EDock-ML

EDock-ML predicts compound activity against protein targets by combining ensemble docking with machine learning to account for receptor flexibility in molecular docking.

Key Features:

• Ensemble docking: Uses ensemble docking to simulate multiple conformations of a target protein to account for receptor flexibility.
• Machine learning scoring: Applies machine learning algorithms to analyze docking scores and evaluate potential compound efficacy.
• Protein-specific models: Employs a bottom-up approach developing machine-learning models for individual proteins in its database.
• Input formats: Accepts compounds via ZINC database identifiers (Sterling, T.; Irwin, J. J., J Chem Inf Model 2015, 55[11], 2,324-2,337) or by uploaded files from chemical drawing software.
• Probabilistic output: Provides probabilistic assessments classifying compounds as likely active or inactive against specified targets.

Scientific Applications:

• Virtual screening: Supports initial stages of drug discovery by screening compound libraries against protein targets using ensemble docking and ML.
• Activity prediction: Predicts whether individual compounds are likely to be active or inactive against specified drug targets.
• Lead prioritization for medicinal chemistry and pharmacology: Prioritizes candidate molecules for follow-up experimental validation and optimization.

Methodology:

Per-protein machine-learning models are trained on docking scores generated from ensemble docking that simulates multiple protein conformations and produce probabilistic assessments of compound activity.