Background Inflammatory bowel disease (IBD), including Crohn’s disease and ulcerative colitis, is a chronic disorder that markedly impairs quality of life. Current diagnostic and monitoring tools rely on invasive procedures such as endoscopy, which are costly and burdensome. Volatile organic compounds (VOCs) in breath and feces, reflecting host–microbiota metabolism, have emerged as promising non-invasive biomarkers, but their clinical utility remains underexplored.This study aimed to identify breath- and feces-derived VOCs as novel biomarkers for IBD and to establish artificial intelligence (AI)-based predictive models for non-invasive diagnosis and disease activity monitoring. The relationship between VOC alterations and gut microbiota dysbiosis was also investigated. Methods A total of 279 participants (131 IBD patients, 148 healthy controls) were enrolled. VOCs from breath and fecal samples were analyzed using gas chromatography–ion mobility spectrometry (GC-IMS). AI-based machine learning models were developed for diagnosis and monitoring. Furthermore, the differences in breath VOCs identified in human cohorts were validated in a DSS-induced colitis mouse model (2% DSS for 7 days). In a subset of 62 individuals, 16S rDNA sequencing characterized gut microbiota composition and its correlation with VOCs. Results Distinct VOC profiles were identified in IBD. Ethyl sulfide and furfural were elevated in breath samples, while hexanoic acid, pentanoic acid, thiophene, and ethyl acetate were reduced. In fecal samples, dimethyl trisulfide increased, whereas several short-chain fatty acids(SCFAs) and alcohols decreased. The diagnostic model achieved an AUC of 0.92 (sensitivity 96%, specificity 71%), and the monitoring model an AUC of 0.88, both outperforming C-reactive protein and fecal calprotectin. Validation in a DSS-induced colitis model confirmed eight discriminatory VOCs, characterized by depleted SCFA-related VOCs and elevated sulfide VOCs, underscoring their robust correlation with the development and severity of intestinal inflammation. IBD patients showed reduced microbial diversity and depletion of short-chain fatty acid-producing bacteria, closely correlated with altered VOC profiles. Conclusions This study demonstrates that integrating volatomics with AI-based modeling enables accurate, non-invasive diagnosis and monitoring of IBD. The cross-species consistency observed in our human cohorts and DSS-induced colitis mice confirms the reliability of specific VOCs as conserved inflammatory biomarkers. These findings, coupled with VOC–microbiota associations, offer profound mechanistic insights and a promising platform for biomarker-guided precision care. Trial registration ChiCTR, ChiCTR2300073475. Registered 12 July 2023—Prospectively registered, <a href="https://www.chictr.org.cn/bin/project/edit?pid=201603">https://www.chictr.org.cn/bin/project/edit?pid=201603</a>.

Volatomics-based biomarkers for non-invasive diagnosis and monitoring of inflammatory bowel disease

期刊论文
发表日期：

2026年4月9日

查看全文

Xiaowen Li,
Siyuan Pan,
Qingshang Li,
Borong Yu,
Jianxia Ma,
Yuanwen Chen

摘要

Background

Inflammatory bowel disease (IBD), including Crohn’s disease and ulcerative colitis, is a chronic disorder that markedly impairs quality of life. Current diagnostic and monitoring tools rely on invasive procedures such as endoscopy, which are costly and burdensome. Volatile organic compounds (VOCs) in breath and feces, reflecting host–microbiota metabolism, have emerged as promising non-invasive biomarkers, but their clinical utility remains underexplored.This study aimed to identify breath- and feces-derived VOCs as novel biomarkers for IBD and to establish artificial intelligence (AI)-based predictive models for non-invasive diagnosis and disease activity monitoring. The relationship between VOC alterations and gut microbiota dysbiosis was also investigated.

Methods

A total of 279 participants (131 IBD patients, 148 healthy controls) were enrolled. VOCs from breath and fecal samples were analyzed using gas chromatography–ion mobility spectrometry (GC-IMS). AI-based machine learning models were developed for diagnosis and monitoring. Furthermore, the differences in breath VOCs identified in human cohorts were validated in a DSS-induced colitis mouse model (2% DSS for 7 days). In a subset of 62 individuals, 16S rDNA sequencing characterized gut microbiota composition and its correlation with VOCs.

Results

Distinct VOC profiles were identified in IBD. Ethyl sulfide and furfural were elevated in breath samples, while hexanoic acid, pentanoic acid, thiophene, and ethyl acetate were reduced. In fecal samples, dimethyl trisulfide increased, whereas several short-chain fatty acids(SCFAs) and alcohols decreased. The diagnostic model achieved an AUC of 0.92 (sensitivity 96%, specificity 71%), and the monitoring model an AUC of 0.88, both outperforming C-reactive protein and fecal calprotectin. Validation in a DSS-induced colitis model confirmed eight discriminatory VOCs, characterized by depleted SCFA-related VOCs and elevated sulfide VOCs, underscoring their robust correlation with the development and severity of intestinal inflammation. IBD patients showed reduced microbial diversity and depletion of short-chain fatty acid-producing bacteria, closely correlated with altered VOC profiles.

Conclusions

This study demonstrates that integrating volatomics with AI-based modeling enables accurate, non-invasive diagnosis and monitoring of IBD. The cross-species consistency observed in our human cohorts and DSS-induced colitis mice confirms the reliability of specific VOCs as conserved inflammatory biomarkers. These findings, coupled with VOC–microbiota associations, offer profound mechanistic insights and a promising platform for biomarker-guided precision care.

Trial registration

ChiCTR, ChiCTR2300073475. Registered 12 July 2023—Prospectively registered, https://www.chictr.org.cn/bin/project/edit?pid=201603.