Searching the Druggable Genome using Large Language Models

Area of Science:

• Bioinformatics
• Computational Biology
• Pharmacogenomics

Background:

• The druggable genome includes genes interacting with drugs.
• The Drug-Gene Interaction Database (DGIdb) is a key resource for drug-gene interactions.
• Current DGIdb access requires structured queries, limiting natural language use.

Purpose of the Study:

• To enable natural language querying of the DGIdb.
• To integrate large language models (LLMs) with DGIdb for enhanced biomedical knowledge access.
• To improve the accuracy and accessibility of drug-gene interaction data.

Main Methods:

• Development of the DGIdb Model Context Protocol (MCP) server.
• Facilitating LLM access to the DGIdb API.
• Demonstrating enhanced LLM performance through natural language queries.

Main Results:

• The MCP server allows LLMs to access up-to-date information from DGIdb.
• LLMs demonstrate improved ability to answer biomedical questions using the MCP server.
• Enhanced accuracy in retrieving drug-gene interaction data.

Conclusions:

• The DGIdb MCP server effectively bridges LLMs and structured biomedical databases.
• This facilitates more intuitive and accurate access to drug-gene interaction information.
• Supports broader research and clinical applications by democratizing data access.