A probabilistic knowledge graph for target identification.

Area of Science:

• Computational biology
• Machine learning in drug discovery
• Bioinformatics

Background:

• Drug discovery is costly and failure-prone.
• Experimental target identification methods are labor-intensive.
• Computational approaches, particularly machine learning, show promise for drug discovery.

Purpose of the Study:

• To introduce Progeni, a novel machine learning framework for identifying safe and efficacious disease targets.
• To improve the efficiency and success rate of drug discovery.

Main Methods:

• Progeni integrates heterogeneous biological networks from diverse sources.
• It constructs a probabilistic knowledge graph by incorporating literature evidence.
• Graph neural networks learn feature embeddings for biological entity identification.

Main Results:

• Progeni demonstrated superior predictive performance compared to baseline methods.
• The framework showed robustness against exposure bias.
• Progeni identified novel targets strongly supported by existing literature.
• Wet lab experiments validated the biological significance of predicted targets for melanoma and colorectal cancer.

Conclusions:

• Progeni is a powerful tool for advancing drug discovery.
• It effectively identifies biologically relevant and validated drug targets.
• The framework offers a more efficient and reliable alternative to traditional methods.