Deep Learning-Powered Plasmonic Platform for Decoding Dynamic Protein Aggregation Landscapes in Parkinson's Disease Progression

In-depth characterization of pathological protein aggregate with its subcomponents is pivotal for early diagnosis and staging of neurodegenerative disorders. Here, we introduce the Protein Aggregate NanoDynamics Analyzer (PANDA), a plasmonic nanotechnology-powered system that translates the supramolecular size distribution of protein aggregates into distinct scattering signatures via sterically constrained immunogold clustering. In this work, PANDA enabled the profiling of α-synuclein (α-syn) aggregates with diverse supramolecular architectures from human serum samples and experimentally revealed biological stage-dependent distribution patterns in Parkinson's disease (PD). Aggregate size profiles from PANDA readouts showed strong correlation with clinical scores (r_max = 0.6) and dopaminergic PET imaging (r_max = 0.7). Notably, it also confirmed the pathophysiological transition of "oligomer-to-fibril" in the course of PD progression. To leverage this transition, we incorporated a deep neural network (DNN) to classify PD stages. The network achieved high accuracy and enabled an objective reference to evaluate PD progression. PANDA system thus offers a noninvasive, artificial intelligence (AI)-augmented framework for molecular diagnosis and stratification of neurodegenerative diseases such as PD.