MSMDL-DDI: Multi-Layer Soft Mask Dual-View Learning for Drug-Drug Interactions.

Area of Science:

• Computational chemistry
• Pharmacology
• Artificial intelligence in medicine

Background:

• Drug-drug interactions (DDIs) pose risks to patient safety.
• Current DDI prediction models often fail to capture complex intra- and inter-molecular interactions.
• Incomplete characterization of molecular properties limits existing DDI prediction methods.

Purpose of the Study:

• To develop an advanced model for accurate drug-drug interaction prediction.
• To address the limitations of existing methods in capturing intricate molecular substructure interactions.
• To enhance the comprehensive understanding of intra- and inter-molecular interactions in drug pairs.

Main Methods:

• Proposed a novel Multi-Layer Soft Mask Dual-View Learning for Drug-Drug Interactions (MSMDL-DDI) model.
• Utilized multi-layer soft-masked graph neural networks to extract key drug substructures.
• Implemented a dual-view learning strategy for enriched drug pair representations and attention-based prediction.

Main Results:

• MSMDL-DDI outperformed nine state-of-the-art methods on three real-world datasets.
• Achieved high accuracy (0.9647) on the Twosides dataset for transductive DDI prediction.
• Demonstrated superior performance in both transductive and inductive DDI prediction tasks.

Conclusions:

• The MSMDL-DDI model effectively captures complex molecular interactions for improved DDI prediction.
• The proposed method offers a significant advancement in computational approaches to drug safety.
• This approach holds promise for enhancing patient safety through more accurate DDI risk assessment.