SSI-DDI

SSI-DDI predicts drug-drug interactions by modeling interactions between chemical substructures using deep learning on raw molecular graph representations to improve DDI prediction accuracy.

Key Features:

• Deep Learning Framework: Employs a deep learning approach that operates directly on raw molecular graph representations of drugs for feature extraction.
• Substructure-Based Analysis: Identifies interactions between specific chemical substructures within each drug rather than treating whole-molecule representations alone.
• Evaluation on Real-World Data: Demonstrates improved prediction performance compared to existing state-of-the-art methods based on evaluation with real-world datasets.

Scientific Applications:

• Pharmacology: Predicts adverse drug-drug interactions to inform drug safety assessments and regimen design.
• Toxicology: Identifies potential toxic interactions at the substructural level to support risk assessment in polypharmacy contexts.

Methodology:

Applies deep learning to raw molecular graph representations to identify substructure–substructure interactions for DDI prediction.