DeepRoot: A KG-Coordinated Multi-Agent System for Therapeutic Reasoning over Historical Medical Texts

2026-06-14 · Source: Artificial Intelligence · Field: Science & Research — Health & Medical Research, Life Sciences & Biology, Artificial Intelligence & Machine Learning · Depth: Expert, quick

Summary

DeepRoot, a multi-agent LLM system, addresses the challenge of extracting verifiable drug-discovery leads from historical medical archives and traditional medicines, which often contain pre-ontological prose and idiosyncratic taxonomies. This system jointly builds and utilizes a verified knowledge graph, demonstrating that grounding and reasoning can be composed as separable axes for therapeutic reasoning. Applied to the Shen Nong Ben Cao Jing, DeepRoot successfully recovers 10 of 21 held-out compound-disease treatment pairs at R@20 (47.6%), significantly outperforming a raw corpus LLM (4.8%) and random (~2.4%). Furthermore, DeepRoot exhibits superior reasoning quality in an LLM-as-judge audit and drastically reduces hallucination, with 7-10% of claims hallucinated compared to 87% for tool-using LLMs.

Key takeaway

For Research Scientists evaluating LLM-based systems for drug discovery from historical medical texts, DeepRoot offers a compelling solution to overcome significant hallucination issues. Your teams should consider adopting KG-coordinated multi-agent architectures to achieve higher accuracy and reasoning coherence, as demonstrated by DeepRoot's 7-10% hallucination rate compared to 87% for tool-using LLMs. This approach can systematically mine and repurpose valuable traditional medical knowledge.

Key insights

DeepRoot's KG-coordinated multi-agent LLM system extracts drug leads from historical texts, separating grounding and reasoning for accuracy.

Principles

• Grounding and reasoning are separable axes in therapeutic reasoning.
• KG coordination significantly reduces LLM hallucination rates.
• Joint KG building and utilization enhances LLM system performance.

Method

DeepRoot jointly builds and utilizes a verified knowledge graph, composing grounding and reasoning for therapeutic analysis. It queries APIs for evidence and integrates LLM reasoning to process historical medical texts.

In practice

• Mine historical medical archives for drug discovery.
• Repurpose traditional medicine knowledge systematically.
• Improve LLM agent reliability in biomedical research.

Topics

• Multi-agent Systems
• Knowledge Graphs
• Large Language Models
• Drug Discovery
• Therapeutic Reasoning
• Medical Text Analysis

Best for: AI Scientist, Research Scientist, Domain Expert

Related on AIssential

• CuraView: A Multi-Agent Framework for Medical Hallucination Detection with GraphRAG-Enhanced Knowledge Verification — CuraView uses GraphRAG and multi-agent verification to detect and explain medical LLM hallucinations, improving factual reliability.
• SMDD-Bench: Can LLMs Solve Real-World Small Molecule Drug Design Tasks? — SMDD-Bench reveals current LLMs solve only 40.2% of complex drug design tasks, underscoring significant challenges in autonomous scientific discovery.
• DeepER-Med: Advancing Deep Evidence-Based Research in Medicine Through Agentic AI — DeepER-Med enhances medical AI trustworthiness via explicit evidence appraisal and agentic collaboration.
• SPOKE: A massive biomedical knowledge graph for precision health and drug discovery — SPOKE integrates vast biomedical data into a knowledge graph for drug discovery and precision medicine.
• A domain-adapted large language model to support clinicians in psychiatric clinical practice — Domain-adapted LLMs can significantly enhance psychiatric clinical workflows, improving accuracy and efficiency.
• Mozi: Governed Autonomy for Drug Discovery LLM Agents — Mozi unifies LLM flexibility with scientific rigor via dual-layer governed autonomy for drug discovery.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.