Machine learning-assisted high-performance immunoSERS platform using silk fibroin as a natural etching mask for early diagnosis of Alzheimer's disease

Abstract

Early and accurate diagnosis of Alzheimer's disease (AD) is a major stride toward pharmacological interventions to delay the onset or progression of the disease in patients with mild symptoms. In this study, we developed a silk fibroin-templated surface-enhanced Raman spectroscopy (SERS)-activated double-sandwich immunoassay (immunoSERS) platform that enhances plasmonic hotspot formation for the ultrasensitive detection of biomarkers. Silk fibroin, acting as a natural etching mask, facilitates the direct fabrication of Au nanocavity (AuNC) substrates and enables the immunoSERS platform to achieve attomolar-level detection (limit of detection: 35.8 aM) with high reproducibility (relative standard deviation: ∼2.5 %) due to its unique structural characteristics. This platform effectively detects four core AD biomarkers—amyloid beta 42 (Aβ42), total tau (t-tau), phosphorylated tau (p-tau), and brain-derived neurotrophic factor (BDNF)—in human plasma. Moreover, by introducing a k-nearest neighbors (KNN)-based machine learning algorithm, the suggested platform could classify disease progression stages with 94.0 % accuracy. These results indicate that this silk fibroin-driven immunoSERS platform is a viable alternative to existing diagnostic techniques for the effective early screening of AD and are a potential therapy to delay AD incidence in clinical practice. © 2025 Elsevier B.V.