Scientists have developed a new artificial intelligence-based model capable of predicting the risk of developing type 1 diabetes more accurately by analyzing genetic data. The study was conducted by researchers at University of California San Diego and published in the journal Nature Genetics.

The key idea behind the research is that traditional methods of assessing type 1 diabetes risk are often not precise enough, especially in people who do not carry the classic “high-risk” genetic markers.

To address this, the researchers used machine learning to create a model called T1GRS, which analyzes complex interactions between hundreds of genetic variants.

The model was trained using data from a large-scale genetic analysis involving more than 20,000 patients with type 1 diabetes and nearly 800,000 individuals of European ancestry without the disease. Additional datasets focusing on the MHC region — one of the most important parts of the genome linked to the immune system — were also included.

In total, the researchers identified around 199 genetic risk variants, including previously unknown loci associated with immune function and the regulation of gene activity. Using this information, the model was trained to estimate an individual’s genetic risk of developing the disease.

The findings showed that T1GRS significantly improves risk classification accuracy compared with previous methods. The model not only distinguishes more effectively between people with and without diabetes, but also identifies hidden gene combinations that had not previously been considered in clinical practice.

One of the most notable findings was that the algorithm made it possible to identify several genetic subtypes of type 1 diabetes. These subtypes differ in the age at which the disease begins and in the likelihood of complications: in some patients the disease develops during childhood, while in others it appears later in life, with different immune response mechanisms involved.

The researchers say that this level of detail could eventually help advance personalized medicine — allowing doctors not only to predict diabetes risk, but also to better understand how the disease may progress in each individual patient.