From data silos to insights: the PRINCE multi-agent knowledge engine for preclinical drug development

Area of Science:

• Pharmaceutical research and development
• Artificial intelligence in drug discovery
• Data integration and management

Background:

• The pharmaceutical industry requires efficient drug development processes amidst increasing costs and regulatory scrutiny.
• Accessing and synthesizing decades of safety study data presents a significant challenge.
• Existing data retrieval methods often lack the sophistication to handle complex, unstructured information.

Purpose of the Study:

• To introduce the Preclinical Information Center (PRINCE), a novel cloud-hosted data integration platform.
• To detail the evolution of PRINCE from a basic search tool to an AI-powered research assistant.
• To demonstrate the application of Large Language Models (LLMs) and advanced retrieval techniques in pharmaceutical data management.

Main Methods:

• Development of PRINCE using a multi-agent architecture powered by LLMs.
• Implementation of advanced data retrieval techniques: Retrieval-Augmented Generation and Text-to-SQL.
• Iterative development guided by user feedback, incorporating a human-in-the-loop approach for accuracy and accountability.

Main Results:

• PRINCE successfully integrates vast amounts of structured and unstructured safety study data.
• The platform evolved to answer complex research questions and assist in drafting regulatory documents.
• Enhanced data accessibility and research efficiency were achieved through AI integration.

Conclusions:

• The PRINCE platform showcases the transformative potential of AI in the pharmaceutical sector.
• AI-driven solutions can significantly improve data accessibility and research efficiency.
• Prioritizing trust, transparency, explainability, and human oversight is crucial for successful AI deployment in regulated industries.