QSAR STUDIES ON DERIVATIVES OF QUINAZOLINE-4(3H)-ONES WITH ANTICONVULSANT ACTIVITIES

Abstract

Quantitative structure-activity relationship study was done on some quinazoline-4(3H)-ones derivatives with anticonvulsant activity against maximal electroshock-induced seizure. The quinazoline derivatives used as dataset and their anticonvulsant activity value were obtained from the literature. The molecular structure of the dataset compounds was generated with Spartan 14 software. This was optimized with PM3 semi-empirical quantum mechanical method available in the software. Molecular descriptors were obtained from the optimized structures using the PaDEL-Descriptor software. Activity values of the compounds and molecular descriptors obtained from the optimized structure made up the database for the study. The database was divided into training and test sets with Kennard Stone algorithm. Genetic function algorithm was used to develop quantitative structure-activity relationship models. The best model obtained was stable, robust and had good statistical parameters including determination coefficient R2 (0.899), adjusted determination coefficient R2adj (0.888), variance ratio F (82.03), leave one out cross-validated determination coefficient Q2 (0.866) and predicted determination coefficient for the test set R2pred (0.7406). The model indicated that the anticonvulsant activity of the studied compounds was dependent on Broto-Moreau autocorrelation-lag2/weighted by Vander Waals volume (ATS2v), average coefficient sum of the last eigenvector from Barysz matrix/weighted by Vander Waals volume (VE2_DZv), largest absolute eigenvalue of Burden matrix-6/weighted by relative atomic mass (SpMax6_Bhm), average valence path of order 6 and radial distribution function at 4.5 interatomic distance weighted by first ionization potential (RDF45i).