DrugReX: an explainable drug repurposing system powered by large language models and literature-based knowledge graph

Area of Science:

• Computational drug discovery and development
• Pharmacology and pharmaceutical sciences
• Artificial intelligence in medicine

Background:

• Drug repurposing offers a faster, cheaper route to new therapies by finding novel uses for existing medications.
• A key challenge in drug repurposing is the lack of explainability, which hinders researcher trust and understanding of AI-driven predictions.
• Existing computational methods often lack transparency in their decision-making processes.

Purpose of the Study:

• To develop and validate DrugReX, an integrated system for explainable drug repurposing.
• To leverage large language models (LLMs) for enhanced transparency and reliability in therapeutic development.
• To identify potential drug candidates for Alzheimer's disease and related dementias (ADRD).

Main Methods:

• Integration of a literature-based knowledge graph, embedding, and scoring system within the DrugReX platform.
• Utilization of large language models (LLMs) for generating explainable insights and predictions.
• Validation on 15 established drug repurposing cases and application to ADRD candidate identification.

Main Results:

• DrugReX achieved significantly high scores in validating established drug repurposing cases.
• The system identified 25 promising drug candidates for ADRD, with 9 clustering with FDA-approved drugs and 10 linked to clinical trials.
• LLM-generated explanations, supported by the knowledge graph, were rated superior in quality and clarity by domain experts compared to LLM-only explanations.

Conclusions:

• DrugReX successfully bridges computational precision with explainability in drug repurposing.
• The integration of LLMs provides unprecedented transparency, enhancing trust and reliability in therapeutic development.
• This work pioneers the use of LLMs for explainable drug repurposing, paving the way for more informed decision-making.