Metabolic BMI model uses AI to reveal hidden metabolic disorders

Researchers at Leipzig University and the University of Gothenburg have developed a novel approach to assessing an individual's risk of metabolic diseases such as diabetes or fatty liver disease more precisely. Instead of relying solely on the widely used body mass index (BMI), the team developed an AI-based computational model using metabolic measurements.

Systems view of metabolic obesity: integrating multi-organ and multi-omics signatures. Credit: Nature Medicine (2026). DOI: 10.1038/s41591-025-04009-7

This so-called metabolic BMI (metBMI) shows that people of normal weight with a high metabolic BMI have up to a fivefold higher risk of metabolic disease. The findings have been published in the journal Nature Medicine.

The conventional body mass index, calculated using height and weight, may indicate overweight but does not reflect how healthy or unhealthy body fat actually is. According to BMI classifications, up to 30% of people are considered to be of normal weight but already show dangerous metabolic changes. Conversely, there are individuals with an elevated BMI whose metabolism remains largely unremarkable. This discrepancy can lead to at-risk patients being identified and treated too late.

For the current scientific study, the international research team sdata from two large Swedish population studies involving a total of almost 2,000 participants. In addition to standard health and lifestyle parameters, extensive laboratory data from blood samples and analyses of the gut microbiome were collected. Based on this dataset, the researchers developed a computational model that predicts metabolic BMI.

"Our metabolic BMI reveals a hidden metabolic disorder that is not always visible on the scales. Two people with the same BMI can have completely different risk profiles depending on how their metabolism and fat tissue function," explained Dr Rima Chakaroun, a researcher at the University of Leipzig Medical Center and first author of the study. She led the research project together with Professor Fredrik Bäckhed during a research stay at the University of Gothenburg.

The results show that an unexpectedly high metabolic BMI (metBMI) is associated with up to a fivefold increased risk of a range of diseases and conditions, including fatty liver disease, diabetes, visceral fat accumulation and insulin resistance. In addition, people with a high metBMI lost 30% less weight following bariatric surgery. These patients underwent surgery at the University of Leipzig Medical Center, which made it possible to collect particularly comprehensive data for the study.

One of the key findings was the close relationship between metabolic profiles and the composition of gut bacteria. Individuals with a higher metBMI showed lower bacterial diversity and a reduced capacity of their gut microbiota to convert dietary fibre into health-promoting fatty acids such as butyrate. The study also highlights that genetic factors play a less important role in metabolic BMI than lifestyle and environmental influences.

The metabolic BMI developed by the researchers is based on extensive measurements of hundreds of small molecules in the blood that reflect cellular metabolism. From more than 1,000 metabolic products initially analyzed, the team was able to identify a reduced panel of just 66 metabolites that retained almost the same explanatory power. These molecules primarily reflect the close interaction between the body's own metabolism and gut bacteria.

Multi-omics prediction of adiposity. Credit: Nature Medicine (2026). DOI: 10.1038/s41591-025-04009-7

"Traditional BMI often overlooks people who are of normal weight but nevertheless have a high metabolic risk. The metBMI can contribute to a fairer and more accurate assessment of disease risk," added Chakaroun.

The model can therefore help to identify affected individuals at an earlier stage, refine the selection of patients for surgical or pharmacological interventions, and personalize therapeutic decision-making. In future, the models are to be further improved by incorporating dynamic markers of insulin secretion and by initiating experimental studies on the gut microbiome–metabolite axis.

The findings were reported in the paper, ‘Multi-omic definition of metabolic obesity through adipose tissue–microbiome interactions’, Nature Medicine. To access this paper, please click here