Emerging drug interaction prediction enabled by a flow-based graph neural network with biomedical network.

Area of Science:

• Pharmacology
• Computational Biology
• Bioinformatics

Background:

• Drug-drug interactions (DDIs) are crucial for disease treatment and drug development.
• Predicting DDIs computationally can improve patient care but is challenging for emerging drugs due to limited data.
• Existing methods often require extensive known DDI information, which is scarce for new drug candidates.

Purpose of the Study:

• To develop an accurate computational method for predicting drug-drug interactions (DDIs) specifically for emerging drugs.
• To leverage rich information within biomedical networks to overcome data limitations for new drug interactions.
• To improve the efficiency of drug development and patient care through enhanced DDI prediction.

Main Methods:

• Proposed EmerGNN, a graph neural network model designed for DDI prediction.
• Learned pairwise drug representations by extracting paths between drug pairs in biomedical networks.
• Incorporated relevant biomedical concepts along paths and weighted network edges for DDI relevance.

Main Results:

• EmerGNN demonstrated higher accuracy in predicting interactions for emerging drugs compared to existing approaches.
• The model effectively leveraged information from biomedical networks to make predictions.
• EmerGNN successfully identified the most relevant information within the biomedical network for DDI prediction.

Conclusions:

• EmerGNN offers a robust solution for predicting DDIs of emerging drugs, addressing the challenge of data scarcity.
• The method enhances the potential for improved patient care and more efficient drug development pipelines.
• Utilizing graph neural networks on biomedical networks is a promising strategy for future DDI prediction research.