Varietal classification and antioxidant activity prediction of Osmanthus fragrans Lour. flowers using UPLC-PDA/QTOF-MS and multivariable analysis

This study was aimed to classify the varieties and predict the antioxidant activity of Osmanthus fragrans flowers by UPLC-PDA/QTOF-MS and multivariable analysis. The PLS-DA model successfully classified the four varieties based on both the 21 identified compounds and the effective compounds. For the antioxidant activity prediction, PLS performed well to predict the antioxidant activity of O. fragrans flowers. Furthermore, acteoside, suspensaside A, ligustroside, forsythoside A, phillygenin and caffeic acid were selected as effective compounds by UVE-SPA for prediction. On the basis of effective compounds, PLS, MLR and PCR were applied to establish the calibration models. The UVE-SPA-MLR model was the optimal method to predict the antioxidant activity values with Rp of 0.9200, 0.9010 and 0.8905 for DPPH, ABTS and FRAP assays, respectively. The results revealed that the UPLC-PDA/QTOF-MS combined with chemometrics could be a new method to classify the varieties and predict the antioxidant activity of O. fragrans flowers.