Revealing and predicting the relationship between the molecular structure and antioxidant activity of flavonoids

Flavonoids have potential antioxidant activity and commercial value. Exploration of the relationship between the properties of flavonoids and their molecular structure provides the theoretical basis for screening and synthesizing flavonoids with more efficient antioxidant activities. In this article, the quantitative structure-activity relationships (QSAR) of 20 flavonoids with cellular antioxidant activity (CAA) were investigated by support vector regression. Planarity is introduced into the QSAR model as a molecular descriptor. The correlation coefficient (R2) and root mean square error (RMSE) are calculated for the training and test datasets. For the training dataset, the R2train and RMSEtrain are 0.85 and 0.090, respectively, showing that our model fits well. For the test dataset, R2test and RMSEtest are 0.61 and 0.095, indicating that our model has great predictive ability. Furthermore, the effect of different glycoside substitutions on antioxidant activity was also revealed. The atomic dipole corrected Hirshfeld charge (ADCH) distribution and bond dissociation enthalpy (BDE) values show that the addition of glycoside can significantly decrease the activity of H on the phenolic hydroxyl group of the B ring, further reducing the antioxidant activity. Moreover, the formation of intramolecular hydrogen bonds is not conducive to the improvement of antioxidant activity.