QSAR study of novel indole derivatives in hepatitis treatment by stepwise- multiple linear regression and support vector machine

The quantitative structure – activity relationship (QSAR) of the novel indole derivatives for prediction of the half maximal inhibitory concentration (IC50), in hepatitis treatment was studied. After calculating the 1481 molecular descriptors, the stepwise (SW) was used as variable selection method to select the most appropriate molecular descriptors. The selected descriptors are Mp, MATS6e, GATS8e, Mor22v, R7v+ and MLOGP, which are of the Constitutional, 2D autocorrelations, 3D-MoRSE, GETAWAY, Molecular properties groups. Modeling was then performed using multiple linear regression (MLR) and support vector machine (SVM). The robustness and the predictive performance of the developed models was tested using both the internal and external statistical validation (test set) of ten compounds, randomly chosen out of 48 compounds. The SVM model with optimal parameters C of 11, γ of 1 and ε of 0.07 has the R2 (0.993, 0.844) and RMS errors (0.068, 0.269) for the training and test sets, respectively, which are better than MLR method (R2=0.886, 0.583 and RMS error=0.272, 0.441). Based on the information derived from the model, some key features for increasing the activity of compounds have been identified and can be utilized to designing new indole derivatives in hepatitis treatment.