A rapid integrated bioactivity evaluation system based on near-infrared spectroscopy for quality control of Flos Chrysanthemi.

For quality control of herbal medicines or functional foods, integral activity evaluation has become more popular in recent studies. The majority of researchers focus on the relationship between chromatography/mass spectroscopy and bioactivity, but the connection with spectrum-activity is easily ignored. In this paper, the near infrared reflection spectra (NIRS) of Flos Chrysanthemi samples were collected as a representative spectrum technology, and corresponding anti-inflammation activities were utilized to illustrate the spectrum-activity study. HPLC/Q-TOF-MS identification and heat map clustering were used to select the quality markers (Q-marker) from five cultivars of Flos Chrysanthemi. Using boxplot analysis and the interval limits of detection (LODs) theory, six crucial markers, namely, chlorogenic acid, 3,5-dicaffeoylquinic acid, 1,5-dicaffeoylquinic acid, luteoloside, apigenin-7-O-ß-d-glucoside, and luteolin-7-O-6-malonylglucoside were screened out. Then partial least squares regression (PLSR) calibration models combined with synergy interval partial least squares (siPLS) and 12 different spectral pretreatment methods were developed for the parameters optimization of these Q-markers in Flos Chrysanthemi powder. After comparing the relationship between Q-marker contents and anti-inflammation activity via three machine learning approaches and PLSR, back-propagation neural network (BP-ANN) displayed a more excellent non-linear fitting effect, as its R for new batches reached 0.89. These results indicated that the integrated NIRS and bioactive strategy was suitable for fast quality management in Flos Chrysanthemi, and also applied to other botanical food quality control.