Rapid quantitative detection of the discrepant compounds in differently processed Curcumae Rhizoma products by FT-NIR combined with VCPA-GA technology.

Curcumae Rhizoma (CR) and vinegar processed Curcumae Rhizoma (PCR) are common medicinal materials widely used in clinical practice in China. There are sliced CR (SCR) and two kinds of PCR products which are processed by different methods: WPCR-prepared with the whole CR root boiled in vinegar and then sliced, and SPCR-prepared with the whole CR root steamed and sliced before boiled with vinegar. In this study, the feasibility of Fourier transform near infrared spectrum (FT-NIR) used to determine the main discrepant components of SCR, WPCR and SPCR were investigated. High performance liquid chromatography (HPLC) was used to identified five discrepant compounds in the three kinds of CR products-curzerene, curcumenol, curdione, furanodienone and demethoxycurin. Pretreatment of NIR qualitative data by different methods revealed that the second derivative in combination with 9 points of Savitzky-Golay smooth (2D9S) could accurately distinguish SCR, SPCR and WPCR from each other, and the discrimination ability was improved significantly by wavebands selection. Then a model with great accuracy was established by combining with wavebands selection and partial least squares regression (PLSR). Compared with the competitive adaptive reweighted sampling (CARS) selection method, 2D9S- variable combination population analysis (VCPA)-Genetic algorithm (GA)-PLSR model was evidently more accurate in prediction of the content of curzerene, curcumenol, curdione and furanodienone, with an R2p of 0.9558, 0.9129, 0.9098 and 0.9350, as well as a ratio of performance to deviation (RPD) of 4.8454, 3.4640, 3.3020 and 4.0082, respectively. Whereas, the content of demethoxycurin failed to be well predicted. The correlation analysis revealed that the results of wavebands selection were consistent with the trend of changes in the content of these target compounds and the findings of NIR absorption analysis, and the characteristic chemical bonds of these compounds corresponded to the areas with significant correlation in the heat map. It can be concluded that the NIR system, combined with appropriate variable selection and linear regression method, can precisely distinguish SCR, SPCR and WPCR from each other, and can accurately and rapidly determine the four discrepant compounds in the three CR products, suggesting a potential of being routinely used for a more diversified analysis in medicinal herbs study.