Gut Microbe Virulence Factors Power New Machine Learning Test for IBD

Researchers have built machine learning models that distinguish patients with inflammatory bowel disease (IBD) from healthy people using profiles of virulence factors in the gut microbiome, reporting high diagnostic accuracy in an independent validation cohort. The study, published in the Journal of Translational Medicine, combined eight metagenomic datasets into a discovery cohort and then tested the models on three separate datasets.

IBD, which includes Crohn’s disease and ulcerative colitis, is a chronic intestinal condition that affects the quality of life of millions of patients worldwide. Current diagnostic methods are predominantly invasive and costly, which has driven interest in non-invasive, microbiome-based alternatives. Recent studies indicate that the structure and function of the gut microbial community are significantly altered in IBD patients, and that the abundance of certain virulence factors (VFs) is much greater in these patients than in healthy controls.

The team, led by Peijie Li and Jiancheng Wang, established a Discovery cohort by integrating eight metagenomic cohorts comprising 1,062 fecal samples, of which 737 came from patients with IBD and 325 from healthy controls. They implemented a standardized virulence factor profiling workflow, termed VFfinder, to quantify gut microbial virulence factor profiles across the multiple cohorts.

Using species abundance and virulence factor abundance matrices, the researchers developed three machine learning models for IBD classification: a Species Model based on microbial species, a VFs Model based on virulence factors, and IBDseek, which integrates both feature types. A separate Validation cohort, drawn from three metagenomic datasets with 170 IBD samples and 125 healthy controls, was used to evaluate the performance and generalizability of each model.

The results showed that virulence factors in the gut microbiome were significantly enriched in patients with IBD, supporting their potential as stable diagnostic features across cohorts. In the Validation cohort, the 79-species Species Model achieved an area under the curve (AUC) of 0.94 and an accuracy of 88.81%.

The VFs Model, built on 45 virulence factor features, achieved a sensitivity of 96.47% (95% CI, 92.48–98.69%) and an accuracy of 89.49% (95% CI, 85.42–92.75%) in the Validation cohort. Sensitivity measures how reliably a test detects true disease cases, and the narrow confidence intervals reported indicate the precision of these estimates in the validation sample.

The combined IBDseek Model, integrating 45 VF features and 79 species features, delivered the strongest overall performance. It achieved a sensitivity of 96.47% (95% CI, 92.48–98.69%), a specificity of 81.60% (95% CI, 73.68–87.96%), and an accuracy of 90.17% (95% CI, 86.19–93.32%). Specificity reflects how well the model correctly identifies healthy individuals, and the wider confidence interval around this estimate shows somewhat greater uncertainty in that measure.

The authors conclude that their study demonstrates cross-cohort enrichment of bacterial virulence factors in IBD patients and supports their utility as stable diagnostic features. The IBDseek Model achieved the highest observed accuracy and shared the highest sensitivity with the VFs Model in the Validation cohort, suggesting that combining virulence-associated and species-level microbial features improves classification.

These findings support the potential of virulence-associated microbial features as informative biomarkers for noninvasive IBD classification and provide a basis for future development of microbiome-based diagnostic support and risk stratification tools. The decreasing cost of metagenomic sequencing increases the feasibility of translating such high-resolution data into clinical applications.

However, the authors emphasize that prospective evaluation in clinically relevant populations is still needed to establish clinical utility and to determine how such tools would be integrated into existing diagnostic pathways. The reported performance figures come from retrospective validation on existing metagenomic datasets, and the models’ value in real-world clinical settings remains to be demonstrated.

Subject of Research: Machine learning diagnosis of inflammatory bowel disease using gut microbiome virulence factor profiles across multiple metagenomic cohorts

Article Title: Machine learning-based diagnosis of inflammatory bowel disease using gut microbiome virulence factor profiles: multicohort integration and independent validation

Article References: Li, P., Xu, J., Liu, S., Li, Z., Guo, Y., Li, R., Lei, Y., Zeng, K., Peng, X., & Wang, J. (2026). Machine learning-based diagnosis of inflammatory bowel disease using gut microbiome virulence factor profiles: multicohort integration and independent validation. Journal of Translational Medicine. https://doi.org/10.1186/s12967-026-09127-5

Image Credits: AI Generated

DOI: 10.1186/s12967-026-09127-5

Keywords: inflammatory bowel disease, gut microbiome, virulence factors, machine learning, metagenomics, Crohn's disease, ulcerative colitis, non-invasive diagnosis, diagnostic biomarkers, IBDseek, cross-cohort validation, Machine