Prompt-to-Pill: Multi-Agent Drug Discovery and Clinical Simulation Pipeline

Area of Science:

• Computational chemistry and pharmacology
• Artificial intelligence in medicine
• Drug discovery and development

Background:

• Traditional drug discovery is lengthy and expensive.
• Large Language Models (LLMs) show promise in various scientific applications.
• Integrating LLMs across the drug development pipeline remains a challenge.

Purpose of the Study:

• To present a comprehensive, modular AI framework for drug discovery and clinical trial simulation.
• To demonstrate the feasibility of an LLM-based multi-agent system for virtual drug development.
• To bridge the gap between preclinical research and clinical development using AI.

Main Methods:

• Developed a Prompt-to-Pill architecture using a multi-agent system (MAS) with specialized LLMs.
• Implemented a workflow covering target identification, molecular generation, ADMET prediction, docking, and lead optimization.
• Simulated clinical trial phases including patient recruitment and outcome prediction using EHR data.

Main Results:

• Successfully demonstrated the end-to-end AI framework using Dipeptidyl Peptidase 4 (DPP4) as a use case.
• Integrated generative, predictive, and retrieval-based LLMs for seamless workflow operation.
• Showcased the potential of LLM-based agents to operationalize drug development *in silico*.

Conclusions:

• The Prompt-to-Pill framework offers a novel approach to AI-driven drug discovery and virtual clinical trials.
• LLM-based multi-agent systems can significantly accelerate and optimize the drug development process.
• This work provides a foundation for future AI applications in pharmaceutical research.