Super-Cross-Reactive Recognition Enables Preclinical Diagnosis of Alzheimer's Disease

Early diagnosis of Alzheimer's disease (AD) is impeded by the subtle, heterogeneous nature of fluid biomarkers during the disease's initial phases. Current detection tools, which target single biomarkers, often lack the necessary sensitivity and specificity for presymptomatic screening. To address this, we develop a machine learning-informed supramolecular sensor array that leverages a holistic design to overcome the limitations of "lock-and-key" recognition. By engineering macrocyclic receptors with semiselective properties, we generate a competitive signal transduction process characterized by multiple equilibria and nonlinear responses. This design confers high-dimensional cross-reactivity, delineating distinct fluorescent fingerprints for specific AD stages. The array achieves 100% diagnostic accuracy in cerebrospinal fluid from 1-month-old rat models─predating conventional ELISA detection by 3 months and behavioral symptoms by 4 months. Crucially, the array identifies human patients at the subjective cognitive decline stage with >95% accuracy. This work presents a clinically viable tool for early AD diagnosis and establishes a generalizable, modular framework for multiplexed biomarker analysis in neurodegenerative diseases.