Quantitative Detection Of Binary Drug Residues In Milk Based On Surface-Enhanced Raman Spectroscopy

Milk is a rich source of minerals,calcium and phosphorus for the body.However,the non-standard production process of a small number of milk may cause drug residues in the milk,causing harm to consumers’bodies.Although the current conventional detection methods for drug residues in milk have the advantages of high sensitivity and accuracy,the detection costs are high,the cycle is long,professional training is required for operators,and they are mainly for single substance detection,which cannot meet the requirements of milk processing and circulation.demand for rapid quantitative detection of multiple substances.The high sensitivity and rich fingerprint information of surface-enhanced Raman scattering（SERS）are conducive to the rapid detection and identification of multiple analytes.Therefore,in this paper,the drugs thiabendazole（TBZ）and ciprofloxacin（CIP）that may exist in milk are used as the detection object.Based on SERS technology,combined with chemometrics and other technical means,the rapid and high-precision quantitative detection of various drug residues in milk was achieved.The main research contents of this paper are as follows:1.A simple and portable droplet detection system was constructed.After fully mixing the prepared silver nanoparticles with the analyte,pipetting a small amount of the mixture onto the substrate can be used for SERS detection.Taking the mixed solution of 50 mg/L TBZ solution and 5 mg/L CIP solution as an example,the effects of substrate type,sample volume,and volume ratio of analyte and silver nanoparticles on the SERS detection system were investigated.The experimental conditions were optimized.The uniformity of the SERS signal was characterized under the optimized conditions,and the average relative standard deviation（RSD）values of the SERS spectral intensities corresponding to the three characteristic peaks were 2.2%,respectively.The good repeatability indicates that the detection system can provide reliable detection data in actual detection.2.A method for high-precision quantitative analysis of binary analytes in milk was developed.The self-modeling mixture analysis（SMA）algorithm was used to extract each component in the mixture spectrum,which reduced the influence of the overlapping peaks in the mixture spectrum on quantitative analysis.A scheme of taking part of spectral information as known data is proposed,which solves the problem that SMA cannot extract pure components from a single mixed spectrum to be measured due to matrix operations.A nonlinear Support Vector Machine Regression（SVR）model was established using the TBZ and CIP pure component spectra extracted by SMA as the training set of the model.The types of kernel functions in the SVR model and the influence of kernel function parameters on the model are analyzed,and the best parameter combination is found through grid optimization.Using this method to predict the mixture of TBZ and CIP in milk solution samples,the²of the SVR model for TBZ concentration prediction is0.9889,the²of the CIP concentration prediction model is 0.9918,and the²of TBZ and CIP concentrations are both close to 0.99,indicating that the method has high predictive reliability in complex systems and can be used for high-precision quantitative analysis of multiple analytes in milk.3.A software for rapid one-step quantitative analysis of binary analytes in milk was developed.Based on Raman characteristic peaks,a partial least squares regression（PLSR）concentration prediction model was established to realize the concentration prediction of TBZ and CIP in mixtures;automatic variable selection and spectral intelligent analysis schemes were constructed.The influence of characteristic wavenumbers extracted based on full spectrum,competitive adaptive reweighted sampling algorithm（CARS）and continuous projection algorithm（SPA）on the accuracy of the prediction results of the PLSR concentration prediction model was analyzed and compared,and the spectral information extraction scheme based on SPA was determined.Using this method to quantitatively analyze the mixture of TBZ and CIP in milk solution samples,the coefficient of determination(²)of the test set of the TBZ concentration prediction model is 0.9724,and the²of the CIP concentration prediction model is 0.9814,which proves the effectiveness of the method.reliability.A user-friendly multi-analyte quantitative analysis software was developed,and on the premise of ensuring the basic accuracy of the prediction,one-step rapid quantitative analysis of each component in the multi-analyte was realized.