Quantitative structure-activity relationships (QSAR) is a mathematical model of a compounds chemical structure and its relationship with the biological activity. Molecular structures are represented by numerical descriptors include various physicochemical, quantum chemical and partial charge.

Mordred are open source and can be easily installed as a high‑flexibility Python2/3 package. Furthermore, mordred can generate total 1825 2D and 3D molecular descriptors in a high calculation speed.

In vivo and in vitro, numerous methods have been developed to measure the antioxidant capacity and effectiveness. Among free radical scavenging methods, DPPH method is furthermore convenient, simple and inexpensive in comparison to other test models. DPPH method was found to be used mostly for the in vitro antioxidant activity evaluation purpose.

The biological activity was expressed by IC50 values (the molar concentration of antioxidants required to inhibit 50% of DPPH activity) ranging from 1.7 to 43600 μmol/L. A lower IC50 value indicates greater antioxidant activity.

Support vector machine model was established using the open source Mordred descriptors; The cost and gamma parameters of the model are 50 and 0.05, respectively.

When the Molecular Generated Moredred Descriptor has missing values, the model prediction will be wrong. Furthermore, it is to be noted that the applicability of these predictions is limited by the training data used for the machine learning models.

The SVR-Mordred model performs robustly when predicting external compounds (R2 =0.83 and RMSE=0.23).