Monte Carlo Method and a Novel Modelling-Optimization Approach on QSAR Study of Doxazolidine Drugs and DNA-Binding

Multiple Linear Regression (MLR), artificial neural networks (ANN), simulated annealing algorithm (SA), genetic algorithm (GA) and imperialist competitive algorithm (ICA) were employed to select the most appropriate descriptors. The obtained results from four combinations of modelling-optimization methods were compared. A high predictive ability was observed for the MLR-ICA model with the best number of empires/imperialists (nEmp = 30) with root-mean-sum-squared errors (RMSE) of 0.013941 in gas phase. In the Monte Carlo method, CORAL software was used and the correlation coefficient (R-2), cross-validated correlation coefficient (Q(2)) and standard error of the model were calculated to be respectively 0.9827, 0.9761, 0.173 for the training set; and 0.9855, 0.9648 and 0.460 for the test set with an optimum threshold of 3. The best Doxazolidine derivative (according to -log IC50) was exposed to reaction with DNA at a 5'-NGC-3' using B3lyp/lanl2dz to investigate the stability of the formed complexes. It was concluded that the combination of different theoretical methods which was employed in this work, can be practically useful in designing new drugs by providing useful information on their physical and structural molecular properties, and understanding of the relation between physico-chemical, structural or theoretical molecular descriptors of drugs to their biological activities.