Prediction of the anti-cancer activity of spiro derivatives of parthenin based on molecular modeling methods and docking

Abstract

A quantitative structure–activity relationship (QSAR) study has been done on the anti-cancer activity (IC50) of 66 spiro derivatives of parthenin against three human cancer cell lines, SW-620, DU-145, and PC-3. QSAR models were based on multiple linear regression (MLR), partial least square, support vector regression (SVR), and Levenberg–Marquardt back propagation artificial neural network (ANN-LM). First, stepwise MLR was employed as a descriptor selection procedure. Then selected descriptors were used as inputs for SVR and ANN models. Comparison of the results indicates that the SVR and ANN methods have better predictive power than other methods. Finally, an ANN model was developed using common molecular descriptors in three MLR models of PC-3, DU-145, and SW-620 cell lines including hydration energy (HE), G2v, and H3u, simultaneously. In order to show the effect of HE on anti-cancer activity, docking of spiro derivatives of parthenin with Nf-κB transcription factor has been done.