New Genetic Tool May Transform Early Detection and Classification of Type 1 Diabetes
A major new study published in Nature Genetics has leveraged machine learning to create a powerful new genetic risk score for type 1 diabetes, potentially improving early prediction and identifying distinct patient subgroups with different clinical outcomes.
Key Findings at a Glance
- A genome-wide association study was conducted in 20,355 individuals with type 1 diabetes and 797,363 non-diabetic Europeans, with further analysis around the MHC region including 10,107 diabetic and 19,639 non-diabetic individuals.
- The machine learning model, T1GRS was trained using identified risk variants and captures nonlinear interactions between genetic variants, including interactions between MHC and non-MHC loci.
- T1GRS showed 89% sensitivity and 84% specificity for type 1 diabetes at an optimal threshold in the discovery dataset.
- The model identified 79 known loci and 8 previously unreported loci associated with type 1 diabetes.
Four Distinct Genetic Subtypes of Diabetes
Based on genetic risk scores, diabetic individuals were categorized into four distinct subtypes, each with unique genetic drivers and clinical implications:
- T cell-enriched: Genetic variants enriched for T cell-related signals.
- MHC-enriched: Variants mainly within the MHC region.
- Pancreas-enriched: Variants enriched for pancreatic cell-related signals, associated with late-onset disease and higher rates of complications.
- MHC-driven: Variants predominantly in the MHC region.
Clinical Implications
The researchers suggest that T1GRS could improve early prediction of type 1 diabetes risk across diverse populations.
The identified genetic subgroups may help guide clinical practice by highlighting differences in age of onset and complication risk, offering a path toward more personalized management and monitoring of the disease.