Bioinspired Iron Porphyrin Covalent Organic Frameworks-Based Nanozymes Sensor Array: Machine Learning-Assisted Identification and Detection of Thiols

Given the crucial role of thiols in maintaining normal physiological functions, it is essential to establish a high-throughput and sensitive analytical method to identify and quantify various thiols accurately. Inspired by the iron porphyrin active center of natural horseradish peroxidase (HRP), we designed and synthesized two iron porphyrin covalent organic frameworks (Fe-COF-H and Fe-COF-OH) with notable peroxidase-like (POD) activity, capable of catalyzing 3,3',5,5'-tetramethylbenzidine (TMB) into oxidized TMB with three distinct absorption peaks. Based on these, a six-channel nanozyme colorimetric sensor array was constructed, which could map the specific fingerprints of various thiols. Subsequently, machine learning techniques, including supervised learning with linear discriminant analysis (LDA), decision trees (DT) and artificial neural networks (ANN), unsupervised learning with hierarchical cluster analysis (HCA), and ensemble learning with random forests (RF), were used for precise identification of thiols in complex systems, with a detection limit as low as 50 nM. Significantly, the sensor array demonstrated strong potential for practical applications, including analyzing homocysteine (Hcy) in human serum, mercaptoacetic acid (TGA) in depilatory creams, and glutathione (GSH) in cell lysates, thereby showing promise for use in disease diagnosis.