No abstract

Two different approaches, namely the electron conformational and genetic algorithm methods (EC-GA), were combined to identify a pharmacophore group and to predict the antagonist activity of 1,4-dihydropyridines (known calcium channel antagonists) from molecular structure descriptors. To identify the pharmacophore, electron conformational matrices of congruity (ECMC)-which include atomic charges as diagonal elements and bond orders and interatomic distances as off-diagonal elements-were arranged for all compounds. The ECMC of the compound with the highest activity was chosen as a template and compared with the ECMCs of other compounds within given tolerances to reveal the electron conformational submatrix of activity (ECSA) that refers to the pharmacophore. The genetic algorithm was employed to search for the best subset of parameter combinations that contributes the most to activity. Applying the model with the optimum 10 parameters to training (50 compounds) and test (22 compounds) sets gave satisfactory results (R(2)(training)= 0.848, R(2)(test))= 0.904, with a cross-validated q(2) = 0.780).

No abstract

The electron conformational-genetic algorithm (EC-GA), a sophisticated hybrid approach combining the GA and EC methods, has been employed for a 4D-QSAR procedure to identify the pharmacophore for benzotriazines as sarcoma inhibitors and for quantitative prediction of activity. The calculated geometry and electronic structure parameters of every atom and bond of each molecule are arranged in a matrix described as the electron-conformational matrix of contiguity (ECMC). By comparing the ECMC of one of the most active compounds with other ECMCs we were able to obtain the features of the pharmacophore responsible for the activity, as submatrices of the template known as electron conformational submatrices of activity. The GA was used to select the most important descriptors and to predict the theoretical activity of training and test sets. The predictivity of the model was internally validated. The best QSAR model was selected, having r² = 0.9008, standard error = 0.0510 and cross-validated squared correlation coefficient, q² = 0.8192.