Algorithm: Uses a Modified Multi Objective Imperialist Competitive Algorithm (MOICA) as the core evolutionary multi-objective optimizer.
• Objective functions: Integrates six objective functions to evaluate classifier performance, structure evaluation, clustering error, and the minimum Redundancy Maximum Relevance (mRMR) filter, and in practice employs five objective functions (four associated with l_1-SVM performance and one for average mRMR ranking).
• Classifier: Employs an l_1-SVM classifier for performance evaluation within the multi-objective framework.
• Pre-processing filter: Applies a False Discovery Rate (FDR) filter during pre-processing to refine data quality.
• Selected biomarkers: Identifies 11 microRNAs: hsa-miR-6784-5p, hsa-miR-1228-5p, hsa-miR-8073, hsa-miR-6756-5p, hsa-miR-1307-3p, hsa-miR-4697-5p, hsa-miR-3663-3p, hsa-miR-328-5p, hsa-miR-1228-3p, hsa-miR-6821-5p, and hsa-miR-1268a.
• Performance metrics: Reports classification performance of 100% sensitivity, 99.38% specificity, 99.69% accuracy, and 99.39% positive predictive value.
• Biological validation: Validates the biological relevance of selected miRNAs using bioinformatics tools and existing literature with involvement in cancer signaling pathways.
• Comparative performance: Demonstrates superior performance relative to routine state-of-the-art methods in biomarker identification for ovarian cancer.

Scientific Applications:

• Ovarian cancer biomarker discovery: Prioritizes miRNA biomarkers for early detection and diagnostic accuracy in ovarian cancer.
• Diagnostic model evaluation: Provides classification metrics to assess diagnostic performance of selected miRNA signatures.
• Biological interpretation: Supports downstream pathway analysis and literature-based validation of miRNAs implicated in cancer signaling pathways.

Methodology:

Computational methods include a Modified Multi Objective Imperialist Competitive Algorithm (MOICA) with six objective functions (practically five: four tied to l_1-SVM performance and one average mRMR ranking), use of the minimum Redundancy Maximum Relevance (mRMR) filter, and a False Discovery Rate (FDR) pre-processing filter.