Exploring QSARs for Inhibitory Activity of some Antimalarial Compounds by MLR and PC-AN

Abstract

As malaria disease is continuous to be one of the major health problems, and until now no effective vaccines or drugs are available due to the mutation of the plasmodium. So, in order to help in designing a new antimalarial agent, a quantitative structure activity relationship was performed to study the Activity of 79 compound as antimalarial agents. The QSAR models were developed using the multiple linear regression (MLR) as a linear method. The principal component – artificial neural network (PC-ANN) was used as nonlinear method for modeling. The models resulted have a good prediction power. The MLR models (13-17) which have R² >0.6, the best model was model number 17 with correlation coefficient R= 0.889, R²= 0.791, and R²adj.= 0.733. Cross validation LOO and LMO were performed on the resulted MLR models 13 - 17 in which they showed a good predictive power. The PCA was performed to divide the data into three data sets; training, validation and test set. Then the ANN performed on the models 13-17. The resulted ANN models were validated by randomization test, then the conditions that proposed by Golbraikh and Tropsha were applied to confirm that the QSAR models have acceptable prediction power or not. However, the best ANN model with the best predictive power was model number 17, with R value 0.8138.