Scientists at the University of California San Diego have created a powerful new machine learning tool that significantly improves the ability to predict who may develop Type 1 diabetes. The advance could help catch the autoimmune condition earlier in both children and adults, opening the door to monitoring and therapies that might delay or prevent its onset.
Type 1 diabetes occurs when the immune system destroys insulin-producing beta cells in the pancreas, leaving patients dependent on lifelong insulin therapy. While genetics play a major role, previous risk scores worked best only for those with the most obvious high-risk gene variants. Many others slipped through the cracks.
The new tool, called T1GRS, changes that. Researchers analyzed genomes from more than 20,000 people of European ancestry with Type 1 diabetes and nearly 800,000 without the disease. They confirmed known risk sites and uncovered additional variants involved in immune function, gene regulation and blood sugar control. Special attention went to the major histocompatibility complex region on chromosome 6, where the strongest genetic signals reside.
Emily Griffin, a postdoctoral fellow involved in the study, explained the importance of these genetic blocks. “The MHC has blocks of co-inherited genetic information that are very highly enriched in individuals with Type 1 diabetes. If you have them, it does not mean you are going to get diabetes, but if you do not have them, it means you have a very low chance of getting diabetes.”
T1GRS goes beyond simple counting of risk variants. It accounts for complex, non-linear interactions among 199 genetic markers across the genome. The result is a score that performs well not just for the highest-risk group but across a wider range of genetic profiles.
Co-first author TJ Sears noted the improvement. “We were able to identify individuals who get diabetes but do not have known high-risk genetic regions at a much higher rate than the previous diagnostic.”
The model also revealed four distinct genetic subtypes of Type 1 diabetes, each with different ages of onset and complication risks:
- MHC-driven: Strongest link to classic high-risk variants, often appears in early childhood.
- MHC-enriched: Mix of MHC and other variants, with somewhat later onset.
- “T-cell enriched”: Driven by immune response genes outside the MHC region.
- Pancreas-enriched: Variants that directly affect beta cells cause this form, which is associated with a later onset but more frequent serious complications, including kidney disease, nerve damage and heart problems.
When tested on independent datasets, including the NIH All of Us Research Program, T1GRS maintained strong predictive power at 87 percent accuracy and worked reasonably well even in non-European populations.
Carolyn McGrail, another co-first author, highlighted the clinical promise. “Genetic risk scoring allows us to capture a broader pool of both children and adults who are at high risk for Type 1 diabetes but who might otherwise be missed. This supports close monitoring to reduce the risk of complications such as diabetic ketoacidosis at diagnosis and helps identify individuals eligible for preventative therapies like teplizumab.”
The study, published April 30, 2026, in Nature Genetics, was led by senior authors Kyle J. Gaulton and Hannah Carter at UC San Diego School of Medicine. It received funding from the Winkler Endowed Chair in Type 1 Diabetes and the Mark Foundation for Cancer Research.
Experts see T1GRS as a step toward routine genetic screening that could transform how families and doctors approach this serious condition. If we could better predict risk, more people could receive interventions before they lose insulin production.
With continued research, it is hoped that tools like this will help achieve truly personalized strategies in Type 1 diabetes care, to alleviate the burden on patients and improve long-term outcomes. For more information on the study, visit the UC San Diego news site or the full paper in Nature Genetics.
