Skip to main content
Premium Trial:

Request an Annual Quote

Researchers Develop Type 1 Diabetes Genetic Risk Algorithm for Children, Infants


NEW YORK (GenomeWeb) – An international team led by researchers from two German institutions has developed a test that calculates the genetic risk for developing type 1 diabetes in children. Using the algorithm, the group wants to eventually perform type 1 diabetes primary prevention trials.

In a study published earlier this month in PLOS Medicine, the researchers from Technische Universitat Dresden, Helmholtz Zentrum Munchen, and elsewhere determined genetic risk scores derived from 41 gene loci linked to Type 1 diabetes risk.

Lead author and TU Dresden professor Ezio Bonifacio explained that researchers normally perform genetic risk studies on patients with a family history of type 1 diabetes. Taking a whole blood spot sample from the patient, researchers extract the DNA and search for genes that are often associated with a risk for developing the disease.

According to Bonifacio, however, patients with family histories of type 1 diabetes only represent a small number of diagnosed cases, and patient recruitment takes too long for a proper clinical trial.

The researchers therefore collected blood samples from 4,543 children from The Environmental Determinants of Diabetes in the Young (TEDDY), a prospective cohort study that screened more than 400,000 newborns for high-risk HLA genotypes for type 1 diabetes. As part of the TEDDY study, researchers genotyped single nucleotide polymorphisms (SNPs) using the Illumina ImmunoChip. The patients lacked relatives with type 1 diabetes, but instead had one of the two highest risk HLA genotypes —heterozygous DR3 and DR4-DQ8, or homozygous DR4-DQ8.

Gathering all the genes linked to type 1 diabetes that researchers had previously published, Bonifacio's team developed an algorithm called the Winkler score, named after fellow co-author Christiane Winkler, that calculated the genetic risk score of developing multiple islet autoantibodies and type 1 diabetes in patients based on 41 SNPs.

In addition to using the newly developed Winkler score, Bonifacio also added the Oram genetic risk score, a previously published algorithm developed by senior author Richard Oram of the University of Exeter Medical School and National Institute for Health Research at Exeter in the UK.

"Adding up the scores, we looked at both of the [Oram/Winkler] scores, and one that merged them, and asked 'Okay, does it work when applied to real-world data?'" Bonifacio explained. When the team ran the algorithm on the TEDDY study, Bonifacio noted that "the merged scores actually worked better than the separated genetic scores."

Bonifacio's team found that it could stratify the risk for islet autoantibodies and diabetes in children depending on their genetic score and age. In children with the HLA risk genotypes, the risk for developing multiple islet autoantibodies was about 6 percent by age six, and the risk for diabetes by age 10 was about 4 percent.

In addition, Bonifacio and his colleagues saw that the genetic score was higher in children who developed islet autoantibodies by the time they were six, compared to children who remained islet autoantibody negative. The risk for developing multiple islet autoantibodies was 11 percent in children with a merged genetic score of greater than 14.4, compared to about 4 percent in children with a genetic score of less than or equal to 14.4.

The risk for developing diabetes by age 10 was about eight percent in children with a merged score of greater than 14.4, compared to about 3 percent in children with a score of less than or equal to 14.4.

According to the study, the researchers identified 82 children who developed pre-symptomatic or asymptomatic diabetes by age six. The upper quartile of the merged genetic score was associated with a greater than 10 percent risk for the pre-symptomatic stage of multiple islet autoantibodies.

Bonifacio noted, however, that the majority of people who have an elevated risk score will not develop type 1 diabetes.

"We wanted to identify children who had this high risk score, which we said was more than 10 percent," he explained. "While that is pretty high, we still noted that 90 percent of children who have the HLA genotype will never develop the disease."

According to Bonifacio, the overall test, combined with the team's customized algorithm, can technically produce a genetic risk score for patients in less than a day.

"We usually stack the samples up, which are done in plates and sent off to the lab, and we receive results every couple of weeks or so," he said.

Overall, Bonifacio believes that combining genetic information from multiple risk loci may improve the prediction of diseases such as type 1 diabetes. Along with his colleagues, Bonifacio envisions a genetic testing model that could identify up to 25 percent of future childhood cases of type 1 diabetes without a prior family history from less than a single percent of newborns.

"Potentially, we'd identify between a fourth and a fifth of all children who might have type 1 diabetes," Bonifacio said. "It's a tool to use in a setting where you're going to be doing clinical trials, where you're recruiting people who have a reasonable risk ... of getting diabetes."

The team acknowledged certain limitations including a small population of genetic scores, as they were from individuals of mainly European descent. Bonifacio explained that if he and his colleagues applied the test on a different population, whether it was Hispanic, Black, or Asian, the genetic score would work in the same manner. However, the algorithm would have to change based on the different genetic haplotypes present in the specific populations.

In addition, Bonifacio noted that the genetic risk score would not be effective in people who develop type 1 diabetes later in life, as the group's algorithm is geared toward juvenile diabetes.

Bonifacio's team is currently using the risk score method as part of the Global Platform for the Prevention of Autoimmune Diabetes (GPPAD), a group of type 1 diabetes prevention clinical trials in Europe. Funded by the Leona M. and Harry B. Helmsley Charitable Trust, GPPAD aims to perform genetic testing on over 300,000 patient blood spots samples from patients in Belgium, Germany, Poland, Sweden, and the UKm.

Families of infants with the increased risk will be contacted by a local GPPAD team and informed about the meaning of an elevated risk, educated about the disease's symptoms, and asked to participate in a randomized controlled trial to prevent type 1 diabetes.

As part of the GPPAD screening, the team has partnered with regional providers in Germany, Sweden, Poland, and the UK for blood testing. Bonifacio noted the providers will use LGC's KASP genotyping assays to run the genotyping-based tests at a much cheaper cost than Illumina's Immunochip platform.

According to Bonifacio, the team will also recruit over 1,000 infants and children in the primary prevention trial in order to identify the at-risk children. The group will introduce immune tolerance to insulin to patients in the hope that it will prevent development of type 1 diabetes.

University of Chicago postdoc May Sanyoura, who was not involved with the study, believes the merged Oram/Winkler score could be applied to more than the small European population in the study. Specializing in researching rare forms of type 1 diabetes, Sanyoura previously worked with Oram and applied his genetic risk score to differentiate between people who may have the monogenic form of diabetes in patients who are part of UChicago's diabetic genetic registry.

In addition, Sanyoura ran the Oram genetic risk score, genotyping 10 SNPs on most of the the UChicago registry participants, and then compared the results to the European population. Separating individuals into different percentiles, Sanyoura saw that patients with a very low T1D risk score were positive for a monogenic cause of diabetes on her team's separate gene testing panel.

"We found the [genetic] score to be really efficient and cost effective, but that doesn't mean that if a person has a T1D genetic risk above the 95th percentile, he doesn't have a monogenic cause [of diabetes]," Sanyoura explained. "We want to be selective to produce positive results, eventually identifying the cause of diabetes to help with treatment."