Study On The Application Of Infrared Spectroscopy And Chemometrics For Disease Diagnosis And Detection Of Heavy Metals In Traditional Chinese Medicine

In this dissertation, a rapid neonatal screening method for phenylketonuria(PKU)and a rapid method for detection of copper content in traditional Chinese medicine Lycium barbarum were developed, which were based on Attenuated Total internal Refraction Fourier Transform Infrared Spectroscopy(ATR-FTIR) coupled with chemometrics methods. To extract and compress the infrared spectroscopy(IR) data of the samples, wavelength selected methods, such as uninformative variables elimination(UVE), backwards interval partial least squares(BiPLS) and random forest(RF) were applied. Then the quantitative calibration models for target ingredients prediction, such as consensus partial least squares(cPLS) models and least square support vector machine(LSSVM) models were developed. Performance of the models was evaluated by various kinds of evaluation indexes. A methodology reference of applications on disease marker detection and rapid analysis of TCM was provided in this research.The main research contents were as follows:1. Study on rapid neonatal PKU screening methodIn this chapter, quantitative calibration models for the predictions of Phe and Phe/Tyr ratio were developed based on spectra data and reference concentrations of Phe and Tyr which were obtained using LC/MS/MS. With the whole spectra and the selected intervals, two kinds of different calibration models were developed and compared by calculating the correlation coefficient(R), root mean square error of prediction(RMSEP), mean relative error(MRE) and predictive accuracy(Acc),sensitivity(Sens) and specificity(Spec). The results showed that, in terms of prediction of Phe concentration, the best performed model was the RF-LSSVM model developed by derivative spectra, yielding R, RMSEP, MRE as 0.94, 81.72, 0.26 and Acc, Sens, Spec as 98.84, 97.57, 100, respectively; and in terms of prediction of Phe/Tyr ratio, the best model was BiPLS-cPLS model developed by smoothed spectra,yielding R, RMSEP, MRE as 0.93, 3.55, 0.36 and Acc, Sens, Spec all as 100. While the Phe model sometimes failed to classify those samples with concentrations near to the threshold value, the Phe/Tyr model could improve both the Sens and Spec to 100.2. Study on the model of copper concentration in Lycium barbarumIn this chapter, with the spectra of Lycium barbarum powder and the copper concentration measured by ICP/MS, various copper concentration quantitative models were developed. Models with whole spectra and optimized spectra were evaluated and compared by R, RMSEP and MRE. And results showed that the best model was RF-LSSVM model established by vector normalized spectra, which yielded R,RMSEP, MRE as 0.92, 0.42, 0.04. The present method can be applied to the rapid analysis of copper concentration.