D4

D4 predicts drug-drug interactions and their underlying mechanisms by integrating deep learning with neuro-symbolic encoding of phenotypic and functional drug knowledge.

Key Features:

• Mechanistic Identification: Predicts DDIs across eleven distinct mechanisms including pharmacokinetic, pharmacodynamic, multi-pathway, and pharmacogenetic interactions.
• Deep Learning Integration: Employs deep learning to generate predictive features from phenotypic and functional data.
• Neuro-Symbolic Strategy: Encodes background biological knowledge using a neuro-symbolic approach within the learning framework.
• Phenotypic and Functional Feature Generation: Utilizes side effects and drug targets as input features for interaction prediction.
• Evaluation with Adverse Event Reporting Systems: Evaluates predictions against adverse event reporting system data.

Scientific Applications:

• DDI Discovery: Identifies both known and novel potential drug-drug interactions.
• Mechanism Elucidation: Determines mechanisms of interaction to inform reduction of adverse drug reactions and therapeutic optimization.
• Patient Stratification: Supports patient stratification for personalized treatment based on predicted interactions and mechanisms.
• Real-World Validation: Grounds predictions in clinical observations by comparison to adverse event reporting systems.

Methodology:

D4 applies deep learning together with a neuro-symbolic approach to encode background biological knowledge and integrate phenotypic data (side effects) and functional data (drug targets) into detailed drug profiles for predicting interaction mechanisms.