Machine Learning Algorithms Enabled Visual On-Site Intelligent Sensing of Bioactive Components by Histidine-Functionalized Nanozyme Sensor Array

The identification of bioactive components (BCs) in food is essential for understanding their biological activities and the health advantages they may provide. Although nanozyme sensor arrays have been developed and used for the detection of BCs, the performance of certain nanozymes remains less than ideal in practical settings. Therefore, developing nanozymes with higher activity is important for enhancing the sensing performance of sensor arrays and broadening their use in biosensing applications. Here, we developed a histidine-functionalized trimesic acid-copper (TA-Cu-His) nanozyme using a defect-engineering approach. Relative to the original TA-Cu material, the resulting TA-Cu-His nanozyme shows enhanced laccase-like (LAC) activity. Because its catalytic activity varies markedly across different pH conditions, we employed a Recursive Feature Elimination (RFE) strategy to screen for the optimal pH sensing channels and further applied them for the identification of various BCs. In addition, integrating various machine learning (ML) algorithms with the sensor array improved the precision of the concentration-irrelevant classification system from 58.62 to 100%, thereby enabling the recognition of blind samples. Finally, we further developed an ML-assisted intelligent sensing platform using a Residual Network 50 (ResNet-50) neural network model to improve the practical application of BC identification. Overall, this work offers a new strategy for the intelligent and efficient identification of multiple BCs.