ABDpred

ABDpred predicts antimicrobial activity of chemical compounds using machine learning to accelerate antibiotic discovery.

Key Features:

• Machine Learning Algorithms: Employs eight algorithms: extreme gradient boosting, random forest, gradient boosting classifier, deep neural network, support vector machine, multilayer perceptron, decision tree, and logistic regression.
• Training Dataset: Models are trained on a dataset of 312 known antibiotic drugs and 936 non-antibiotic compounds.
• Cross-Validation Approach: Uses five-fold cross-validation to evaluate model performance.
• Ensemble Methodology: Integrates top-performing algorithms—extreme gradient boosting, random forest, gradient boosting classifier, and deep neural network—via a soft-voting ensemble.
• Predictive Performance: The ensemble model achieves accuracy exceeding 80% on both testing and blind datasets.

Scientific Applications:

• High-throughput screening: In silico screening of large chemical libraries to identify compounds with predicted antimicrobial activity.
• Candidate prioritization: Prioritizes compounds for experimental validation in antibiotic discovery pipelines.

Methodology:

Models trained on a dataset of 312 antibiotic drugs and 936 non-antibiotic compounds using eight specified ML algorithms, evaluated with five-fold cross-validation, and combined by a soft-voting ensemble of the four top-performing classifiers.