MACHINE LEARNING EARLY DIAGNOSIS OF ALZHEIMER’S DISEASE WITH SURFACE-ENHANCED RAMAN SCATTERING ON A-BETA SPECIES IN THE CEREBROSPINAL FLUID

Background: disease (AD) is the most prevalent neurodegenerative condition in elderly, where the presence of amyloid-protein (A-Beta) aggregates, forming extracellular amyloid plaques, is one of the primary neuropathological markers of AD. Currently, the NINCDS- ADRDA criteria are used for clinical diagnosis of AD, categorizing the disease as possible or probable but not definite (which still requires neuropathological examinations) [1]. Part of this can be attributed to the fact that the clinical and laboratory biomarkers evaluated are not unique for AD with respect to other neurodegenerative disorders [2]. Material and Methods: In this work we present an innovative approach based on Machine Learning [3] applied to data coming from Surface Enhanced Raman Spectroscopy (SERS) for the analysis of cerebrospinal fluid (CSF) collected from patients with AD or other neurological conditions, combined with the Real-Time Quaking- Induced Conversion technique, a seeding aggregation assays capable to detect and amplify traces of pathological Ab species in the CSF of patients with AD [4]. Preliminary Results: Dimensionality reduction techniques such as t-SNE confirm the discriminating power of our approach, while Support Vector Machine shows performances of 81% for AUROC and 0.78 for accuracy. This work demonstrates the importance of ML for SERS analysis, revealing chemo-structural information to distinguish AD from the other diseases.