DDI-CPI

DDI-CPI predicts drug–drug interactions by analyzing the chemical–protein interactome (CPI) to identify drug–human protein interactions that mediate DDIs.

Key Features:

• Structure-based prediction: Predicts DDIs using only the molecular structure of a submitted compound.
• Docking across 611 human proteins: Performs molecular docking of the submitted molecule against 611 human proteins to generate CPI profiles.
• CPI profile as feature vector: Generates a chemical–protein interactome (CPI) profile that serves as the feature vector for prediction.
• Comparison with 2515-drug library: Assesses potential DDIs between the submitted molecule and a library of 2515 drug molecules.
• Machine learning prediction model: Uses a pre-constructed prediction model based on machine learning to predict DDIs from CPI profiles.
• Predictive performance: Demonstrates Area Under the Curve (AUC) > 0.85 in cross-validation and independent validation tests.
• PK/PD protein exploration: Enables identification of pharmacokinetic/pharmacodynamic (PK/PD) proteins implicated in predicted DDIs via CPI profiles.
• 3D molecular visualization: Provides 3D visualization of drug–protein interactions to inspect binding poses and interaction geometry.

Scientific Applications:

• Pharmacology and toxicology: Predicts potential adverse DDIs to inform safety assessment and toxicity investigations.
• Drug development: Identifies DDI risks early in discovery to guide compound selection and optimization.
• Mechanistic investigation: Reveals protein mediators of DDIs to elucidate mechanisms of action and interaction pathways.
• Combination therapy design: Supports selection and evaluation of drug combinations by predicting interaction profiles.

Methodology:

User-submitted molecules are docked against a predefined set of 611 human proteins to generate CPI profiles, and those CPI profiles are used as input to a pre-constructed machine learning prediction model to assess potential DDIs with known drugs; the workflow integrates computational docking, machine learning, and 3D visualization.