Deep-PK

Deep-PK predicts pharmacokinetic and toxicity properties of small molecules using graph neural networks and graph-based signatures to optimize and understand drug candidates.

Key Features:

• ADMET Property Prediction: Predicts 64 ADMET endpoints covering absorption, distribution, metabolism, excretion, and toxicity.
• General Property Prediction: Predicts 9 additional general molecular properties alongside ADMET endpoints.
• Molecular Optimization: Supports optimization of input molecules to improve pharmacokinetic and toxicity profiles.
• Model Interpretation: Provides interpretation capabilities to aid understanding of pharmacokinetic and toxicity predictions for given molecules.
• Graph-Level Feature Representation: Utilizes graph-based signatures as graph-level features for prediction tasks.

Scientific Applications:

• Drug Development: Evaluates pharmacokinetic properties of small molecules in drug development and high-throughput screening processes.
• ADMET Prediction: Predicts absorption, distribution, metabolism, excretion, and toxicity properties to replace time-consuming and expensive in vitro, in vivo, or pre-clinical data generation.
• Candidate Optimization: Optimizes and understands pharmacokinetics and toxicity for given input molecules in early-stage drug discovery.

Methodology:

Deep-PK applies graph neural networks and graph-based signatures as a graph-level feature to predict pharmacokinetic and toxicity properties.