Multiplexed discrimination and ultrasensitive determination of pesticides using a Cu-O-Mo nanozyme tri-channel array driven by machine learning.

Nanozyme-based multichannel sensors integrated with machine learning (ML) often operate as "black boxes," lacking interpretability between analytical outputs and underlying molecular mechanisms. To address this, we developed a polyoxometalate-functionalized MOF featuring Cu-O-Mo bridging motifs, which integrates strong surface-enhanced Raman scattering (SERS) enhancement with synergistic peroxidase (POD)-and polyphenol oxidase (PPO)-like nanozyme activities. A tri-channel assay system (POD-SERS, POD-UV, PPO-UV) coupled with ML algorithms (PCA, LDA, confusion matrix, ROC-AUC) enabled accurate discrimination of seven distinct pesticides and their mixtures, achieving ultrasensitive detection of pirimicarb at 0.1 μg/mL. The platform successfully tracked pesticide residue metabolism on citrus peels over 10 days. Mechanistic studies revealed that POD inhibition is governed by coordination to Mo6+/Mo5+ centers, while PPO inactivation occurs via high-affinity binding to Cu+/Cu2+ sites, with secondary steric and redox effects enhancing specificity. The catalytic mechanism was confirmed through Mo K-edge EXAFS and XANES analyses, competitive binding assays, and IC50 determinations. This work establishes a field-deployable and interpretable framework for multiplex pesticide screening in agricultural and environmental monitoring.