The use of knowledge graphs for drug repurposing: From classical machine learning algorithms to graph neural networks

Area of Science:

• Biomedical informatics
• Computational pharmacology
• Artificial intelligence in drug discovery

Background:

• Drug repurposing accelerates the development of new therapies by identifying novel indications for existing drugs.
• Computational approaches, including artificial intelligence (AI), are increasingly vital for discovering drug repurposing candidates.
• Knowledge graphs (KGs) provide a robust framework for modeling complex biomedical knowledge and predicting drug-disease associations.

Purpose of the Study:

• To provide a comprehensive review of computational drug repurposing methods that leverage knowledge graphs.
• To explore the rationale behind using KG-based representations in drug discovery.
• To analyze various KG-based tools, their methodologies, and performance in predicting drug-disease links.

Main Methods:

• Review and synthesis of existing literature on KG-based computational drug repurposing.
• Discussion of traditional machine learning and deep learning techniques applied to KGs for drug repurposing.
• Analysis of specific KG-based tools, focusing on their construction, link prediction accuracy, and limitations.

Main Results:

• KG-based methods offer an intuitive approach to integrate and exploit diverse biomedical data for drug repurposing.
• Both traditional machine learning and advanced deep learning models show promise in KG-based link prediction for identifying drug-disease relationships.
• A variety of KG-based tools exist, each with unique strengths and weaknesses in terms of data integration and predictive power.

Conclusions:

• Knowledge graphs are instrumental in advancing computational drug repurposing by effectively modeling biomedical knowledge.
• The integration of AI, machine learning, and KGs represents a significant frontier in accelerating the identification of new therapeutic uses for existing drugs.
• Further development and application of KG-based methodologies are crucial for optimizing drug discovery pipelines.