BACKGROUND: Electrical brain stimulation is recognized as a promising therapeutic approach for treating brain disorders such as epilepsy. However, the use of this technique is still largely empirical, since stimulation parameters and targets are chosen using a trial-and-error approach. Therefore, there is a pressing need to design optimal, rationale-based multi-site brain stimulation protocols to control epileptiform activity. APPROACH: Here, we developed biologically-inspired models of brain activity receiving stimulation at two levels of description (single- and multi-population epileptogenic networks). First, we used bifurcation analysis to determine optimal parameters able to abort epileptiform patterns. Second, we present a graph-theory based method to classify network populations in an epileptogenic network based on their contribution to seizure generation and propagation. MAIN RESULTS: The best therapeutic effects (i.e., reduction of epileptiform discharges duration and occurence rate) were obtained by the specific targeting of populations with the highest eigenvector centrality values. The timing of stimulation was also found to be critical in seizure abortion impact. SIGNIFICANCE: Overall, our results provide a proof-of-concept that using network neuroscience combined with physiology-based computational models of brain activity can provide an effective method for the rational design of brain stimulation protocols in epilepsy.