Machine learning-assisted nanozyme-coded sensor array for recognition and quantification of biothiols

Abstract: Biothiols play an essential role in antioxidant defense and the maintenance of normal cell function, and they are also biomarkers of many diseases. It is of great significance to establish an effective, reliable and simple method to accurately distinguish and quantify biothiols. Here, we design a quasi - amorphous material Mn_x(DTPMP) which exhibits excellent oxidase-like activity to catalyze the oxidation of colorless 3, 3', 5, 5' - tetramethylbenzidine (TMB) to blue oxTMB for the machine learning-assisted nanozyme-coded recognition and quantification of biothiols. The addition of biothiols inhibits the conversion process of TMB, leading to time-resolved reduction in the color signal of oxTMB. Due to their different abilities to inhibit oxidation over time, specific fingerprints can be drawn for each target. On this basis, a unified stepwise prediction model is established using pattern recognition and classification and regression algorithms in support vector machine (SVM), enabling qualitative identification as well as the precise determination of biothiols simultaneously. The time-resolved nanozyme-coded pattern recognition can not only differentiate cancer cells from normal ones according to intracellular glutathione (GSH), but also evaluate the severity of disease according to serum homocysteine (Hcy), showing a promising application prospect in disease diagnosis.