Rapid Discrimination of Pork Contaminated with Different Pathogens by Using SERS

• Published in: Food analytical methods 2024-02, Vol.17 (2), p.309-321
• Main Authors: Chen, Yahui, Peng, Yankun, Guo, Qinghui, Zhuang, Qibin, Zuo, Jiewen, Zhao, Xinlong
• Format: Article
• Language: English
• Subjects: Accuracy; Analytical Chemistry; Bacteria; Chemical composition; Chemistry; Chemistry and Materials Science; Chemistry/Food Science; Classification; Contamination; Discriminant analysis; E coli; Escherichia coli; Food Science; Foodborne pathogens; Listeria monocytogenes; Microbiology; Mixtures; Pathogens; Pork; prediction; Raman spectra; Raman spectroscopy; Salmonella; Standardization; stoichiometry; Substrates; synergism; Synergistic effect
• ISSN: 1936-9751; 1936-976X
• DOI: 10.1007/s12161-023-02567-5

Escherichia coli , Salmonella , and Listeria monocytogenes are three of the most common foodborne pathogens found in pork. SERS technology enables the rapid acquisition of molecular information by harnessing the synergistic effect of Raman scattering and enhanced substrate surface plasmonics. However, because of the similar chemical composition between bacteria and between pathogens and pork, it is difficult to discriminate pork contaminated with different pathogens. In view of this problem, Raman spectra of three bacteria, bacteria mixtures, and fresh pork contaminated with three pathogens were obtained by using Au @Ag NPs as the enhancement substrate. After data pretreatment, Raman characteristics of pathogens were analyzed by the PCA. PLS-DA and SVM-DA classifiers were employed to discriminate the feature data. To reduce the impact of sample variations on the modeling results, the influence of three data standardization methods, SNV, max-normalization, and area-normalization, on the predictive ability of the models was discussed. The predicted results demonstrated that SVM-DA, combining with SNV, achieved superior discrimination for bacteria mixtures. The sensitivity, specificity, and accuracy of prediction validation were all above 0.93. When PLS-DA was combined with normalization techniques, better results were obtained for discriminating pork contaminated with different bacteria. In the predictive discriminant analysis, the classification accuracy reached 0.85. This work not only achieved rapid discrimination of biological cells in complex environments but also improved the accuracy of the classification model, providing a feasible idea for the analysis of pork contaminated with microorganism based on SERS by incorporating stoichiometric methods.