Peer-reviewed article
1775. Microbiome-Based Classifiers Accurately Differentiate Infectious Diarrhea From Functional Gastrointestinal Disorders and Provide Population-Scale Confidence Measures of Fecal Microbiota Restoration in Recurrent C. DifficileInfection
Wu Q, Ross C, Jones C, Blount K, Savidge T ·
Summary
Written by AI, not yet AICS's view · claude-haiku-4-5-20251001This study develops microbiome-based classifiers to distinguish C. difficile infection from irritable bowel syndrome and evaluates fecal microbiota therapy (RBX2660) in adults and children with recurrent C. difficile infection. While the research includes 447 children, the focus is on microbiota analysis and diagnostic classification rather than child well-being outcomes. The study does not evaluate interventions for improving child health, development, protection, or participation.
Key points
- Random Forest classifiers achieved 97.7% accuracy in predicting C. difficile cases and distinguishing CDI from IBS
- Study analyzed fecal microbiota from 2,129 subjects including 447 children across multiple cohorts
- RBX2660 treatment significantly restored microbiota composition in recurrent CDI cases compared to placebo
- Microbiota restoration accelerated markedly after 30 days of treatment
- Classifiers could reduce disease classification scores and distinguish between conditions in both adults and children
Limitations
This is a laboratory-based microbiota analysis study focused on diagnostic classification rather than clinical child well-being outcomes. The document does not report on health impacts, developmental outcomes, or other indicators relevant to child well-being beyond microbiota composition.
Abstract
Abstract Background Fecal microbiota therapy is being actively pursued as treatment for recurrent C. difficile infection (rCDI), as well as for other GI disease indications associated with dysbiosis, for example, irritable bowel syndrome (IBS). RBX2660 is a microbiota-based drug designed to restore a healthier microbiome and has demonstrated clinical efficacy for preventing rCDI. Despite this and other treatment successes, our understanding of functional microbiota reconstitution at the population scale is still evolving, as is the ability to distinguish IBS from CDI recurrence. Herein we describe development of a Random Forest classifier for CDI diagnosis, and we evaluate microbiome restoration in participants of the Phase 2 trial of RBX2660. Methods Fecal 16S rDNA sequences from 2,129 subjects enrolled in diverse multi-center cohorts were analyzed (1,235 adults and 447 children with CDI, AAD, IBS, or controls). Technical variations due to different DNA extraction, primer region coverage, and sequencing platforms were addressed using closed-reference OTU picking with UCLUST. The RDP classifier and SILVA database assigned taxonomy for each OTU sequence. Stratified random sampling with 50 repeated tests of microbiota training sets was performed for supervised learning. Microbiota signatures of patients in the RBX2660 PUNCH CD2 trial were then assessed using classifiers built to predict CDI treatment outcomes and IBS misdiagnosis. Results Random Forest built the best classifiers accurately predicting 97.7% of CDI cases, and confidently distinguished CDI from IBS patients based on their microbiome signatures (figure). RBX2660 treatment significantly restored microbiota community composition in rCDI cases compared with placebo controls. Conclusion Random Forest classifiers built on a population-scale study of microbiota composition in patients with GI disease provide a highly accurate predictor of CDI cases versus potential IBS misdiagnosis in adults and children. RBX2660 significantly reduced disease classification scores in rCDI patients with a healthy-like microbiota reconstitution markedly accelerating after 30 days of treatment. This study was funded by 1UO1 AI24290-01 and Rebiotix, Inc. Disclosures C. Jones, Rebiotix, Inc.: Employee, Salary. K. Blount, Rebiotix, Inc.: Employee, Salary. T. Savidge, Rebiotix: Grant Investigator, Research grant.