UCSD - University of California - San Diego

04/30/2026 | Press release | Distributed by Public on 04/30/2026 08:23

Predicting Genetic Risk for Type 1 Diabetes Just Got More Accurate Thanks to UC San Diego Study

Published Date

April 30, 2026

Article Content

Key Takeaways

  • A new machine learning tool called T1GRS accurately identifies people at high risk for developing Type 1 diabetes by analyzing complex interactions between many different genes.
  • Though it was built using data from people of European descent, T1GRS is accurate at predicting risk for non-European populations as well.
  • Four genetic sub-types of Type 1 diabetes emerged, each with different patterns of disease onset and complications that could inform personalized treatment of the condition.

In people with Type 1 diabetes (T1D), the immune system shuts down the body's ability to make the hormone insulin, responsible for regulating blood sugar and providing cells with glucose to produce energy. As a result, they are dependent on external sources of the hormone for the rest of their lives. Predicting who will develop T1D remains difficult, as existing genetic risk scores are largely limited to individuals with well-known high-risk variants.

Now, researchers at University of California San Diego and their colleagues have developed a machine learning model called T1GRS that analyzes complex interactions between many different genes to calculate more accurate genetic risk scores for a broader population. The study demonstrates that this tool can identify children and adults at high risk for T1D earlier than current methods, enabling the use of preventive measures before the disease fully develops. The study was published on April 30, 2026 in Nature Genetics.

Predicting genetic risk

The researchers combed through a dataset of genomes from more than 20,000 people of European ancestry with T1D - and nearly 800,000 without the autoimmune disorder - for disease-associated genetic variants. They confirmed known risk variants at 79 loci (physical locations of genes on chromosomes) and at 13 loci not previously linked to the condition that are involved with gene regulation, immune function, and blood sugar control.

The team also mapped out specific genetic variants in the major histocompatibility complex (MHC), a region on chromosome 6 that contains the strongest known genetic associations with T1D. By analyzing data from more than 29,000 people, they uncovered several novel variants associated with T1D that influence immune function and gene activation.

"The MHC has "blocks" of co-inherited genetic information that are very highly enriched in individuals with Type 1 diabetes," said co-first author Emily Griffin, PhD, a postdoctoral fellow in the lab of Kyle J. Gaulton, PhD, associate professor of pediatrics at UC San Diego School of Medicine. "If you have them, it doesn't mean that you're going to get diabetes, but if you don't have them, it means you have a very low chance of getting diabetes."

The researchers then developed a machine learning model, called T1GRS, which incorporates non-linear interactions between 199 of the risk variants they identified across the genome and within the MHC region. The model calculates a person's risk of having T1D, their T1GRS score, with improved accuracy.

Previous tools for predicting genetic risk for T1D are most accurate in the highest risk individuals, according to co-first author TJ Sears, PhD, a postdoctoral fellow and former graduate student in the lab of Hannah Carter, PhD, associate professor of medicine at UC San Diego School of Medicine. In contrast, T1GRS had high accuracy across a larger set of individuals, including those with more complex genetic risk.

"We were able to identify individuals who get diabetes but don't have known high-risk genetic regions at a much higher rate than the previous diagnostic." TJ Sears, PhD, postdoctoral fellow at UC San Diego School of Medicine

"We were able to identify individuals who get diabetes but don't have known high-risk genetic regions at a much higher rate than the previous diagnostic," said Sears.

Four Type 1 diabetes sub-types

An analysis of genetic features most strongly influencing each person's T1GRS score enabled the grouping of individuals with T1D into four sub-types, each with unique clinical profiles and outcomes:

  • MHC-driven group: Primarily characterized by well-known high-risk genetic variants of T1D. These individuals typically experience early onset in childhood.
  • MHC-enriched group: Influenced by a mix of genetic variants both within and outside the MHC region. Disease onset is slightly later than the MHC-driven group, with an intermediate pattern of disease severity.
  • T-cell-enriched group: Driven largely by non-MHC variants that affect the adaptive immune response. These individuals also tend to have an intermediate age of onset.
  • Pancreas-enriched group: Primarily influenced by non-MHC gene variants that impact pancreatic cells, including beta cells which produce and secrete insulin. Despite having a later age of onset, individuals in this group face the highest rates of complications, including kidney disease, nerve damage and heart problems.

To test whether the T1GRS model could predict risk outside of the dataset it was trained on, the team applied it to genetic datasets from the National Institutes of Health All of Us Research Program and the National Pancreatic Organ Donor (nPOD) biobank. While the accuracy of the model was reduced in this smaller sample size, it still predicted risk with 87% accuracy.

"We were able to do a really good job of predicting risk in non-European populations as well, even though T1GRS was developed in individuals of European descent," said Griffin.

What's more, individuals with T1D from these independent datasets were classified into the same four genetic sub-types identified in the original dataset, reinforcing the clinical relevance and generalizability of these groupings.

"We were able to do a really good job of predicting risk in non-European populations as well, even though T1GRS was developed in individuals of European descent," said Griffin. Emily Griffin, PhD, postdoctoral fellow at UC San Diego School of Medicine

Better screening and earlier treatment

The findings reinforce the potential for T1GRS as a widespread clinical screening tool, facilitating earlier detection and eventually personalized treatment for T1D.

"Genetic risk scoring allows us to capture a broader pool of both children and adults who are at high risk for T1D but who might otherwise be missed," said co-first author Carolyn McGrail, PhD, a former graduate student in Gaulton's lab, now a senior associate consultant at L.E.K. Consulting. "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."

Additional co-authors on the study include: Alexandra L. Ghaben and Parul Kudtarkar at UC San Diego; Patrick Smadbeck at Broad Institute; Jason Flannick at Harvard Medical School, Broad Institute and Boston Children's Hospital.

The study was funded, in part, by the Winkler Endowed Chair in Type 1 Diabetes to Kyle J. Gaulton and the Mark Foundation for Cancer Research to Hannah Carter.

Disclosures: Kyle J. Gaulton has done consulting for Genentech, received honoraria from Pfizer, and is a shareholder of Neurocrine biosciences. Carolyn McGrail, Timothy J. Sears, Hannah Carter, and Kyle J. Gaulton have a pending patent 19/281,499 related to machine learning methods developed for this study. The remaining authors declare no competing interests.

UCSD - University of California - San Diego published this content on April 30, 2026, and is solely responsible for the information contained herein. Distributed via Public Technologies (PUBT), unedited and unaltered, on April 30, 2026 at 14:24 UTC. If you believe the information included in the content is inaccurate or outdated and requires editing or removal, please contact us at [email protected]