Radboudumc researchers in the EU-funded HEREDITARY project combine gut bacteria, brain scans and explainable AI to study mental health.
Most of us have felt our stomach tighten before an exam or a difficult conversation. That everyday sensation points to something researchers now take very seriously: the gut and the brain are closely linked, and both matter for brain health. When the relationship between them changes, the effects can sometimes show up in our mental health: in our mood, in how we cope with stress or in how anxious we feel.
Scientists call this link the gut-brain axis. Within the EU-funded HEREDITARY project, a team at Radboud University Medical Center (Radboudumc) in Nijmegen, the Netherlands, is studying it with a combination of tools that rarely meet in one study: gut microbiome profiles, brain scans and explainable artificial intelligence (AI).
A conversation in both directions
The gut and the brain communicate through several routes, and the messages travel both ways. Some connections are direct. The vagus nerve, which runs from the brainstem down to the abdomen, carries signals between the two. Other routes are indirect. Gut bacteria produce metabolites, such as short-chain fatty acids, that can influence the rest of the body, including the brain. The gut also shapes the body’s inflammatory status, which in turn affects how the brain works.
Understanding these routes could tell us more about brain health and mental health. In the longer term, it could also bring us closer to treatments that support both the gut and the brain.
The problem is that both organs produce high-dimensional data. On one side there are hundreds of groups of bacteria; on the other, tens of thousands of connections between brain regions. Finding the few links that really matter in that volume of data is a challenge in itself.
On top of that, the gut is not static. When the Radboudumc team followed more than 400 healthy adults over a year, they found that a person’s gut bacteria changed considerably over time: the differences within one person reached about 60% of the differences between people. Without knowing how much the microbiome naturally fluctuates, it is hard to tell whether a change is meaningful or simply part of everyday life.
Step one: Understanding the gut itself
The team therefore starts with the gut. They draw on the Healthy Brain Study, a cohort of healthy adults in their thirties from the Nijmegen region who were examined in depth at three time points over a year. In 414 of these participants, the researchers measure how much the gut microbiome changes within each person, and which factors relate to that change, such as bacterial diversity, antibiotic use, body fat, alcohol, and dietary fibre. The result is a reference for what normal variation looks like.
First analyses, presented at the HEREDITARY plenary meeting in Lisbon in February 2026, already show distinct microbiome profiles. The larger of two groups identified (330 people) had more body fat, a higher intake of alcohol and saturated fat, and a higher ratio of pro-inflammatory to anti-inflammatory signalling molecules than the smaller group.
The same cohort is also producing the first gut-brain signals. By screening 4,270 candidate models, the team identified a pattern that combines specific gut bacteria with connectivity in brain networks involved in reward and emotion. This pattern was linked to body mass index, anxiety sensitivity and the impact of stressful life events. It also predicted how people valued and chose unhealthy foods in a separate food-choice task that the model had never been trained on.
A related study of 702 participants, published in 2026 in Reviews in Endocrine and Metabolic Disorders, looked at how brain and gut change over time. Increases in the connectivity of the brain’s salience network (which helps us decide what deserves our attention) predicted greater gut bacterial diversity about four months later. Physical activity, in turn, predicted changes in that network. This points to exercise as a possible, accessible lever on the gut-brain axis, although dedicated intervention studies are still needed to test it.
These findings are early and hypothesis-generating rather than clinical tools. But they show that gut-brain data, analysed together, carry information that neither organ reveals on its own.
Step two: From diagnostic labels to biology
The next step moves from healthy volunteers to patients. The MIND-Set cohort of Radboudumc and the Donders Institute includes more than 600 patients with mood and anxiety disorders, substance use disorders, autism or attention deficit hyperactivity disorder (ADHD), together with more than 100 people without a diagnosis.
Here, the team uses machine learning to find shared gut-brain features that underlie brain and mental health. They combine gut bacteria profiles with brain scans and look at symptom dimensions that cut across diagnoses, such as stress responsivity or how people value rewards, rather than only at diagnostic labels. Explainable AI then shows which bacteria and brain connections drive each prediction, so researchers and clinicians can check whether the results make biological sense.
A first test in MIND-Set is encouraging. Using gut bacteria data alone, standard analyses found no differences between people with and without a current depressive episode. A machine learning framework developed within HEREDITARY, which combines many models and explains its own predictions, could tell the two groups apart with a score of 0.81 on a standard performance measure (AUC), where 0.5 equals chance, and 1 is perfect. A conventional approach reached 0.60.
In the long run, this could help group patients by their biology rather than by their label alone, and point towards more personalised treatments.
What comes next
The project is still very much a work in progress. The team is now analysing gut metabolites, including short-chain fatty acids, in the Healthy Brain Study samples together with Radboud University. This will make it possible to test one of the indirect routes between gut and brain directly. In parallel, the team will bring gut, brain and symptom data from MIND-Set together in multimodal models.
Dr Jeshua Tromp, Radboudumc researcher, said: “No single discipline can untangle the gut-brain axis on its own. By combining the expertise of microbiologists, neuroscientists, AI experts and clinicians, we can find patterns that none of us would see alone. The next step is to bring all these pieces together and see what they tell us about mental health.”
Step by step, the work is building the evidence needed to turn the gut-brain axis from an intriguing idea into something that can inform healthcare.
This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101137074
Please Note: This is a Commercial Profile
Please note, this article will also appear in the 28th edition of our quarterly publication.
