CNN-BLPred

CNN-BLPred classifies β-lactamase (BL) enzymes using a convolutional neural network to improve molecular and functional classification for antimicrobial resistance research.

Key Features:

• Deep Learning (CNN): Uses a convolutional neural network with a simple architecture including a single convolutional layer for BL classification.
• Molecular and Functional Classification: Supports both molecular and functional classification of β-lactamase enzymes.
• Gradient Boosted Feature Selection (GBFS): Applies GBFS using Gradient Boosted Trees to reduce the initial 10,912 features by approximately 95%.
• Benchmarking: Evaluated with leave-one-out cross-validation, 10-fold cross-validation, and independent test sets and compared against RNNs and Random Forests.
• Performance Improvements: Reported a 7% increase in classic β-lactamase prediction accuracy and an average 25.64% improvement across Classes A, B, C, and D during 10-fold cross-validation.
• Class Balancing: Employs Random Oversampling (ROS), Random Undersampling (RUS), and SMOTE to mitigate class imbalance.

Scientific Applications:

• Antimicrobial resistance research: Enables accurate classification of β-lactamase enzymes to support studies of antibiotic resistance prevalence and spread.
• Enzyme function and mechanism studies: Assists investigation of β-lactamase molecular and functional diversity relevant to resistance mechanisms.
• Antibiotic development: Provides classification data that can inform design and evaluation of β-lactamase–targeting antibiotics and inhibitors.

Methodology:

Feature extraction and selection followed by GBFS using Gradient Boosted Trees to reduce 10,912 features by ~95%; classification via a CNN with a single convolutional layer; class balancing using ROS, RUS, and SMOTE; benchmarking with leave-one-out cross-validation, 10-fold cross-validation, and independent test sets with comparisons to RNNs and Random Forests.