MBZUAI uses AI to detect diabetes risk before symptoms

READY
Read by Emma, the Simply Abu Dhabi narrator.
A single fasting glucose test shows normal. Continuous monitoring over two weeks reveals the truth: glucose levels drift into the prediabetic range. This gap between a single measurement and the patterns hidden within it forms the foundation of research at Mohamed bin Zayed University of Artificial Intelligence, where scientists are developing systems to catch disease before it announces itself through conventional screening.
The Human Phenotype Project, a longitudinal study operating in Abu Dhabi, follows participants for up to 25 years. It assembles more than 40 layers of health data from each person: genetic information, blood biomarkers, medical imaging, continuous glucose monitoring and sleep tracking. Open to UAE citizens and residents aged 18 to 70, it draws on the country’s population diversity to establish health reference standards that account for both age and ethnicity. One finding emerged with particular clarity: 40 percent of individuals whose fasting glucose appeared normal subsequently recorded readings within the prediabetes range during continuous monitoring.
At the centre of this work sits the Human Phenotype Project, a longitudinal study operating in Abu Dhabi that follows participants for up to 25 years. The project assembles more than 40 layers of health data from each person, combining genetic information, blood biomarkers, medical imaging, continuous glucose monitoring and sleep tracking. Open to UAE citizens and residents aged 18 to 70, it draws on the country’s population of more than 200 nationalities to establish health reference standards that account for both age and ethnicity. The result is a dataset of unprecedented depth and demographic breadth.
One finding emerged with particular clarity: 40 percent of individuals whose fasting glucose levels initially appeared normal subsequently recorded readings within the prediabetes range during continuous monitoring. These are people conventional screening methods would have cleared. This gap between single measurements and pattern analysis forms the foundation of MBZUAI’s approach.
The university has developed GluFormer, an AI foundation model that analyses daily glucose fluctuations rather than relying solely on average measurements. Published in Nature in January 2026 and trained on more than 10 million continuous glucose monitoring measurements from over 10,800 participants, GluFormer can identify meaningful glucose patterns using just two weeks of data. A multimodal version can also simulate how an individual’s body responds to specific foods, offering insight into personal metabolic health.
A five-year research project running from 2025 to 2030 examines polygenic risk scores, which combine multiple genetic variants to estimate inherited disease risk. Using the Emirati Genome Program’s repository of 900,000 genomes alongside diabetes diagnoses recorded through Malaffi, Abu Dhabi’s health information exchange, researchers aim to develop a population-specific score for Emiratis. Existing models were largely developed using European genetic data, a limitation that shapes their accuracy across different populations.
Work led by Professor Imran Razzak, Associate Professor of Computational Biology at MBZUAI, has uncovered links between oral microbes and the metabolic changes associated with diabetes. Using Human Phenotype Project data, the team identified bacterial and microbial gene patterns associated with glucose levels, liver health and body fat. AI models using selected oral microbial characteristics distinguished individuals with prediabetes more accurately than models based on complete microbial profiles.
The retina provides a non-invasive view of blood vessels and nerve tissue. MBZUAI researchers are developing AI systems capable of detecting signs of disease in retinal images. During a research showcase with Cleveland Clinic Abu Dhabi in October 2025, the university demonstrated how eye scans could help identify early indicators associated with diabetes, hypertension, Alzheimer’s disease and heart disease. Professor Jianing Qiu, Assistant Professor of Personalised Medicine at MBZUAI, leads research exploring how retinal images reflect biomarkers of metabolic and vascular dysfunction. Related work, led by Professor Razzak, is developing AI systems combining retinal vascular imaging with electrocardiogram data to identify early signs of heart failure.
These systems are designed to support clinicians rather than replace them, providing additional information that could help detect previously overlooked warning signs. The university has become a focal point for research addressing the underrepresentation of non-European populations in genetic studies, work that directly shapes the accuracy of tools like polygenic risk scores.
What it means for Abu Dhabi
The UAE’s investment in healthcare technology is advancing a central objective: catching disease before it declares itself. The initiatives underpinning MBZUAI’s research, including the Emirati Genome Program and Malaffi, are strengthening the data foundations needed to identify risk earlier and tailor prevention to the individual. Diabetes affects millions globally, and early intervention changes outcomes. Abu Dhabi possesses both the research capacity and the population diversity to develop tools that work across different genetic backgrounds.
These advances align with the UAE National AI Strategy 2031, which recognises healthcare as a priority sector. The work reflects a broader commitment to positioning the emirates as a hub for precision medicine and artificial intelligence research, drawing talent and partnerships while building systems that serve the local population first.
Source: WAM