CQNR and EPP3D

CQNR and EPP3D identify bacterial effector proteins by applying a cluster quality–based non-reductional oversampling algorithm to balance imbalanced datasets and a structural-feature–based predictor using three-dimensional protein structures.

Key Features:

• CQNR Oversampling Algorithm: Generates new minority class samples near existing minority instances to address class imbalance without removing data as noise.
• 3D Structure-Based Effector Prediction: EPP3D predicts bacterial effector proteins using structural descriptors derived from three-dimensional protein structures.
• Structural Feature Extraction: Utilizes features including convex hull layer count, surface atom composition, radius of gyration, packing density, and compactness.

Scientific Applications:

• Bacterial Effector Protein Identification: Detects and classifies bacterial effector proteins using structural characteristics derived from protein three-dimensional structures.
• Imbalanced Dataset Analysis: Improves predictive modeling for biological datasets with severe class imbalance.
• Structural Bioinformatics: Supports protein function analysis through structural feature-based classification.

Methodology:

CQNR balances imbalanced datasets by generating synthetic minority samples near existing minority instances, and EPP3D analyzes features extracted from protein 3D structures in PDB files to train a classifier for bacterial effector protein prediction.