QSP-Copilot: An AI-Augmented Platform for Accelerating Quantitative Systems Pharmacology Model Development

Area of Science:

• Pharmacology
• Computational Biology
• Artificial Intelligence

Background:

• Quantitative Systems Pharmacology (QSP) aids drug development but faces challenges in knowledge integration, model construction, validation, and scalability.
• Traditional QSP workflows are often slow, labor-intensive, and lack consistent validation, hindering efficient application.

Purpose of the Study:

• To introduce QSP-Copilot, an AI-augmented solution to enhance QSP modeling workflows.
• To address limitations in traditional QSP by automating tasks and improving scalability and transparency.

Main Methods:

• Development of QSP-Copilot, an end-to-end AI solution using a multi-agent system and large language models (LLMs).
• Modular support for QSP tasks including project scoping, model structuring, evaluation, and reporting.
• Application and validation of QSP-Copilot on rare diseases: blood coagulation and Gaucher disease.

Main Results:

• QSP-Copilot reduces QSP model development time by approximately 40% through task automation.
• Achieved high extraction precision: 99.1% for blood coagulation and 100.0% for Gaucher disease.
• Systematic documentation by QSP-Copilot improves methodological transparency and reduces manual curation burden.

Conclusions:

• QSP-Copilot significantly improves efficiency and transparency in QSP modeling workflows.
• AI-augmented workflows like QSP-Copilot are pivotal for enhancing scalability and impact in drug development, especially for rare diseases.
• QSP-Copilot facilitates knowledge integration and model construction in biologically complex or data-sparse areas.