RESUMO
INTRODUCTION: In this paper, we have introduced a new atomic descriptor with Klopman index to determine the local reactive sites of the molecular systems during electrophilic, and nucleophilic attacks. This index, similar to other local reactivity descriptors but more advanced, has been used as a realistic descriptor to discover new aspects of molecular structure. METHODS: Nonlinear Least Squares (NLLS) methods to define the parameters maximizing the fit between the observed points and the computed simulation results were performed according to the Levenberg- Marquardt (LM) algorithm. We have attempted to demonstrate the structural properties of compounds that contribute not only basic pharmacophore (b-Pha) but also positive (Auxiliary Group- AG) or negative (Anti-Pharmacophore Shielding-APS) due to the new local atomic reactivity. RESULTS AND CONCLUSION: In the 4D-QSAR study, nonparametric regression analysis was used to determine the adjustable constants. Using the leave One Out Cross-Validation (LOO-CV), antibacterial activities (pEC50-µM) were predicted as r2 loo-cv (q2) = 0.979, r2 pred (r2) = 0.911, respectively, for 27 training sets and 9 test set compounds. Also, the rm2 (overall) value, which indicates the closeness between the predicted and corresponding observed data, was calculated to be 0.957. The model obtained by the Molecular Conformer Electron Topological (MCET) method was compared with the q2 loo-cv and R2 non-cv values determined by the CoMFA and CoMSIA methods and more satisfactory results were obtained.