Gut Microbiome Study Finds Subtle Microbial Shifts, Not Global Dysbiosis, in Gut–Brain Disorders

People with disorders of gut–brain interaction (DGBI) — a family of chronic conditions that includes irritable bowel syndrome (IBS), functional dyspepsia and functional constipation — do not show a dramatic, across-the-board disruption of their gut microbiome, according to a new metagenomic study published in Frontiers in Microbiology. Instead, the research suggests that microbial changes in these conditions are subtler and more fragmented than the word “dysbiosis” often implies, with distinct microbial signatures emerging only when patients are separated by clinical subtype. The finding matters because DGBI affect up to 40 percent of the global population, yet their biological underpinnings remain frustratingly elusive, and microbiome research has been one of the most active fronts in the search for mechanisms.

The study, led by researchers at the Affiliated Guangdong Second Provincial General Hospital of Jinan University together with collaborators at the 01Life Institute and the Research Institute of Tsinghua University in Shenzhen, enrolled 86 patients diagnosed according to the Rome V criteria, the current international standard for classifying DGBI by symptom pattern. The cohort included 39 women and spanned six subtypes: 33 patients with functional bloating, 18 with mixed-habit IBS (IBS-M), 14 with diarrhea-predominant IBS (IBS-D), 13 with functional dyspepsia, 5 with functional constipation, and 3 with constipation-predominant IBS (IBS-C). Participants stopped taking medications known to affect the gut or its microbes — including antibiotics, probiotics and proton pump inhibitors — before providing stool samples for shotgun metagenomic sequencing, a technique that reads the collective genetic material of the gut community and can resolve microbes to the species level, beyond what coarser 16S rRNA surveys can achieve.

To serve as a comparison group, the team drew on publicly available metagenomic data from the NCBI database (accession PRJNA557323), selecting 86 healthy controls matched to the patients for age and body mass index. The two groups were statistically indistinguishable in mean age (roughly 42 versus 44 years) and BMI (about 23 in both). Because the patient and control samples were sequenced in different projects, the researchers applied a batch-effect correction method called MMUPHin to the abundance data at the phylum, genus and species levels and to functional profiles, while preserving the biological group difference as a covariate. Even so, the authors acknowledge that residual technical confounding from differences in DNA extraction, library preparation and sequencing platforms cannot be fully excluded — a caveat that shapes how the headline results should be read.

The core comparison delivered a strikingly null result at the community level. Alpha diversity — a measure of how many species are present and how evenly they are distributed — showed no significant differences between patients and controls in observed richness (p = 0.36), Pielou evenness (p = 0.24) or Shannon diversity (p = 0.55). Beta diversity, which asks whether whole communities differ in composition, likewise showed substantial overlap between the groups: a principal coordinate analysis of Bray–Curtis distances yielded a PERMANOVA effect size of just R² = 0.0077 with p = 0.165, meaning group membership explained less than one percent of the variation in community structure. At the genus level, the five most abundant taxa — Bacteroides, Prevotella, Eubacterium, Alistipes and Faecalibacterium — were largely similar between groups, with Prevotella more abundant in healthy controls.

Yet the picture was not entirely flat. Co-occurrence network analysis, which maps statistical correlations between species, revealed different patterns of interspecies association in the two groups. Taxa associated with DGBI, exemplified by Clostridium citroniae and Hungatella hathewayi, tended to cluster within a relatively compact module, whereas control-associated taxa such as Coprococcus comes and Oscillibacter sp. 57_20 occupied a partially distinct module. Species-to-pathway correlation analysis further suggested that DGBI-linked modules were connected to multiple metabolic pathways, while control-associated taxa formed smaller, more separated modules. The authors stress that such networks are built on statistical associations and cannot establish direct causal interactions between microbes, but the results hint that DGBI may involve a remodeling of ecological relationships within the gut community rather than the loss or gain of individual species.

Functional profiling based on the MetaCyc database pointed in the same direction. Relative to controls, the DGBI group showed higher estimated abundance of pathways related to bacterial cell envelope biosynthesis — including lipid A/LPS-related and peptidoglycan biosynthesis pathways — as well as carbon-source utilization, glycol metabolism and degradation, aromatic compound degradation, allantoin and urate-related metabolism, and isoprenoid precursor biosynthesis. Controls, by contrast, showed higher relative abundance of an allantoin degradation pathway. Because these estimates derive from metagenomic sequencing of microbial DNA, they reflect functional potential rather than measured metabolic activity; the authors note that confirming actual metabolic output would require metabolomic assessment.

To test whether the microbiome could distinguish patients from controls, the team built a random forest classifier on species-level abundances, using 70 percent of samples for training with Bayesian hyperparameter optimization and five-fold cross-validation, and selecting features with the Boruta algorithm. On the held-out 30 percent, the model achieved an overall accuracy of 73.1 percent, with an area under the ROC curve of 0.769 — moderate discrimination, but not enough for clinical use. The authors are explicit that, given the potential batch effects between the patient cohort and the external control dataset, the model should be regarded as exploratory and its performance is insufficient to establish a diagnostic signature for DGBI.

The most biologically interesting results emerged when the patients were split by subtype. Functional dyspepsia patients showed significantly lower observed species richness than IBS-C and IBS-M patients, while IBS-M patients displayed significantly higher Pielou evenness and Shannon diversity than those with functional bloating or functional dyspepsia. LEfSe analysis, which identifies taxa that discriminate between groups using effect sizes, flagged subtype-specific signatures: the top discriminative species in IBS-C included Prevotella stercorea, Ruminococcus torques, Bacteroides finegoldii, Asaccharobacter celatus and Alistipes timonensis, while Victivallis vadensis was strongly associated with functional constipation, all at linear discriminant analysis scores above 3.0 and false-discovery-rate-adjusted p values below 0.05. Bacteroides vulgatus dominated across all subtypes, Prevotella copri was more abundant in functional bloating but undetected in IBS-C, and the beneficial commensal Faecalibacterium prausnitzii was relatively more abundant in IBS-C patients. Notably, no significant beta-diversity differences were found among subtypes at any taxonomic level.

An exploratory correlation analysis added a clinical dimension: in the IBS-C subtype, the abundance of Asaccharobacter celatus — an obligate anaerobe in the family Coriobacteriaceae that participates in intestinal equol biosynthesis — was negatively correlated with the Wexner constipation score, a validated eight-item scale on which higher scores indicate more severe constipation. The authors urge caution on this point, since the IBS-C group contained only three patients. Distance-based redundancy analysis showed that clinical parameters, particularly anxiety score, BMI and Wexner score, were closely associated with variation in microbial community structure across subtypes, while anxiety and bloating scores themselves did not differ significantly between subtypes. Constipation-predominant patients (FC and IBS-C) scored above 20 on the Wexner scale, whereas other subtypes all scored below 10.

The study’s limitations frame its contribution honestly. Beyond the batch effects inherent in comparing sequenced cohorts, several subtypes — especially the constipation-predominant ones — had very small sample sizes, limiting statistical power, and the Abdominal Bloating Score used in the study was developed by the authors and has not been formally validated. The inconsistent findings on alpha diversity in IBS across the wider literature, which a systematic review has attributed to differences in populations, subtypes and methods, suggest that pooling heterogeneous DGBI phenotypes may mask real microbial differences. Taken together, the results support the idea that DGBI involve specific, subtype-linked microbiota alterations rather than a single, pronounced global dysbiosis — a reframing that could help explain why microbiome studies of functional gut disorders have so often produced conflicting results, and one that the authors say now requires validation in larger, independently recruited cohorts with standardized sampling and sequencing protocols.

Subject of Research: Gut microbiota profiles in patients with disorders of gut–brain interaction characterized by shotgun metagenomic sequencing

Article Title: Metagenomic characterization of gut microbiota in patients with Disorders of Gut–Brain Interaction

Article References: Tan, W., Shi, W., Song, J., Chen, X., Lin, M., Xu, Y., Ling, Y., Ma, Y., Xia, Y., Wang, Z., Liang, B., & Yang, X. (2026). Metagenomic characterization of gut microbiota in patients with Disorders of Gut–Brain Interaction. Frontiers in Microbiology, 17, Article 1890519. https://doi.org/10.3389/fmicb.2026.1890519

Image Credits: AI Generated

DOI: 10.3389/fmicb.2026.1890519

Keywords: disorders of gut-brain interaction, gut microbiota, metagenomics, irritable bowel syndrome, functional dyspepsia, functional constipation, dysbiosis, Rome V criteria, random forest, Faecalibacterium, alpha diversity, co-occurrence networks