Mathematical action potential (AP) models describe the changes in the membrane voltage due to a complex interplay between ionic currents, and their interactions with drug compounds. Therefore, these models can guide preclinical risk assessments for drug-induced cardiac arrhythmia, and extract more information from animal-based experiments. The rabbit Purkinje fiber has been used in preclinical studies, as it includes the major currents present in human ventricular myocytes. A recently proposed mathematical AP model of the rabbit Purkinje fiber, combined with ion channel screening data, predicted drug effects on AP changes, with an agreement of up to 80%. To explain the 20% mismatch, we first attempt to improve the AP model by calibrating its parameters to fit control AP traces and their changes in the presence of reference drug compounds with well-studied channel block properties. Subsequently, by fitting AP changes in the presence of a new set of compounds we infer block effects on various ion channels. We are currently validating our method in terms of uncertainty quantification of the control parameters and prediction of previous ion channel screening results. The aim is finally to perform an experimental test for any inferred block which has not been measured before, particularly for those drug compounds whose action potential changes are not explained by the existing ion channel screening data.