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

2026-06-14

AI Summary

Historical medical texts are difficult to integrate into modern drug discovery due to inconsistent terminology and the absence of structured ontologies, while existing large language models (LLMs) suffer from high hallucination rates and low reliability when extracting verifiable insights from such sources. To address this, this work proposes DeepRoot—a multi-agent LLM framework that, for the first time, decouples and synergistically integrates knowledge graph construction with therapeutic reasoning. By grounding inference in an empirically validated knowledge graph, DeepRoot enables retrieval-augmented reasoning that substantially reduces hallucination rates to 7–10%, compared to a baseline of 87%. Evaluated on the *Shennong Ben Cao Jing*, it achieves a recall@20 of 47.6%, vastly outperforming the baseline of 4.8%, and demonstrates superior reasoning quality and accuracy across all metrics, establishing a new paradigm for mining historical medical knowledge.

Abstract

Historical medical archives and traditional medicines hold immense potential for drug discovery and remain a primary source for current drug development. However, pre-ontological prose and idiosyncratic taxonomies prevent the standardization and medical modernization of the data for use in current biomedical pipelines. Furthermore, no existing LLM agent system, whether tool-calling, retrieval-augmented, or agentic deep-research, can convert such text into verifiable drug-discovery leads at scale. We close this gap with DeepRoot, a multi-agent LLM system that jointly builds and utilizes a verified knowledge graph, showing that grounding and reasoning -- often conflated -- are separable axes the system can compose for therapeutic reasoning. Applied to the Shen Nong Ben Cao Jing, DeepRoot recovers $10$ of $21$ held-out compound-disease treatment pairs at R@$20$ ($47.6\%$ vs $4.8\%$ for a raw corpus LLM and $\sim\!2.4\%$ random) and dominates an LLM-as-judge audit for reasoning quality over baseline LLMs and LLMs with direct tool-call access to the same APIs DeepRoot itself queries. Tool-using LLMs hallucinate evidence on $87\%$ of claims, versus 7-10% for DeepRoot. Graph-only inference hallucinates $0\%$ but ranks lowest on reasoning coherence; DeepRoot KG+LLM is the only condition to win on both axes, pointing toward a route for systematic mining and repurposing of historical medical knowledge.

Problem

• historical medical texts
• therapeutic reasoning
• knowledge graph
• drug discovery
• LLM hallucination

Innovation

• multi-agent LLM
• knowledge graph coordination
• therapeutic reasoning
• historical medical text mining
• hallucination reduction