Published in late 2025 in the journal Nature, a study of nearly 35,000 people represents a major step up in scale for human microbiome research: It identified and ranked gut bacterial species — both previously known and uncharacterized — that are associated with BMI, clinical health status, and dietary patterns.

To better understand how the gut microbiome is linked to diet and metabolic health, the researchers developed a score that ranks bacteria according to their association with good or poor health. They also carried out dietary interventions to examine their impact on bacterial profiles.

Those results were presented by the study’s lead investigator, Nicola Segata, Department of Cellular, Computational and Integrative Biology, University of Trento, Trento, Italy; European Institute of Oncology, Scientific Institute for Research, Hospitalization and Healthcare, Milan, Italy; Department of Twins Research and Genetic Epidemiology, King’s College London, London, England, at the 14th Gut Microbiota for Health World Summit, held on March 14-15, 2026, in Porto, Portugal.

A New Quantitative Threshold

This study reaches a new quantitative threshold by mobilizing 34,694 individuals drawn from five US and UK cohorts, with linked data (diet, gut metagenome, and clinical biomarkers). The scale enables robust modeling of diet-microbiome-health interactions.

Within these cohorts, the researchers identified clear relationships between certain bacteria and indicators such as weight, blood glucose, and blood lipids. To structure these findings, they built a ranking called the “ZOE Microbiome Health Ranking 2025,” which orders bacteria according to their association with good or poor health.

The score compresses the microbiome’s complex composition into a single indicator. Its construction drew on data from 661 non-rare microbial species (each present in more than 20% of participants), 37 clinical markers — including glucose, lipids, and measures of inflammation — and three physiopathological dimensions: anthropometric measures, fasting metabolism, and postprandial response (after eating).

The result is clear and stable across the five cohorts: Healthy individuals have more favorable bacteria, whereas people who are overweight or ill have more unfavorable bacteria.

“This study identifies intestinal microbiome species — both those already known and those that have not yet been cultured (and whose existence we did not know) — that are closely linked to certain health parameters, to different dietary patterns, and to risk factors. It is the largest dataset of its kind to date, which is a major contribution to the field,” said Lindsey Edwards, research director for The Faecal Microbiota Transplant Programme and lecturer in Microbiology, King’s College London, to the Science Media Centre.

A Global Hierarchy of the Microbiome

Among the results is the finding that 50 favorable species have a mean cumulative abundance of 5.98%, whereas 50 unfavorable species have a mean cumulative abundance of 13.64%.

The imbalance suggests that harmful signatures occupy a larger ecological niche within the gut microbiome. Taxonomically, 92% of the species at the extremes belong to the phylum Firmicutes, with a heavy concentration in the class Clostridia (80 species). Notably, 22 of the 50 species associated with better health remain unknown (uncultured), and among the 28 known favorable species, 24 still lack phenotypic characterization. In short, the bacteria most strongly linked to health outcomes, whether beneficial or harmful, are the least well understood.

A Quantified Link With BMI and Metabolic Phenotypes

The microbiome ranking shows a strong relationship with BMI: The more unfavorable the microbiome profile, the higher the BMI tends to be.

In a meta-analysis of 5348 individuals, differences were clear. People with normal weight had, on average, 5.2 more favorable species than people with obesity (P = .0003). Conversely, people with obesity had 1.95 more unfavorable species (P = .0005). Overall, the microbiome does not operate on a simple binary logic (healthy vs pathological) but along a continuous risk gradient.

The signal persists when the analysis is broadened to disease. Using 25 publicly available case-control microbiome studies (4816 total samples, with n = 2707 controls and n = 2109 cases) covering five diseases with varying levels of association with the gut microbiome (colorectal cancer, inflammatory bowel diseases, type 2 diabetes, glucose intolerance, and cardiovascular diseases), the researchers showed that healthy individuals had, on average, 3.6 more favorable species and -1.6 fewer unfavorable species than patients.

The count of the top 50 ZOE Microbiome health-ranked species was higher in control individuals than in cases for 21 of 25 cohorts. These results exceed those of conventional indicators, demonstrating that the nature of the species present matters more than their quantity.

Dietary Alignment, but Not Perfect Overlap

Regarding diet, a second ranking based on diet (diet rank) was strongly aligned with the health ranking, with a correlation of Spearman rank correlation coefficient of 0.72. However, the relationship is not perfect: 65 species out of 661 showed large discrepancies (difference ≥ 0.3). That means some bacteria can be associated with diets considered unfavorable while producing beneficial effects, for example, by synthesizing protective metabolites such as short-chain fatty acids.

Interventional Validation: Partial Causality

Two randomized trials including 746 participants confirmed the observed dynamics, with an increase in favorable species and a decrease in unfavorable species under dietary intervention.

In the BIOME trial, 57 species were modified in the “prebiotic” arm compared with four in the “probiotic” arm and 14 in the control arm, whereas in the METHOD trial, 46 species were modified under personalized dietary intervention. Species that increased after intervention had significantly more favorable ranks.

Thus, this observation paves the way for a mechanistic approach to nutritional medicine. The microbiome, until now studied descriptively, becomes quantifiable through a unidimensional score applicable in clinical practice. The effect of diet is therefore not limited to absorbed nutrients but includes indirect effects mediated by microbial transformation.

“Modulating the microbiome to improve health is not just about reintroducing microbes but also understanding how we can help them thrive in ways that strengthen our resilience and improve health,” Edwards said.

Remaining Structural Limits

However, structural limitations persist, including the absence of formal causality (with correlation remaining dominant), a high proportion of unknown species, and significant interindividual variability. The next step will be to link this topology to specific biochemical mechanisms, including metabolites and enzymatic pathways, in order to move from robust correlation to exploitable causality.

Segata reported being a consultant for ZOE Ltd. (ZOE Ltd. is a UK-based nutrition company) and having received stock options from the company.

This story was translated from Medscape's French edition.