An intelligent sensing system for thermal processing contaminants via decoding nucleophilic reactivity fingerprints with a rationally designed sensor array.

Thermal processing contaminants like chloropropanols and glycidol are hazardous foodborne carcinogens. Existing detection methods are hindered by complexity and laboratory dependency. In this work, we introduce an intelligent sensing system that decodes the nucleophilic reactivity fingerprints of these contaminants via a rationally designed nanozyme-based sensor array. The mechanism centers on contaminants' specific restoration of nanozyme activity inhibited by nucleophilic reductants, generating unique fingerprint responses. An optimal sensor array is designed under the guidance of a dual-task feature selection framework, rationally refining the initial 15 sensing pathways down to an optimal 6-element array. This system enables high-throughput identification (F1 score > 0.95 on an independent test set) and precise quantification (R2 > 0.99) for individual and mixed contaminants. It exhibited remarkable robustness in complex food matrices (soy sauce, biscuit, edible oil), with quantification accuracy (R2 > 0.98) after simple sample preparation. This work provides an intelligent platform integrating chemical principles with AI for precision food safety and on-site decision-making.