Machine Learning-Enhanced Three-Channel Cu-Pt Bimetallic Nanozyme Sensor Array for Multi-Class Pesticide Fingerprint Discrimination

Abstract

Pesticide residues with diverse toxicities and synergistic effects demand rapid, multiplexed discrimination methods beyond conventional single-analyte assays. Here, we report a machine-learning-enhanced three-channel colorimetric sensor array based on Cu-Pt bimetallic nanozymes for multi-class pesticide fingerprint recognition. The single CuPtBN receptor integrates oxidase-, laccase-, and superoxide dismutase-like activities, generating three orthogonal optical response channels that capture differential inhibition and promotion effects from structurally similar pesticides. By coupling these multidimensional signals with “classification-regression” dual-loop machine learning models, the platform achieves concentration-independent qualitative identification and precise quantitative prediction, with 100% classification accuracy for nine representative pesticides and reliable predictive capability for nonlinear-response analytes. In blind tests and real agricultural samples, the system accurately identified pesticide species with prediction errors below 5% and recoveries of 85−105%. This work establishes a simplified yet information-rich multichannel nanozyme sensing strategy, highlighting its promise for high-throughput food safety screening and intelligent agrochemical monitoring.