To this point, determination of genetic risk for Type 1 diabetes mostly has been limited to people with well-known and well-documented risk profiles.

But with a machine learning tool created by researchers at UC San Diego in La Jolla and their colleagues, scientists hope they can cast a wider net.

People with Type 1 diabetes are unable to produce insulin, a hormone responsible for regulating blood sugar and supplying cells with glucose for energy. Therefore, they need external sources.

A study released April 30 in the scientific journal Nature Genetics looks beyond existing genetic risk scores and details findings from the researchers’ machine learning model, T1GRS.

The UC San Diego team behind the paper includes co-senior author Kyle Gaulton, an associate professor of pediatrics in the School of Medicine; Emily Griffin, a postdoctoral fellow in Gaulton’s lab; Carolyn McGrail, a former graduate student in Gaulton’s lab and a current senior associate at L.E.K. Consulting; and TJ Sears, a postdoctoral fellow and former graduate student in the lab of co-senior author and School of Medicine associate professor Hannah Carter.

Other collaborators include Alexandra Ghaben and Parul Kudtarkar from UC San Diego; Patrick Smadbeck of the Broad Institute of MIT and Harvard; and Jason Flannick of Harvard Medical School, Broad Institute and Boston Children’s Hospital.

T1GRS moves beyond known high-risk variants to evaluate interactions among genes. As a result, scientists can better assess genetic risk scores beyond the more obvious cases.

“Those previous studies were mostly focused on the highest-risk information that captures a subset of individuals that are going to develop Type 1 diabetes,” Gaulton said. “Part of the innovation of our study is that it captures more complex relationships and allows you to do a better job of predicting whether a broader set of individuals is going to develop Type 1 diabetes.”

The machine learning model was fueled by data from 20,000 Type 1 diabetes patients with European ancestry and roughly 800,000 patients without it. The scientists identified 79 loci, or the location of a gene on a chromosome, that qualify as risk variants. Of that group, 13 were not previously linked with Type 1 diabetes.

Scientists also mapped a region on chromosome 6 most genetically tied to Type 1 diabetes called the major histocompatibility complex, or MHC, with data from 29,000 people. Doing so enabled them to find novel variants linked to Type 1 diabetes that impact gene activation as well as immune function.

People with Type 1 diabetes were placed in four different groups: MHC-driven, MHC-enriched, T-cell-enriched and pancreas-enriched.

“The MHC has ‘blocks’ of co-inherited genetic information that are very highly enriched in individuals with Type 1 diabetes,” Griffin said in a statement. “If you have them, it doesn’t mean you’re going to get diabetes, but if you don’t have them, it means you have a very low chance of getting diabetes.”

Gaulton told the La Jolla Light that the researchers hope to combine genetic data with biomarker data to gain a better understanding of Type 1 diabetes risk. Other goals include avoiding misdiagnoses, properly managing disease, identifying people for clinical trials and making headway on preventive therapies.

“There’s a lot of Type 1 diabetes that’s not covered by that very, very early age and super-high-risk genetic variants,” Gaulton said. “In our tools, it helps expand the set of individuals across the spectrum of disease that you can predict disease for.”

“As new therapies are developed and new clinical trials are created and successful, I think that would help with the eventual development of permanent prevention therapies that people could qualify for,” he added. “I’d say for any disease, earlier prediction that’s accurate … would be important. Type 1 diabetes definitely qualifies for that as well.” ♦