A Combined Virtual Screening 2D and 3D QSAR Methodology for the Selection of New Anticonvulsant Candidates from a Natural Product Library

A virtual screening methodology combining a discriminant function (df) based on Dragon 2D-descriptors, general ADME filters, and a pharmacophore identified through superposition of rigid analogs was applied in order to identify new anticonvulsant agents among 10 903 natural products. Fifty-six compounds were selected from the application of this df, Lipinski's rule-of-five and other criteria for prediction of oral bioavailability, and the optimal value of log P for a compound to diffuse passively through the blood–brain barrier. Seven of these compounds were removed because they did not belong to descriptor space defined by the training set. The remaining 49 compounds were fitted to the pharmacophore structural constrains, selecting 7 structures with RMS distance value below 0.2. Systematic conformational analysis was performed to find the global minimum conformation, using the PM3 base included in Hyperchem 6.03. The global minimum conformations were further refined at a 6-31G** level with Gaussian 03. Restricted optimization was then performed to determine the energy difference between the energy minimum conformation and the active conformation defined by the pharmacophore, retaining those structures whose energy differences were below 7 kcal/mol. Four new potential anticonvulsants which fulfill the df and the pharmacophore requisites were selected. One of these structures, 2-(4,6-dimethyl-1-benzofuran-3-yl)acetic acid, was acquired and assayed in the MES and Rotorod tests, confirming anticonvulsant activity in mice at 30 and 100 mg/kg (0.5 and 4 h, i.p.) and absence of neurotoxicity at the tested doses.