Path-based reasoning for biomedical knowledge graphs with BioPathNet.

Area of Science:

• Biomedical informatics
• Network biology
• Machine learning in healthcare

Background:

• Biomedical network analysis is vital for progress, but traditional link prediction methods lack complexity and interpretability.
• Representation learning improves accuracy but faces scalability and interpretability challenges.

Purpose of the Study:

• To introduce BioPathNet, a novel graph neural network framework for biomedical link prediction.
• To enhance accuracy and interpretability in biomedical knowledge graphs using path-based reasoning.

Main Methods:

• BioPathNet utilizes the Neural Bellman-Ford Network (NBFNet) for path-based reasoning in link prediction.
• It incorporates a background regulatory graph (BRG) for improved message passing and stringent negative sampling.
• Learns representations between node pairs by considering all relations along paths.

Main Results:

• BioPathNet outperformed existing methods in gene function annotation, drug-disease indication, synthetic lethality, and lncRNA-mRNA interaction prediction.
• Predicted novel drug indications for ALL and Alzheimer's, validated by experts and clinical trials.
• Identified new synthetic lethality gene pairs and regulatory interactions, confirmed via literature review.

Conclusions:

• BioPathNet offers superior performance and versatility in biomedical link prediction tasks.
• Its interpretability allows researchers to trace prediction paths for molecular insights.
• BioPathNet is a valuable tool for accelerating drug discovery and personalized medicine.