RESUMO
Washing and sanitation are vital steps during the postharvest processing of fresh produce to reduce the microbial load on the produce surface. Although current process control and validation tools effectively predict sanitizer concentrations in wash water, they have significant limitations in assessing sanitizer effectiveness for reducing microbial counts on produce surfaces. These challenges highlight the urgent need to improve the validation of sanitation processes, especially considering the presence of dynamic organic contaminants and complex surface topographies. This study aims to provide the fresh produce industry with a novel, reliable, and highly accurate method for validating the sanitation efficacy on the produce surface. Our results demonstrate the feasibility of using a food-grade, catalase (CAT)-immobilized biomimetic leaf in combination with vibrational spectroscopy and machine learning to predict microbial inactivation on microgreen surfaces. This was tested using two sanitizers: sodium hypochlorite (NaClO) and hydrogen peroxide (H2O2). The developed CAT-immobilized leaf-replicated PDMS (CAT@L-PDMS) effectively mimics the microscale topographies and bacterial distribution on the leaf surface. Alterations in the FTIR spectra of CAT@L-PDMS, following simulated sanitation processes, indicate chemical changes due to CAT oxidation induced by NaClO or H2O2 treatments, facilitating the subsequent machine learning modeling. Among the five algorithms tested, the competitive adaptive reweighted sampling partial least squares discriminant analysis (CARS-PLSDA) algorithm was the most effective for classifying the inactivation efficacy of E. coli on microgreen leaf surfaces. It predicted bacterial reduction on microgreen surfaces with 100% accuracy in both training and prediction sets for NaClO, and 95% in the training set and 86% in the prediction set for H2O2. This approach can improve the validation of fresh produce sanitation processes and pave the way for future research.
