UC San Diego Researchers Launch Machine Learning Tool to Predict Type 1 Diabetes Risk Early

Researchers at UC San Diego develop T1GRS, a machine learning model that accurately identifies Type 1 diabetes risk to enable earlier medical intervention.

By: AXL Media

Published: May 1, 2026, 7:25 AM EDT

Source: Information for this report was sourced from EurekAlert

A Novel Computational Path to Early Diabetes Detection

In a study published on April 30, 2026, researchers at the University of California San Diego introduced a machine learning model capable of predicting Type 1 diabetes risk with unprecedented precision. The tool, designated as T1GRS, represents a significant shift from traditional genetic risk scores that often rely on a narrow set of high-risk variants. By examining the non-linear interactions between 199 different genetic risk factors, the model allows for the identification of vulnerable individuals, both children and adults, well before the immune system begins its irreversible attack on insulin-producing cells.

Expanding the Genetic Map of Autoimmune Susceptibility

The development of the T1GRS model was underpinned by a massive genomic analysis involving over 20,000 individuals with Type 1 diabetes and approximately 800,000 healthy controls. Through this extensive dataset, the research team verified 79 known genetic loci and uncovered 13 new locations on the genome associated with immune regulation and blood sugar control. According to Emily Griffin, a postdoctoral fellow at UC San Diego, the major histocompatibility complex on chromosome 6 remains the primary area of interest, containing specific blocks of genetic information that are highly concentrated in those who develop the condition.

Bridging the Gap for Complex Genetic Profiles

One of the primary advantages of the T1GRS system is its ability to provide accurate risk assessments for individuals who lack the most common high-risk genetic markers. While previous diagnostic tools were effective for the most obvious cases, they often failed to capture those with more nuanced or complex genetic backgrounds. TJ Sears, a postdoctoral fellow involved in the research, noted that the team was able to identify individuals who develop diabetes but do not possess the well-known high-risk regions at a much higher rate than any previous diagnostic method, effectively widening the net for clinical screening.