Near infrared spectroscopy model development and variable importance in projection assignment of particle size and lobetyolin content of Codonopsis radix

Near infrared (NIR) diffuse reflectance spectroscopy was investigated to simultaneously determine the particle size (physical attribute) and active ingredient lobetyolin (chemical attribute) of Codonopsis radix. Laser diffraction and high-performance liquid chromatography were used as reference methods to determine particle size and lobetyolin content, respectively. Several spectral pretreatment methods were compared, with first derivative combined with nine-point Savitzky-Golay smoothing filter as the best method for establishing the partial least-squares models of particle size and lobetyolin. Then, synergy interval partial least squares (SiPLS) and backward interval partial least squares (BiPLS) were compared. The results showed that BiPLS was the appropriate method for establishing the particle size model; the root mean square error of calibration (RMSEC), root mean square error of cross-validation (RMSECV) and root mean square error of prediction (RMSEP) values were 34.3 mu m, 36.1 mu m and 29.2 mu m, respectively, and the values of R-cal(2), R-cv(2) and r(pre)(2) were 0.92, 0.91 and 0.94, respectively. The ratio of performance to deviation(RPD) was 4.1. Meanwhile, SiPLS was the optimal method for establishing the lobetyolin model; the RMSEC, RMSECV, RMSEP values were 0.052 mg g(-1), 0.059 mg g(-1) and 0.054 mg g(-1), respectively, and the values of R-cal(2), R-cv(2) and r(pre)(2) were 0.87, 0.84 and 0.83, respectively. The RPD was 2.5. According to the variable importance in projection( VIP) scores and the variable selection method of SiPLS, 1210-1296 nm was the second overtone of C-H; 2070-2156 nm and 2242-2328 nm were the combination of C-O, O-H and C-H. Therefore, the results showed that NIR could be used to determine physical and chemical properties simultaneously.