AOMP

AOMP predicts aldehyde oxidase (AOX) metabolism for small molecules to identify AOX substrates and metabolic sites, informing drug metabolism and pharmacokinetic risk assessment.

Key Features:

• Graph Neural Network Architecture: Employs a graph neural network (GNN) that represents molecular structures as graphs with atoms as nodes and bonds as edges.
• Integrated Task Prediction: Simultaneously performs metabolic substrate/non-substrate classification and metabolic site prediction.
• Transfer Learning Enhancement: Leverages transfer learning from ^13C nuclear magnetic resonance data to improve predictive performance for both tasks.
• Performance Superiority: Demonstrated superior performance versus benchmark methods using cross-validation and external testing protocols.
• Application in Drug Discovery: Systematically evaluated AOX metabolism of common fragments in kinase inhibitors and identified novel scaffolds with AOX metabolism liability, substantiated by in vitro experimental validation.

Scientific Applications:

• AOX Metabolic Risk Assessment: Rapid evaluation of AOX-mediated metabolic liabilities during drug development.
• Compound Design Optimization: Enables medicinal chemists and pharmacologists to optimize compounds to mitigate AOX-related clearance and toxic metabolite formation.
• Scaffold Identification and Prioritization: Identifies scaffolds with specific AOX metabolism profiles to guide synthesis and lead selection.

Methodology:

Uses a graph neural network on molecular graphs with joint substrate classification and metabolic site prediction, incorporates transfer learning from ^13C NMR data, and evaluates models via cross-validation and external testing.