Machine Learning-Assisted Dual-pH Electrochemical Sensor for Rapid Detection of Quercetin, Rutin and Glucose in Litchi Fruit

Electrochemical sensing provides an alternative approach for the trace detection of bioactive substances in fruits. However, the complex matrix in fruit tissues, the coexistence of multiple active components, and the varied pH environments limit the sensing performance and accurate quantitative detection of conventional electrochemical sensors. Herein, a dual-mode electrochemical sensor based on a Co3O4@N-MWCNTs modified glassy carbon electrode was developed for the sequential detection of quercetin, rutin, and glucose in fruits under acidic and alkaline conditions. The as-prepared electrode exhibited improved charge transfer efficiency and favorable electrocatalytic activity toward the three target analytes. Under optimal conditions, the sensor displayed wide linear ranges of 0.5~70 μM for quercetin and 0.5~5 μM for rutin in acidic environment, with low detection limits of 0.124 μM and 0.045 μM, respectively. In alkaline environment, the detection limit for glucose was determined to be 8.86 μM. Moreover, four combined machine learning models with feature selection algorithms were established, among which the CARS-RFE+RFR model achieved the best prediction accuracy and robustness for multicomponent quantification. Furthermore, the proposed sensing system was applied to the rapid determination of quercetin, rutin, and glucose in real litchi samples, with recoveries ranging from 98.4% to 105.4%. This study provides a feasible electrochemical strategy for multicomponent detection in complex plant matrices, showing good applicability for rapid on-site analysis in agricultural and food-related applications.