Abstract
Malaria has been known as one of the major causes of morbidity and mortality on a large scale in tropical countries until now. In the past decades, many scientific groups have focused their attention on looking for ideal drugs to this disease. So far, this research area is still a hot topic. In the present study, the antimalarial activity of 1, 4- naphthoquinonyl derivatives was modeled by linear and nonlinear statistical methods, that is to say, by forward stepwise multilinear regression (MLR) and radial basis function neural networks (RBFNN). The derived QSAR models have been statistically validated both internally - by means of the Leave One Out (LOO) and Leave Many Out (LMO) crossvalidation, and Y-scrambling techniques, as well as externally (by means of an external test set). The statistical parameters provided by the MLR model were R2 =0.7876, LOOq2 =0.7068, RMS =0.3377, R0 2 =0.7876, k =1.0000 for the training set, and R2 =0.7648, q2 ext =0.7597, RMS=0.2556, R0 2=0.7598, k=1.0417 for the external test set. The RBFNN model gave the following statistical results, namely: R2=0.8338, LOOq2=0.5869, RMS=0.2781, R0 2 = 0.8335, k=1.0000 for the training set, and R2 =0.7586, q2 ext =0.7189, RMS=0.2788, R0 2=0.7129, k=1.0284 for the external test set. Overall, these results suggest that the QSAR MLR-based model is a simple, reliable, credible and fast tool for the prediction and virtual screening of 1, 4-naphoquinone derivatives with high antimalarial activity. In addition, the energies of the highest occupied molecular orbital were found to have high correlation with the activity.
Keywords: Antimalarial activity, dug design, malaria, molecular design, multiple linear regression (MLR), 1, 4- Naphthoquinonyl derivatives, QSAR, radial basis function neural network (RBFNN), statistical methods, model
29