ESPE Yearbook of Paediatric Endocrinology

Nat Med 2025. Jun 5. Online ahead of print. PMID: 40473952 doi: 10.1038/s41591-025-03730-7

Brief summary: This study developed an artificial intelligence–driven, microRNA-based multi-context dynamic risk score (DRS) using 5,983 samples from people with and without type 1 diabetes (T1D) from 7 countries. This DRS was validated for its ability to enable T1D discrimination, correct misdiagnoses, and predict therapeutic responses, thereby establishing a personalized, adaptive framework for T1D risk monitoring and intervention.

Identifying individuals at high risk of T1D and how they progress over time is crucial to guide risk-based monitoring and treatment strategies (1). Current T1D screening relies on autoantibodies and/or genetic risk scores (GRS), but these have limitations. Autoantibodies clearly identify individuals with early stages of T1D but they mark only the initial disease stage and do not provide information on when progression will occur (1). GRS offers static risk prediction, despite T1D risk being influenced over time by environmental and therapeutic factors. Thus, there is a need for a dynamic risk score (DRS).

These researchers developed a DRS for T1D based on circulating microRNAs, which are known to reflect environmental influences. Using data from diverse, multicenter, multiethnic, and multinational ("multicontext") cohorts, they identified 50 microRNAs associated with functional β-cell loss through both experimental and computational analyses. These microRNAs were measured in 2,204 individuals across four distinct contexts (Australia, Denmark, Hong Kong SAR China, and India), resulting in a four-context miRNA-based DRS that effectively distinguished individuals with and without T1D. Generative artificial intelligence was then used to enhance this model, achieving strong predictive accuracy (AUC = 0.84) in an independent validation cohort of 662 participants. The DRS also accurately predicted exogenous insulin requirement within one hour of islet transplantation. Additionally, in a clinical trial of imatinib, the baseline miRNA signature, but not traditional clinical metrics, successfully differentiated responders from non-responders at one year.

This study introduced a microRNA-based DRS capable of distinguishing T1D, correcting misdiagnoses (T1D vs Type 2 diabetes), and predicting therapeutic outcomes. This DRS offers a promising tool for longitudinal risk screening and more precise management of individuals with T1D. There is now a need for its validation and refinement by including additional biomarkers (e.g. T1D antibodies) to improve its predictive performance.

Reference: 1. Joglekar MV, Kaur S, Pociot F, Hardikar AA. Prediction of progression to type 1 diabetes with dynamic biomarkers and risk scores. Lancet Diabetes Endocrinol 2024;12(7):483-492.