Identifying novel drug indications through automated reasoning

Area of Science:

• Pharmacology
• Computational Biology
• Bioinformatics

Background:

• In silico drug repurposing leverages vast pharmacological and biological data.
• Traditional literature-based methods generate numerous hypotheses from co-occurrence networks, complicating novel drug indication identification.
• Existing methods often lack fine-grained analysis for identifying indirect drug-disease relationships.

Purpose of the Study:

• To develop an automated reasoning method for acquiring facts from literature and knowledge bases to identify novel drug indications.
• To enhance the identification of indirect relationships for drug indications, improving upon existing literature-based approaches.

Main Methods:

• Developed a novel method using automated reasoning with AnsProlog to encode molecular effects of drug-target interactions and disease links.
• Acquired necessary facts from literature and knowledge bases to build a domain knowledge base.
• Focused on fine-grained analysis to identify indirect relationships for drug indications.

Main Results:

• Applied the method to 943 drugs from DrugBank to assess potential anti-cancer activities.
• Identified 507 drugs with potential for cancer treatment.
• Achieved 82.7% recall for known cancer drugs (67/81) and 49.8% recall for non-cancer drugs in clinical trials for cancer (144/289).

Conclusions:

• The developed method effectively infers drug indications, including novel ones, based solely on molecular targets and interactions.
• Demonstrated the method's capability to discover potential new uses for existing drugs, particularly in cancer treatment.
• The approach shows significant potential for accelerating drug repurposing and identifying alternative therapeutic applications.