Analysis of Antioxidant Capacity Based on Fe-ZIF67 Nanozyme and Machine Learning

Abstract: This study employed a surface coordination modification strategy to introduce Fe2+ onto the surface of ZIF67, using uncoordinated imidazole nitrogen atoms to anchor metal ions, thereby preparing Fe-ZIF67 nanozyme materials. The morphology,structure and elemental composition of the materials were characterised by TEM, XPS and other methods, with XPS confirming that Fe2+ was successfully anchored on the surface of ZIF67, forming new Fe-N active centres. The correlation coefficients between the colour change value (B-B₀)/B₀ and the concentrations of three herbal components (MMP,LBP,DOPs) reached 0.99, demonstrating that this index can be used as a quantitative measure of antioxidant capacity for subsequent analysis. Principal component analysis showed that samples of different types and concentrations displayed clear clustering in the score plots, verifying the distinguishability of the colour signal data. By inputting colour images and the corresponding antioxidant indices into a laboratory-constructed LSTM model for training, the determination coefficient R2for the test set exceeded 0.95, achieving accurate prediction of antioxidant capacity using photographs alone. Applying this model to real tawny wine samples, the resulting antioxidant indices were consistent with the trends of herbal component additions in the samples, validating the feasibility of this method for evaluating actual samples.