Comprehensive Evaluation Of Plantaginis Semen Quality Based On Near-infrared Spectroscopy And Chemometrics

Plantaginis Semen(PS),the dried mature seed of Plantago asiatica L.or Plantago depressa Wild.(Plantaginaceae),was firstly recorded in“Shen Nong’s Herbal Classic”,and is a common traditional Chinese medicine as well as a drug homologous food.In view of the shortcomings of the current 2020 edition of the“Pharmacopoeia of the People’s Republic of China”(Ch P)for the quality evaluation of PS,such as the content determination of only two single components,this paper explores the establishment of a simple and rapid analysis method,which is based on near-infrared spectroscopy(NIRS)combined with chemometrics,to establish an analytical model,and to simultaneously determine the content of geniposidic acid,verbascoside,isoverbascoside and total flavonoids in PS,as well as the antioxidant activities,in order to comprehensively evaluate the quality of PS.The main research contents and results are as follows:1.Rapid and simultaneous determination of the content of a variety of components in PS by near-infrared spectroscopy combined with chemometrics.First,seventy-four batches of PS samples were collected from four different places of production.After pretreatment,the content of geniposidic acid,verbascoside and isoverbascoside in PS were determined by HPLC method,while the content of total flavonoids in PS was determined by UV-Vis spectrophotometry method.Second,NIR spectral data were collected for all batches of samples.Finally,the sample content data was correlated with the NIR spectral data,and then a multivariate calibration model was established using the chemometrics method.A variety of spectral preprocessing methods,such as standard normal variation(SNV),multiplicative scatter correction(MSC),smoothing,derivative and their combination methods were investigated,and a variety of wavelength selection methods were compared,such as genetic algorithm(GA),particle swarm optimization(PSO),and competitive adaptive reweighted sampling(CARS).All samples were divided into a calibration set and a prediction set at the ratio of 3:1 by the descending order of reference values.The coefficients of determination(R~2),root mean square error(RMSE),mean absolute error(MAE)and residual predictive deviation(RPD)were calculated to comprehensively evaluate the model performance.The results show that after optimization,the performance of each model was greatly improved,and the GA-PLSR model based on genetic algorithm optimized wavelength combined with partial least squares has the best performance,where the R~2of calibration set and prediction set were both greater than 0.835,and the PRD were both greater than 2.54,indicating that the models had good robustness and prediction accuracy,and could be adopted to rapidly and accurately determine the content of geniposidic acid,verbascoside,isoverbascoside and total flavonoids in PS.2.Rapid determination of antioxidant activities of PS by NIRS combined with chemometrics.First,the antioxidant capacities of PS were characterized by DPPH radical scavenging assay,ABTS radical scavenging assay and ferric reducing antioxidant power(FRAP)assay,respectively.Then,the measured values of antioxidant capacity of each group were used as the reference values,which were correlated with the NIR spectral data of the corresponding sample,and then a multivariate calibration model was established.In order to obtain better modelling accuracy,ten spectral preprocessing methods were also investigated in this paper,and three wavelength variable selection methods,such as GA,PSO and CARS were optimized.The results showed that after optimization,the performance of the three antioxidant capacity prediction models were greatly improved,and the GA-PLSR model based on the optimal wavelength of the genetic algorithm combined with the partial least squares method showed the best performance,where the R~2of calibration set and prediction set were both greater than 0.880,and the PRD were both greater than 2.97,indicating that the models had good robustness and prediction accuracy,and can be applied for quick and accurate determination of the antioxidant capacity of PS.3.Study on the application of bionic swarm intelligence optimization algorithm to wavelength selection in NIRS modeling.Wavelength selection is an important step in the modelling of NIRS,which is of great significance to reduce model complexity and to improve model performance.In the modelling work of the first two chapters,we found that the wavelength selection strategies based on swarm intelligence optimization algorithms such as GA and PSO performed well in improving the prediction quality of the NIR model.Motivated by this,in the present chapter we further explores 10 bionic swarm intelligence optimization algorithms inspired by natural creatures,and evaluates their performance for wavelength selection in NIR spectral modelling.With three benchmark NIRS datasets and PS datasets as the research objects,and with the R~2,RMSE and RPD in the calibration set and prediction set as the model quality evaluation indexes,the performance of the bionic swarm intelligent optimization algorithm is comprehensively evaluated,and is also compared with the classical wavelength selection algorithm.The results obtained showed that the 10 swarm intelligent optimization algorithms can significantly reduce the number of wavelengths with the compression ratio>50%(namely less than half of the wavelengths retained).Compared with the full spectra models,the present models not only simplified the model structures,but improved the model performances.Furthermore,the performances were generally better than the ones by some popular and classic wavelength selection algorithms.