MADD: Multi-Agent Drug Discovery Orchestra

2025-11-11

Conference on Empirical Methods in Natural Language Processing

AI Summary

In early-stage drug discovery, hit identification remains costly, and AI tools exhibit high usability barriers for wet-lab researchers. This work introduces the first multi-agent collaborative system tailored for drug discovery, which automatically translates natural-language instructions into end-to-end molecular generation and virtual screening workflows by integrating large language models (LLMs) with domain-specific molecular modeling tools—including molecular docking, generative modeling, and property-based filtering. Key contributions include: (1) AI-first de novo design against five novel therapeutic targets; (2) construction of the first benchmark dataset comprising over 3 million query–molecule pairs; and (3) statistically significant performance gains across seven evaluation tasks versus state-of-the-art LLM-based approaches, culminating in the discovery and public release of multiple validated hit candidates. The system substantially enhances accessibility and practical utility of AI-driven drug discovery for experimental biologists and medicinal chemists.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Cognitive Modeling & Cognitive Systems: Agent Architectures
• Humans and AI: AI for Accessibility

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Economics, Online Markets and Human Computation: Uses of LLMs and GenAI for marketplace design, bidding, and strategic interactions

Abstract

Hit identification is a central challenge in early drug discovery, traditionally requiring substantial experimental resources. Recent advances in artificial intelligence, particularly large language models (LLMs), have enabled virtual screening methods that reduce costs and improve efficiency. However, the growing complexity of these tools has limited their accessibility to wet-lab researchers. Multi-agent systems offer a promising solution by combining the interpretability of LLMs with the precision of specialized models and tools. In this work, we present MADD, a multi-agent system that builds and executes customized hit identification pipelines from natural language queries. MADD employs four coordinated agents to handle key subtasks in de novo compound generation and screening. We evaluate MADD across seven drug discovery cases and demonstrate its superior performance compared to existing LLM-based solutions. Using MADD, we pioneer the application of AI-first drug design to five biological targets and release the identified hit molecules. Finally, we introduce a new benchmark of query-molecule pairs and docking scores for over three million compounds to contribute to the agentic future of drug design.

Problem

• Addresses hit identification challenges in early drug discovery
• Improves accessibility of complex AI tools for wet-lab researchers
• Enhances virtual screening through multi-agent orchestrated pipelines

Innovation

• Multi-agent system orchestrates customized drug discovery pipelines
• Four coordinated agents handle compound generation and screening
• Natural language queries build and execute hit identification workflows