Development On The FTIR-ATR Spectra Quantitative Calibration Model And Its Application In Traditional Chinese Medicine

Fourier Transform Infrared Spectroscopy-Attenuated Total Refraction(FTIR-ATR) quantitative calibration model developing methods were researched in this dissertation, including multiple chemometrics methods of spectra preprocessing, spectral bands optimizing and compression and multivariate calibration. To develop the model, spectra data were preprocessed firstly using methods such as smoothing, derivation, wavelet transformation(WT), Kernel Principal Component Analysis(KPCA) and normalization, and were compressed and optimized after using Backward interval partial least squares(BiPLS), wavelet compression, consensus Uninformative Variables Elimination(cUVE) and Minimal Redundancy Maximal Relevance(mRMR), and prediction results of developed models including consensus Partial Least Squares(cPLS), Support Vector Machine(SVM), Random Forest(RF), Artificial Neural Networks(ANNs) and Least Squares Support Vector Machine(LSSVM) were compared and optimized finally. Being compared, developed models were applied in screening of phenylketonuria(PKU) and detection of Cu ions in TCM Pheretimae, results were positive.(1)The development and optimization of FTIR-ATR calibration model in PKU screening. IR spectra data of the samples were collected using FTIR-ATR, and concentration of target ingredients in the samples, which are phenylalanine(Phe) and tyrosine(Tyr), were determined, based on accomplished achievements of research group.The Phe and Phe/Tyr ratio quantitative calibration model were developed using 9-point smoothing, first differential, 9-point smoothing coupled with first differential, WT, WPT and KPCA as preprocessing methods, wavelet compression and mRMR as spectral bands compression and optimization methods, cPLS and LSSVM as model developing methods after spectral data were normalized.Evaluating indicators of denoised models before and after, such as correlation coefficient(R), root-mean-square error(RMSE), mean relative error(MRE), predictive accuracy(Acc), sensitivity(Sens) and specificity(Spec) were compared, and the results were as follows: The accuracy of WT, WTP and KPCA models were improved apparently, in which Acc of WT models were 100%, and 1D9 S combining with sym12.1 wavelet model was the optimal model, with R, RMSE, MRE were 0.91, 89.17 and 0.28 respectively; Method of spectral bands optimizing of the optimal model was mRMR, using MIQ as evaluate function, 500 as number of characteristic variables, the R, RMSEP, MRE, Acc, Sens and Spec were 0.95, 71.61, 0.28, 98.87, 97.78 and 100 respectively. Therefore, Phe concentration and Phe/Tyr ratio could be screened accurately using the FTIR-ATR quantitative calibration model developed in this dissertation, which is a hopeful new method for rapid, accuracy, and green PKU screening.(2)The development and optimization of FTIR-ATR calibration model in detection of Cu in TCM Pheretimae. First, Cu in the samples were ionized by wet digestion, and were complexed under specific condition, then, complex compounds were filtered in microporous membranes and the FTIR-ATR spectra of which were collected. Cu concentration in the same batch of samples were determined using ICP-MS.The 30 calibration set was randomly assigned as training set and test set in 3:1 ratio. The Cu content quantitative calibration model were developed using smoothing and derivation as preprocessing methods, BiPLS and cUVE as spectral bands optimization methods, RF, SVM, LSSVM and ANNs as model developing methods. The results of methods mentioned above were compared, the optimal model was BP-ANN model, with R, RMSE and MRE were 0.9658, 0.0235, 0.1193 and 0.9457, 0.0155, 0.1432 respectively. Moreover, Cu contents of pheretimae in three different producing area which were Guangdong, Guangxi and Fujian were predicted by optimized model, and were compared with ICP-MS results, It showed that relative error was less than 15% compare to ICP-MS results, and with concentration error less than 1ppm. Results above demonstrated that FTIR-ATR quantitative calibration model in detection of microscale Cu content in TCM Pheretimae was developed combining with membrane enrichment technology, provided a referential method for detection of heavy metals in TCM.