Bagging-SVM

Bagging-SVM predicts potential druggable proteins from protein sequences by extracting PseAAC, dipeptide composition (DPC), and reduced sequence (RS) features and classifying them with a Bagging ensemble of SVM classifiers for drug-target identification.

Key Features:

• Feature Extraction: Extracts Chou's pseudo amino acid composition (PseAAC), dipeptide composition (DPC), and reduced sequence (RS) features to form a 591-dimensional representation of protein sequences.
• Feature Selection: Applies a genetic algorithm (GA) to select the most informative features from the extracted dataset.
• Ensemble Learning with Bagging-SVM: Aggregates multiple Support Vector Machine (SVM) classifiers via bagging, training on different data subsets to improve robustness and mitigate overfitting.
• Validation and Performance: Assesses model performance using 5-fold cross-validation and reports an accuracy of 93.78%.

Scientific Applications:

• Druggable protein prediction: Identifies proteins with properties indicative of being druggable based on sequence-derived features.
• Therapeutic target discovery: Supports the discovery and prioritization of novel therapeutic targets for drug development.
• Computational evaluation of classifiers: Provides a high-accuracy ensemble SVM approach for comparison with other drug-target prediction methods.

Methodology:

Feature extraction using PseAAC, DPC, and RS to produce a 591-dimensional feature set; feature selection via a genetic algorithm; classification using a Bagging ensemble of SVM classifiers; performance evaluated with 5-fold cross-validation (93.78% accuracy).