[Application of DOSC combined with SBC in batches transfer of NIR quantitative model].

Near infrared model established under a certain condition can be applied to the new samples status, environmental conditions or instrument status through the model transfer. Spectral background correction and model update are two types of data process methods of NIR quantitative model transfer, and orthogonal signal regression (OSR) is a method based on spectra background correction, in which virtual standard spectra is used to fit a linear relation between master batches spectra and slave batches spectra, and map the slave batches spectra to the master batch spectra to realize the transfer of near infrared quantitative model. However, the above data processing method requires the represent activeness of the virtual standard spectra, otherwise the big error will occur in the process of regression. Therefore, direct orthogonal signal correction-slope and bias correction (DOSC-SBC) method was proposed in this paper to solve the problem of PLS model's failure to predict accurately the content of target components in the formula of different batches, analyze the difference between the spectra background of the samples from different sources and the prediction error of PLS models. DOSC method was used to eliminate the difference of spectral background unrelated to target value, and after being combined with SBC method, the system errors between the different batches of samples were corrected to make the NIR quantitative model transferred between different batches. After DOSC-SBC method was used in the preparation process of water extraction and ethanol precipitation of Lonicerae Japonicae Flos in this paper, the prediction error of new batches of samples was decreased to 7.30% from 32.3% and to 4.34% from 237%, with significantly improved prediction accuracy, so that the target component in the new batch samples can be quickly quantified. DOSC-SBC model transfer method has realized the transfer of NIR quantitative model between different batches, and this method does not need the standard samples. It is helpful to promote the application of NIR technology in the preparation process of Chinese medicines, and provides references for real-time monitoring of effective components in the preparation process of Chinese medicines.

[Feature Selection and Interpretation in Infrared Quantitative Models of Liquiritin and Glycyrrhizin in Radix Glycyrrhizae].

Feature selection can improve the interpretation of the modeling variables to a certain extent by selecting variables from the complex spectra backgrounds. However, the improvement of models interpretation does not mean that the modeling variables have the exact physical or chemical significance. In this paper, We explore the relation between the chemical characteristics of target components and the spectrum variables selected with 3 kinds of variables selection methods which are moving window partial least squares regression (mwPLS), synergy interval partial least squares regression (siPLS) and competitive adaptive re-weighted sampling (CARS), and compare the interpretation difference of the variables selected with the above variables selection methods. The results show that the variables selected with mwPLS accord with ν(φ)C=C of liquiritin and δCH3 or δCH2 of glycyrrhizin, which are the obvious spectra differences between the flavonoids and saponins in Radix Glycyrrhizae, and the variables selected with siPLS are the characteristic intervals combinations of the flavonoids or saponins in Radix Glycyrrhizae, which is the combination of ν(ø)C=C, ν(ø)C-O, ν(ø)C-H of flavonoids or the combination of νC-O vC-H, νO-H of saponins while the variables selected with CARS can better accord with most of the characteristic peaks from 1000 to 4000 cm(-1) of liquiritin or glycyrrhizin in Radix Glycyrrhizae, and the predict performance of the infrared quantitative model established on the spectroscopic variables selected with CARS can be improved. Therefore, most of the variables selected with CARS can be interpreted by the characteristic peaks in the infrared characteristic region of the target components, which is beneficial to improve the interpretation of the quantitative model.

[Effect of algorithms for calibration set selection on quantitatively determining asiaticoside content in Centella total glucosides by near infrared spectroscopy].

The appropriate algorithm for calibration set selection was one of the key technologies for a good NIR quantitative model. There are different algorithms for calibration set selection, such as Random Sampling (RS) algorithm, Conventional Selection (CS) algorithm, Kennard-Stone(KS) algorithm and Sample set Portioning based on joint x-y distance (SPXY) algorithm, et al. However, there lack systematic comparisons between two algorithms of the above algorithms. The NIR quantitative models to determine the asiaticoside content in Centella total glucosides were established in the present paper, of which 7 indexes were classified and selected, and the effects of CS algorithm, KS algorithm and SPXY algorithm for calibration set selection on the accuracy and robustness of NIR quantitative models were investigated. The accuracy indexes of NIR quantitative models with calibration set selected by SPXY algorithm were significantly different from that with calibration set selected by CS algorithm or KS algorithm, while the robustness indexes, such as RMSECV and |RMSEP-RMSEC|, were not significantly different. Therefore, SPXY algorithm for calibration set selection could improve the predicative accuracy of NIR quantitative models to determine asiaticoside content in Centella total glucosides, and have no significant effect on the robustness of the models, which provides a reference to determine the appropriate algorithm for calibration set selection when NIR quantitative models are established for the solid system of traditional Chinese medcine.