Reviewing the Computational Landscape of Drug Repurposing: Evolution from Structure-Based Methods to LLM-Based...

Area of Science:

• Computational drug discovery
• Pharmacology
• Bioinformatics

Background:

• Traditional drug discovery is lengthy, expensive, and high-risk.
• Drug repurposing offers a faster, cheaper, and safer alternative by identifying new uses for approved drugs.
• Computational strategies are crucial for efficient drug repurposing.

Purpose of the Study:

• To provide a comprehensive survey of computational drug repurposing methodologies.
• To clarify the principles, applications, and limitations of various computational approaches.
• To offer insights into future research directions in drug repurposing.

Main Methods:

• Categorization of methods into biological mechanism-driven (structure-based, omics-based, fuzzy logic-based, adverse event-based), network-based (graph mining, matrix factorization/completion), and data-driven (text mining, large language models).
• Elaboration on the principles, advantages, and challenges of each methodological category.
• Discussion on the integration of multi-source data and the evolution of text mining to LLM-based methods.

Main Results:

• Biological mechanism-driven methods offer deep mechanistic insights.
• Network-based methods facilitate systematic prediction and integration of diverse data.
• Data-driven methods, especially LLMs, enhance information extraction from literature.
• Each method has unique strengths and limitations.

Conclusions:

• The future of drug repurposing relies on the intelligent integration of diverse computational methodologies.
• Network-based and data-driven methods are poised to enable large-scale drug repurposing.
• Biological mechanism-driven methods remain essential for rigorous validation and explanation of repurposed drugs.