Rapid detection of Rosa laevigata polysaccharide content by near-infrared spectroscopy

A method of rapid detecting Rosa laevigata polysaccharide content on the basis of near infrared spectroscopy (NIRS) was established to achieve the purpose of controlling quality of R. laevigata. A total of 129 batches of R. laevigata samples were randomly divided into calibration set and prediction set, number of which were 65 and 64 respectively. The polysaccharide content was measured according to the method provided by Chinese pharmacopoeia, and was 26.05±5.44(%). The Near-infrared diffuse reflectance spectra of R. laevigata were preprocessed by first-order derivative and autoscaling, and was built model with PLS. When 6 Latent variables (LVs) were used in model, the smallest root mean square error of cross validation (RMSECV) was 1.18%, and root mean square error of prediction (RMSEP) was 1.21%. The uninformative variables in spectrum were eliminated by UVE-PLS, and 383 variables were obtained. The prediction accuracy was improved, and RMSECV and RMSEP were 0.90% and 0.99% respectively. Then, 383 variables were further optimized by genetic algorithm (UVE-GA-PLS), and 179 variables were obtained; under this condition, the RMSECV and RMSEP were 0.93% and 1.07% respectively. In this work, result of UVE-PLS was the best. Analyzing variables VIP score in PLS, variable selected by UVE-PLS and UVE-GA-PLS, spectrum region from 7200 cm(-1) to 6700 cm(-1) of the combinations of first and second overtone of C-H and from 5300 cm(-1) to 5000 cm(-1) related to the first overtone of O-H could play an important role in the detection R. laevigata polysaccharide by NIR. Therefore, it was feasible of rapid detection of R. laevigata polysaccharide content by NIRS.