Machine-Learning-Assisted Impedance Component Analysis Enables Standardizable Surface Protein Analysis of Extracellular Vesicles Using Engineered Nanovesicles.

Characterizing surface protein heterogeneity on extracellular vesicles remains challenging but essential for understanding their biological functions and clinical applications. Here, this study introduces an integrated platform that combines engineered cell-derived nanovesicles, used as model standards with controlled surface protein states, and machine learning-optimized impedance spectroscopy. Cell-derived nanovesicles with defined surface protein densities are generated by extruding HeLa cells expressing 1, 3, or 9 copies of amyloid-β 42, establishing a series of reference vesicles with precisely controlled oligomeric configurations. Through systematic evaluation of impedance features across frequencies from 10 Hz to 1 MHz, machine learning identifies reactance changes at 1 kHz as optimal for distinguishing oligomeric states on the vesicle membrane. Equivalent-circuit modeling reveals that membrane capacitance correlates with protein oligomerization, and structural predictions explain the mechanistic basis. The platform also enables label-free, time-resolved analysis of Aβ oligomer formation directly on vesicular membranes, providing insights into aggregation dynamics. This platform establishes standardizable reference materials using engineered nanovesicles and a quantitative framework for extracellular vesicle surface protein analysis, offering a broadly applicable method for studying membrane-associated protein dynamics relevant to neurodegenerative diseases and advancing extracellular vesicle-based diagnostics.