Data Fusion Strategy Combined With Chemometrics To Trace The Origins Of Paris Polyphylla Smith Var.yunnanensis

Under the concept of comprehensive health,traditional Chinese medicine shows the predominant advantage in prevent,healthcare,therapy,rehabilitation and so forth,which causes the extensive attention around domestic and international.However,the quality of medicine is the crucial factor to influence the safety and effectiveness of clinical medication.Hence,evaluating the quality of traditional Chinese medicine is the premise of clinical medication,and is also the core of the research and development of traditional Chinese medicine.The quality of traditional Chinese medicine is affected by many factors,such as variety,origins,harvesting and so forth,which can affect the accumulation of secondary metabolites and result in the uneven quality of medicinal materials.The spectroscopic technique can reflect the quality of medicinal materials in the whole form of the chemical information basing on the quickly and nondestructive procedures,which conforms to the holistic view of traditional Chinese medicine.As a conventional quantitative analysis instrument,chromatography can directly evaluate the quality of medicinal materials by analyzing the index components.Both of spectroscopic and chromatography techniques combined with chemometrics,a rapid,simple and comprehensive method for the quality evaluation of traditional Chinese medicine can be established.This paper concentrated on tracing the provenance and botanical origins of genus Paris?mainly Paris polyphylla Smith var.yunnanensis?.Analyzing instrument of attenuate total reflection Fourier transform infrared spectrometry?ATR-FTIR?,ultraviolet-visiblespectroscopy?UV-Vis?,ultra-high-performanceliquid chromatography?UPLC?,were used to analyze samples.Besides,principal component analysis?PCA?,partial least squares discriminant analysis?PLS-DA?,support vector machine gird search?SVM-GS?,random forest and other chemometrics methods,were employed to comprehensively uncover the differences among different origins?wild and cultivated?and different botanical?wild?samples.This study aimed to establish a systematic model for quality evaluation of P.polyphylla Smith var.yunnanensis and provide a basis for its quality assessment.This paper was consisted by five chapters.The first chapter was identifying and tracing the geographical origins of 177 wild P.polyphylla Smith var.yunnanensis samples obtained from 14 counties and belonged to 3 production areas of Yunnan,basing on the analysis of FTIR and UV-Vis spectra,low-and mid-level data fusion strategies combined with chemometrics.In chapter 2,ATR-FTIR and UV-Vis instruments were carried out to analyze 161 batch wild P.polyphylla Smith var.yunnanensis samples collected from 6 producing areas of Yunnan,and correctly distinguished samples by high-level data fusion strategy combined with chemometrics.In chapter 3,FTIR and UPLC analysis were applied to determine 87 batch wild Paris samples?6 various species?and properly identified with the help of data fusion method combination of chemometrics.In the fourth chapter,the ATR-FTIR spectra of rhizome and leaf tissues for 219 batch cultivated P.polyphylla Smith var.yunnanensis?collected from 5 origins and covered 3 cultivation years?and the corresponding biomass data were used to recognize different provenances samples.In chapter 5,reviewed the application of spectrum and chromatography in the quality evaluation of P.polyphylla Smith var.yunnanensis.In this chapter,FTIR and UV-Vis spectra of P.polyphylla Smith var.yunnanensis were collected.The FTIR spectral was range from 4000 to 400 cm^-1 and the UV-Vis was range from 200 to 500 nm.Spectra were pretreated by standard normal variate?SNV?,second derivative?SD?and savitsky-golay?SG?.Then,the pretreated spectra?FTIR and UV-Vis?and the combinations?low-and mid-level data fusion?were used to establish PLS-DA and SVM-GS models.Compared with the single spectrum and low-level data fusion models,results of PLS-DA and SVM-GS models basing on mid-level data fusion strategy showed a better identification performance for northwest and southeast Yunnan samples.For the large sample size?177 batch samples from three origins?and multi-classification variables?samples obtained from central Yunnan belonged to seven microhabitats?,the mid-level data fusion method combining with chemometric method could correctly identify samples and latent variables?LVs?was more suitable for screening characteristic variables than principal components?PCs?.Therefore,compared the classification results of PLS-DA and SVM-GS models established by a single spectrum matrix,the model performance was more robust based on data fusion strategy,which could comprehensively reflect the difference of chemical components among samples.The SVM-GS and random forest model identification results of wild P.polyphylla Smith var.yunnanensis were compared basing on low-,mid-and high-level data fusion strategies.Results showed that the SVM-GS model established basing on low-level data fusion strategy could not correct identify samples collected from different origins,and the results of mid-level data fusion method was superior to the former.The model established by high-level data fusion method showed the most satisfying result and the accuracy for training and test sets of random forest model were 99%and 98%,respectively,both of which were significantly higher than the results of ATR-FTIR,UV-Vis,low-and mid-levle data fusion strategies.According to the misclassified individuals,the accuracy of training and test sets,the performance of three pattern recognition methods?PLS-DA,SVM-GS and random forest?were compared basing on identification results of different botanical and geographical origins of Paris samples.Results revealed that the accuracy among different species was higher than that of P.polyphylla Smith var.yunnanensis collected from various origins,indicating that the difference of chemical profiles among different species was greater than that of samples picked from different origins,conclusion of which was consistent with the quantitative results of polyphyllin saponins and one-way analysis of variance?ANOVA?results.Additionally,FTIR spectrum showed a better classification result than UPLC chromatogram for the identification of different species.By comparing the model performance established by ten data matrices,conclusion could be drawing that the model established basing on data fusion strategies?especially for the mid-level?provides higher accuracy and robustness than the polyphyllin content,FTIR spectrum,and UPLC chromatogram.Tracing the geographic origins of cultivated P.polyphylla Smith var.yunnanensis basing on fused different tissues ATR-FTIR spectra and combined the chemometrics of PLS-DA and random forest.Compared with the model result established by single tissue ATR-FTIR spectra,the classification model established basing on data fusion strategy showed a high accuracy.Besides,none sample was misclassified for both training and test sets showing in PLS-DA model,results of which were verified by the5-,6-and 7-year cultivated P.polyphylla Smith var.yunnanensis samples.Therefore,ATR-FTIR combined with data fusion strategy could be used to trace the geographical origins and this method has the advantages of direct,rapid and easy operation.Additionally,from the perspective of green analysis,infrared spectrum analysis method was a green analysis method,which did not involve organic reagents,reduce waste liquid generation,and the analyte can be recycled.