From fruit flies to humans, epileptic seizures have been described in virtually all known neuronal model systems. They represent hypersynchronous electrical activity that disrupts normal brain function. Whilst an epileptic seizure is a phenomenon of integrated neuronal networks, identified causes for epilepsy range from molecules (‘epilepsy genes’), to whole-brain abnormalities (‘epileptogenic networks’). Computational models can help linking these different levels of explanation by allowing us insights into otherwise unobservable dynamics. One example approach is using Bayesian model inversions to fit interconnected neuronal population models to electroencephalography (EEG) measures of abnormal neuronal dynamics in epilepsy. Here I will illustrate two worked examples of how this approach can be applied to clinically relevant questions: by (1) offering insights into the epileptogenic effects of a specific molecular disruption of neuronal function (NMDAreceptor antibodies), and (2) making predictions of the epileptic brain’s response to perturbations (from intracranial recordings in patients undergoing epilepsy surgery).