Synergistic Integration of Enzyme-Free Nanocatalysts and Machine Learning for Economical Glucose Estimation in Human Plasma

Abstract The use of gold nanoparticles (Au NPs) as a glucose oxidase-mimicking nanozyme is highly promising owing to their unique catalytic properties and stability. However, the low peroxidase activity of the Au NPs posed a limitation for translation research, as subsequent reactions reduced oxygen (O2) to H2O2. Herein, a method is developed in which Ag+ ions serve as the terminal electron acceptor and, depending on glucose concentration, are reduced to Ag NPs, which are used as a colorimetric probe for estimating plasma glucose levels. The GOD-mimicking activity of the Au NPs was evident from our experimental results, which followed Michaelis−Menten kinetics. As a proof of principle, 195 patient samples were analyzed and compared with an enzymatic pathological test (GOD/POD systems) and further classified using ML, demonstrating that the liquid-state format yielded better results, delivering highly accurate quantitative glucose estimation comparable to the standard pathological test. Likewise, the method was substantiated for solid-state glucose estimation (n = 60), which delivered modest quantitative precision in estimating blood glucose levels.