Multiplex Cerebrospinal Fluid Proteomics Identifies Biomarkers Predicting Neuropsychiatric Symptom Progression in...

Neuropsychiatric symptoms (NPS), commonly concomitant with Alzheimer's disease (AD), substantially impair the quality of life and accelerate disease progression, yet reliable biomarkers for early identification of individuals at high NPS risk remain elusive. In this study, we leveraged the data from the Alzheimer's Disease Neuroimaging Initiative (ADNI), incorporating longitudinal data from 509 participants diagnosed with mild cognitive impairment (MCI) or mild AD at baseline and followed up for 1 and 2 years. The dataset included cerebrospinal fluid (CSF) proteomic profiles comprising 6361 proteins, along with comprehensive data on NPS diagnosis, cognitive function, and AD pathology. LASSO regression and recursive feature elimination were applied to identify NPS-related CSF proteins, followed by random forest modeling to predict NPS risk at baseline, 1 year, and 2 years. Incorporating selected CSF proteins significantly improved NPS prediction compared to the reference model, with AUCs increasing from 0.64 to 0.76 at baseline, 0.63 to 0.80 at 1 year, and 0.63 to 0.81 at 2 years. Notably, Cyclin-Dependent Kinase-Like 2 (CDKL2), Nidogen 2 (NID2), and Lin-7 Homolog B (LIN7B) were consistently associated with NPS across all time points. Among them, CDKL2 and NID2 were significantly associated with AD biomarkers and cognitive scores, and their expression changes were independently validated in cerebrospinal fluid from a mouse model, highlighting their potential as stable predictive biomarkers. Our findings highlight CSF proteomic signatures that robustly predict NPS progression in individuals with MCI and mild AD, offering a framework for early risk stratification and precision intervention in NPS.