0
0Researchers recently utilized transformer-based deep learning and unsupervised clustering to analyze longitudinal health records from over 100,000 patients in the UK. By examining decades of prediagnostic medical data, the study successfully identified five distinct subtypes for both Alzheimer’s disease and Parkinson’s disease. These categories are defined by unique combinations of comorbidities, genetic profiles, and symptom trajectories, such as metabolic-inflammatory or vascular-psychiatric phenotypes. The findings reveal that certain patient groups face significantly higher mortality and hospitalization rates than others, regardless of their genetic predisposition. This scalable framework demonstrates how routinely collected electronic health records can provide a biologically informed roadmap for early diagnosis. Ultimately, the study advocates for a precision medicine approach to neurodegenerative diseases through targeted interventions tailored to specific clinical subtypes.
References:
Lian J, Fan Z, Petrazzini B O, et al. Subtyping Alzheimer’s disease and Parkinson’s disease using longitudinal electronic health records[J]. Nature Aging, 2026: 1-14.
