TRACEDD: A Tool-grounded Reasoning and Agentic Coordination for Explainable Drug Design

Drug discovery involves a series of coordinated decisions across various stages, including target validation, structure analysis, molecular design, developability assessment, and synthetic feasibility. However, existing computational methods typically function as separate tools, leading to potential gaps in the overall process. To address this challenge, the researchers have developed TRACEDD (Tool-grounded Reasoning and Agentic Coordination for Explainable Drug Design), a framework that aims to overcome these limitations.

TRACEDD introduces three main contributions: a tool-first multi-agentic architecture, a multi-agent system that mirrors expert teams, and an end-to-end workflow from target validation to synthesis planning.

The framework employs a multi-agent system that closely resembles the decision-making process of expert discovery teams. This system operates through a Reason-Act-Observe loop, ensuring transparent and traceable decision-making. TRACEDD decomposes the drug discovery process into specialized agents, each responsible for specific tasks such as target validation, druggability assessment, molecular generation, lead optimization, ADMET evaluation, literature evidence integration, and retrosynthesis. These agents collaborate using the Reason Act Observe workflow.

To demonstrate the effectiveness of TRACEDD, the researchers applied the framework to the case of JAK2 as a representative target. The system successfully retrieved experimental protein structures, utilized AlphaFold when structures were not available, identified druggable pockets, and performed de novo molecular generation. Known JAK2 inhibitors were used to define design hypotheses and guide reinforcement learning-based molecular generation.

The framework incorporated docking scores, predicted pIC50, and other physicochemical/ADMET properties as reward and prioritization signals for the molecular generation process.

By adopting a tool-first approach and integrating agentic orchestration with domain-specific computational tools, TRACEDD offers a transparent, adaptable, and human-verifiable molecular design workflow. This innovative framework provides a solid foundation for more reliable AI-assisted drug discovery, overcoming the limitations of traditional disconnected computational methods.