Digital siblings unveil distinct molecular and clinical subtypes in Alzheimer's disease via omics-driven profiling.

Alzheimer's Disease (AD) is a multifactorial neurodegenerative disorder marked by extensive biological and clinical heterogeneity, complicating prognosis and personalized treatment strategies. Because of this, data-driven methods that characterize patient similarity and subgroup-specific molecular signatures are essential for advancing precision medicine in AD. We applied manifold learning techniques to fuse proteomic, demographic, and clinical data from 438 AD patients, enabling the identification of "digital siblings"-patients with closely related molecular and clinical profiles within a learned latent space. This framework enabled robust clustering and stratification of patient subgroups, revealing distinct pathway enrichments associated with clinical traits such as age, alcohol use, and comorbidities. Moreover, structural and network analyses of key protein interactions within these subgroups provided insights into the molecular mechanisms potentially driving disease heterogeneity. While this approach primarily clustered patients based on comprehensive molecular patterns, it lays critical groundwork for developing predictive models that incorporate longitudinal progression and intervention outcomes. Overall, our results underscore the potential of "digital sibling"-based stratification to refine patient subgroup characterization and serve as a foundation for future dynamic modeling in AD.