MHH researchers aim to investigate the interaction between diet, the microbiome, and metabolism and to provide personalized dietary recommendations.
Molecular biologist Dr. Mattea Müller aims to develop a predictive tool for personalized dietary recommendations using AI. Copyright: Karin Kaiser/MHH
Our diet influences our health. However, this interaction is complex and works differently for each person. The Computational Precision Nutrition (CPN) research group at MHH aims to use artificial intelligence to develop a data-driven predictive tool that will enable personalized dietary recommendations and help prevent diet-related diseases in the future.
The realization that nutrition and health are closely linked is not new. As early as 1850, the German philosopher Ludwig Feuerbach essentially stated: “You are what you eat.” Today, nutritional research is seeking to determine exactly how food influences our metabolome—that is, the metabolic processes in our bodies and their interactions. The gut microbiome also plays a crucial role in this process. This is because the community of microorganisms is not only actively involved in digestion but also produces essential vitamins, strengthens the intestinal barrier, and trains the immune system. Exactly how this interaction works is still unclear, as the relationships between diet and health are complex. In addition, lifestyle factors and pre-existing conditions play an important role.
Dr. Mattea Müller, a postdoc in the “Clinical Data Sciences” research division at the Peter L. Reichertz Institute of Medical Informatics (PLRI) at Hannover Medical School (MHH), is addressing this issue. With her research group “Computational Precision Nutrition (CPN),” she aims to better understand the individual differences in metabolic responses to nutrition and their effects on health. The long-term goal is to develop an AI-based predictive tool that will enable physicians to provide personalized dietary recommendations for the preventive care and treatment of their patients. The Federal Ministry of Research, Technology, and Space (BMFTR) is funding the project with 1.8 million euros over five years.
Finding evidence of cause and effect
“Computer-aided approaches are revolutionizing nutritional science by integrating data from wearables such as smartwatches or fitness trackers and from digital health platforms with multi-omics technologies, which provide insights into genes, their activity, proteins, and metabolic processes in our bodies,” says Dr. Müller. “Traditional statistical models, on the other hand, cannot adequately capture the temporal and individual variations inherent in such data.” This is because the data itself is not uniform. For example, laboratory values from blood tests can yield different results and are generally difficult to compare with one another due to non-identical testing procedures and reference ranges.
The project focuses on developing computer-based models that harmonize the data and highlight the interactions between diet, the microbiome, and the metabolome, as well as their significance for health. “We want to use AI-based methods to derive scientifically verifiable cause-and-effect relationships that can explain why, for example, a certain diet has a positive effect on one person but not on another,” explains the molecular biologist and data scientist.
Not all apples are the same
One of the research group’s goals is therefore not only to understand but also to predict individual metabolic responses to diet. The influence of gender and any pre-existing conditions will also be taken into account. To do this, the researchers must first collect data from as many people as possible to train the AI. This data comes from databases in Germany, the United Kingdom, and the Netherlands. The challenge lies in the diversity—not only among people but also in the food itself. “One apple isn’t necessarily the same as another; there are already enormous differences in nutrient composition between varieties and growing regions worldwide,” the molecular biologist points out.
Interactive platform for predictions and recommendations
Ultimately, an interactive platform for precision nutrition—the CPN-Map—will be made available as a user-friendly tool, initially for research purposes. It remains to be seen whether the analyses will consistently deliver reliable predictions. If so, the CPN-Map could move into clinical use. With individually tailored, personalized nutritional recommendations, it aims to help improve patients’ metabolic health and prevent diseases such as obesity, type 2 diabetes, and neurodegenerative disorders.
Peter L. Reichertz Institute of medical informatics
The Peter L. Reichertz Institute of Medical Informatics (PLRI) is a joint institute of MHH and the Technical University of Braunschweig. Also participating in the project “Computational Precision Nutrition: Mapping the Interactions Between Diet, Microbiome, and Metabolome in Health and Disease (CPN-Map)” are Leibniz University Hannover and the Lower Saxony Centre for AI and Causal Methods in Medicine (CAIMed), as well as the University of Kiel and Maastricht University.
Text: Kirsten Pötzke