Epilepsy is a chronic neurological disorder characterized by recurrent seizures due to excessive neuronal activity. To date, there are about 30 drugs available for the treatment of epilepsy; however, approximately one-third of the patients do not achieve sustained seizure-free status with adequately chosen antiseizure medications. For this reason, the search for new therapies remains a priority. Computational tools have been extremely useful for the design and discovery of new drugs, reducing costs and time spent in the identification of novel molecular bioactive scaffolds. In this work, ligand-based computational models were developed and validated to be used in virtual screening in order to identify new drugs with promising antiseizure activity in the PTZ kindling model. Training data for the models were obtained from specialized literature, and were then representatively sampled using an in-house clustering procedure (iRaPCA). Linear classifiers based on conformation-independent molecular descriptors were generated using in-house Python routines that combine feature bagging and forward stepwise feature selection. The best classifiers obtained were combined into meta-classifiers and validated by retrospective screening experiments. Finally, the best model ensemble was applied to screen the chemical libraries DrugBank 5.1.8 and Drug Repurposing Hub (DRH), to detect potential drug repurposing opportunities for possible active drugs in the PTZ kindling model.