Generation of a new QSAR model for anticonvulsant activity and its application in virtual screening

A. TALEVI; J.J. PRIETO; E.A. CASTRO; L.B. BLANCH

Int. J. Appl. Chem.

Año: 2007

Abstract.- A QSAR model for the prediction of anticonvulsant activity mediated by sodium channels blockade was generated from topological descriptors. The stability and prediction ability of this QSAR function were evaluated, with good results, through calculation of statistical parameters, cross-validation by use of the leave-one-out technique, randomization and, finally, external validation. Afterwards, the model was applied in a virtual screening study for the selection of the most promissory compounds among a collection of 119,000 molecular structures. A second selection criteria based on the optimal value of the octanol-water partition coefficient for anticonvulsant drugs was applied, leading to a final group of 45 promissory compounds representing only a 0.04 percent of the group on which the in silico screening was performed. Subsequent bibliographic search has shown that some of these compounds have demonstrated anticonvulsant activity in the past, whereas others are structurally close-related to drugs with verified anticonvulsant activity.