计算生物学
小桶
推论
生物
模式(遗传算法)
代谢途径
免疫系统
模块化设计
图形
计算机科学
微生物群
功能(生物学)
代谢网络
相关性(法律)
人工智能
人类微生物组计划
生物信息学
分类
疾病
生物途径
作者
Jianfu Li,Ellen Bu,Andrew Lian,Yiming Li,Cui Tao
标识
DOI:10.1038/s41538-026-01105-x
摘要
Diet profoundly influences the gut microbiome and host metabolism, yet the mechanistic pathways linking dietary exposures to health remain fragmented across the literature. We constructed an evidence-weighted Diet–Microbiome–Health Knowledge Graph (DMH-KG) from 1309 curated PubMed abstracts (2023–2025) using a standardized schema of 11 entity categories and 12 relationship types, each carrying a polarity label denoting the relationship’s valence (beneficial, harmful, or neutral/observational). Manual annotation yielded 10,270 entity mentions and 4866 relationships; expert-guided consolidation and pruning produced 4765 entities connected by 4772 frequency-weighted, polarity-labeled edges. Network analysis revealed a modular (Louvain Q = 0.78), small-world organization centered on microbial and dietary hubs. High-confidence pathways emerged, including the probiotic–SCFA–immunity axis and the high-fat-diet–LPS–endotoxemia cascade, both central to age-related immune and metabolic health. Validation against five KEGG and Reactome pathways showed high fidelity (mean entity coverage 92.0%; 100% for Butanoate metabolism and Secondary Bile Acid biosynthesis). Beyond recovering established biology, systematic cross-study traversal surfaced multi-step chains assembled from separate publications, some with no PubMed co-mention detected, as exploratory hypotheses. A comparative pilot showed that grounding a current large language model (GPT-5.1) in DMH-KG evidence improves the mechanistic specificity of its answers. Preserving evidence provenance, relationship polarity, and biological direction, the DMH-KG offers a scalable, auditable platform for mechanistic inference across diet–microbiome–health interactions, with direct relevance to immune regulation, muscle health, metabolic aging, and chronic-disease prevention.
科研通智能强力驱动
Strongly Powered by AbleSci AI