Six prediabetes clusters
A recent ML study classified prediabetes into six clusters and reported 92% accuracy in identifying them — three clusters were tagged as high‑risk for progression to type 2 diabetes, paving the way for cluster‑specific prevention strategies study post. The result frames prediabetes as heterogeneous, not one-size-fits-all.
The paper appeared) on Jan 17, 2026 in Biomarker Research. Authors profiled) DNA methylation in peripheral blood, reporting a discovery dataset of n=187 and an independent replication cohort of n=146 that underpinned the signature of 1,557 CpG sites. The author list names Amandeep Singh with senior contributors Andreas L. Birkenfeld and Annette Schürmann, and affiliations at the University of Tübingen, Helmholtz Center Munich and the German Center for Diabetes Research (DZD). (lifescience.net) The work explicitly builds on the Tübingen phenotyping program that used MRI, metabolic testing and genetics in roughly 899 deeply phenotyped participants, according to the Tübingen team. (dzd-ev.de) Prior Tübingen analyses linked the higher‑risk phenotypes to specific mechanisms—low beta‑cell function, pronounced insulin resistance often with elevated liver fat, and compensatory high insulin secretion—providing the biological rationale the epigenetic model targets. (conexiant.com) The paper states many of the identified CpG sites overlap) previous epigenome‑wide signals, and the authors call) for broader population‑level validation before the signature is translated into routine clinical screening.
