MGRL-DDI: Multiview Graph Representation Learning for Accurate Drug-Drug Interaction Prediction

Area of Science:

• Computational chemistry
• Pharmacology
• Bioinformatics

Background:

• Drug-drug interactions (DDIs) pose significant clinical challenges, impacting patient safety and treatment efficacy.
• Current prediction methods are limited by single-view drug representations, failing to capture complex drug properties.

Purpose of the Study:

• To develop an advanced framework for predicting drug-drug interactions (DDIs).
• To overcome the limitations of single-view drug representations in DDI prediction.

Main Methods:

• Proposed MGRL-DDI, a multiview graph representation learning framework.
• Integrated three complementary drug structure views: 3D molecular graphs, motif graphs, and molecular graphs.
• Introduced a multiview fusion module to combine information across structural dimensions.

Main Results:

• MGRL-DDI demonstrated superior performance in DDI prediction compared to existing methods.
• Achieved consistent improvements in both warm-start and cold-start scenarios.
• Highlighted the effectiveness of multiview structural modeling for DDI prediction.

Conclusions:

• Multiview graph representation learning offers a more comprehensive approach to modeling drug structures.
• MGRL-DDI significantly enhances the accuracy and robustness of drug-drug interaction prediction.
• The proposed framework holds promise for improving patient safety in clinical practice.