Machine Learning-Powered Tri-mode Nanozyme Sensor Array for the Highly Selective Identification of Quinolone Antibiotics

Abstract

Developing efficient quinolone antibiotics (QNs) detection methods is of great significance for ensuring food safety and public health. In our previous work, a nanozyme sensor array was developed to enable efficient recognition of QNs. However, relying solely on a single colorimetric signal is susceptible to interference from complex matrix backgrounds and cross-contamination. To address this challenge, in this work, we constructed sensing units using colorimetric, photothermal, and fluorescence signals from a nanozyme reaction system. This array can accurately detect and distinguish QNs in the range of 0.1–100 μM. Compared to our previous work, the detection concentration range has been expanded by 10-fold. In addition, due to the cross-validation of three-mode signals, even if interfering substances are present at a concentration 20× higher than QNs, the array still shows excellent selectivity and anti-interference. Meanwhile, by optimizing various machine learning (ML) models, the classification accuracy of the concentration-irrelevant model established by the array rose from 49.13% to 94.80%. Furthermore, the integration of regression models and sensing arrays achieves accurate quantitative determination of various QNs. The present research holds considerable importance in advancing the effective development of array units and enhancing the precise detection of QNs in complex samples.