Research On Raman Sensing Technology For Rapid Detection Of Biohazards In Food

Objective:Foodborne illness caused by biohazardous contaminants is a widespread food safety problem worldwide.Currently,the detection of such contaminants relies on sophisticated laboratory instruments and the operation of professional personnel,which has limitations such as cumbersome and time-consuming detection processes.With the development of new detection technologies,Raman spectroscopy has shown excellent performance in food safety detection,specifically in terms of nondestructive detection,no sample pre-treatment and simple operation.The development of rapid and sensitive detection methods based on Raman spectroscopy is of great importance to ensure food safety and maintain consumer health.To this end,for several typical biohazardous contaminants in food,this thesis establishes biosensing techniques for single and multiple contaminant simultaneous detection in food based on Raman spectroscopy detection by combining it with machine learning,novel substrate materials,and clustered regularly interspaced short palindromic repeats（CRISPR）technology,respectively.We expect that this thesis will provide new technologies for food safety supervision and monitoring,and new ideas for the development and progress of detection technologies.Methods:1.A rapid detection method for foodborne pathogens and their virulent proteins based on Raman spectroscopy and machine learning:（1）For common foodborne pathogens（E.coli,Salmonella and Vibrio parahaemolyticus）,a Raman based generative warfare network（GAN）combined with support vector machine（SVM）detection technology was constructed,and 30000 training iterations were conducted using the GAN model to obtain pseudo spectral data close to the true spectral data,Using it as input data for the classification model,establish three precise classification methods for foodborne pathogens by optimizing the penalty parameters and kernel functions of the SVM model.（2）Aiming at the typical biological toxin Staphylococcus aureus enterotoxin B（SEB）in food,based on the pretreatment of its Raman spectrum,the competitive adaptive weighting algorithm（CARS）was used to obtain its characteristic bands,and different prediction models competitive adaptive weighting algorithm-partial least squares regression（CARS-PLR）,backward interval partial least squares（Bi PLS）,and partial least squares regression（PLR）were compared.（3）In order to meet the requirements of joint detection of food borne pathogens and multi target pollutants of their virulent proteins,Sn S₂ material was synthesized by hydrothermal method.With its characteristics of stacking nano enrichment,layer spacing expansion,and sulfur vacancy surface morphological defects,Sn S₂ material was used as SERS substrate,and machine learning pattern recognition was used to establish detection methods for four target pollutants,namely S.aureus,E.coli,SEB,and stc E.2.To improve detection sensitivity and specificity,a new SERS substrate material was introduced and combined with the CRISPR/Cas system to establish a rapid detection method for ARG pollutants and typical foodborne viruses in food:（1）For mac B,a resistance gene of macrolide antibiotics prevalent in food,gold nanoparticles were used as SERS substrates to establish the detection of mac B gene by synthesizing SERS probes combined with the ability of d Cas9-sg RNA to specifically recognize target nucleotide sequences,using mosaic methylene blue（MB）as Raman reporter molecules.（2）For the detection of hepatitis a virus（HAV）,a typical food-borne virus in food,the synthesized MXene-Au self-assembled material was used as the SERS substrate,and the detection of food-borne viruses was achieved by characterizing the self-assembled material,characterizing the Au NPs@tamra probe and verifying the ability of non-specific cleavage of ss DNA after Cas14a activation,and then analyzing the viral DNA load by Raman spectroscopy.Results:1.Raman spectroscopy combined with machine learning has good application in the detection of foodborne pathogenic bacteria and their metabolites:（1）The overall accuracy of the GAN-SVM model in the detection of foodborne pathogenic bacteria was 90%,and the probabilities of their being correctly classified were 85%for E.coli,91.2%for Salmonella and 94%for Vibrio parahaemolyticus,and 100%for the validation by the leave-one-out crossover method.（2）CARS-SVR proved to be suitable for the prediction of SEB concentration after comparing with different prediction models,and its prediction accuracy increased with the increase of SEB concentration,and the prediction accuracy was 89%,90%and 92%,respectively.（3）In the process of detecting the four target contaminants,S.aureus,E.coli,SEB and stc E,the SERS substrate based on Sn S₂combined with machine learning pattern recognition method obtained the detection ranges of 1-10⁶ CFU/m L,7.458-7.458×10⁴ CFU/m L,10^-6-10^-10 mol/L and 10^-6-10^-10mol/L.The final classification accuracy of 95%for S.aureus and E.coli and 97.43%for SEB and stc E was obtained by the SVM classification algorithm.2.At the same time,the combination of SERS and CRISPR technology has high specificity and sensitivity in the detection of nucleic acid harmful pollutants in food:（1）A detection technology for the resistance gene mac B was constructed based on the SERS-d Cas9 system,with a standard curve of y=8.79x+545.37,a correlation coefficient of R²=0.996,and a minimum detection limit of 11.9 fmol/L.（2）A detection technology for foodborne viruses was constructed based on the SERS-Cas14a system.Under optimal conditions,the detection range for hepatitis A virus was 0.0001 pmol/L-100 pmol/L,with a linear relationship of y=144.65x+968.2 and a correlation coefficient of R²=0.9877.Conclusion:In this thesis,a rapid and convenient semi-quantitative detection technique for several typical biohazard contaminants in food was successfully constructed based on Raman spectroscopy combined with machine learning algorithms.The GAN-SVM model not only improves the accuracy of classification,but also solves the problem of requiring a large number of data samples for support;The CARS-SVR model achieves rapid detection of SEB by combining Raman spectroscopy with multiple regression metrology algorithms;The Sn S₂-based SERS substrate combined with SVM achieves highly sensitive detection of the target contaminants;SERS combined with CRISPR technology further improves the specificity and sensitivity of Raman spectroscopy detection technology.The SERS-d Cas9 system has simple operation and high sensitivity in the process of detecting mac B,which can meet practical detection needs,and the detection limit reaches the level of fmol/L;The SERS-Cas14a system detects HAV technology,achieving highly sensitive detection without nucleic acid amplification.In summary,by combining Raman spectroscopy with machine learning,novel substrate materials and gene editing systems,this thesis successfully establishes biosensing techniques for different types of contaminants such as foodborne pathogenic bacteria and their virulence proteins,drug resistance genes and foodborne viruses,laying the foundation for the subsequent detection of relevant biohazardous contaminants in food.