MCPmed: a call for Model Context Protocol-enabled bioinformatics web services for LLM-driven discovery

Area of Science:

• Bioinformatics
• Computational Biology
• Biomedical Research

Background:

• Bioinformatics web servers are vital for data exploration in biomedical research.
• Current human-centric designs limit machine readability for AI agents.
• Lack of machine-actionable interfaces hinders advanced automation and data integration.

Purpose of the Study:

• To enhance machine readability of bioinformatics web servers for large language models (LLMs) and research agents.
• To develop a standardized, machine-actionable layer for bioinformatics web services.
• To improve automation, reproducibility, and interoperability in biomedical data exploration.

Main Methods:

• Adapted the Model Context Protocol (MCP) to bioinformatics web server backends.
• Created a standardized, machine-actionable layer associating endpoints with scientific concepts and metadata.
• Implemented MCP on widely used databases: GEO, STRING, and UCSC Cell Browser.

Main Results:

• Demonstrated enhanced data exploration capabilities through MCP-enabled LLMs.
• Successfully integrated MCP into established bioinformatics databases.
• Showcased improved machine readability and interaction with bioinformatics resources.

Conclusions:

• MCP provides a standardized approach to make bioinformatics web servers machine-actionable.
• MCP facilitates enhanced exploration of biomedical data by LLMs and research agents.
• MCPmed initiative aims to accelerate adoption and prepare services for future AI-driven research.