Evaluation of computational modelling as a preclinical proarrhythmic safety assay

Abstract

Many cardiac ion channels are prone to reduced current flow following blockade by pharmaceutical compounds. High throughput screening technologies now allow us to establish this reduction in activity routinely and relatively cheaply. It would be advantageous to integrate this information to provide an indication of the likely affect of a compound at the whole cell, or indeed, whole organ level. Computational cardiac electrophysiology models offer this opportunity, by expressing the complex feedback between ion currents and transmembrane potential in mathematical form. To date, a thorough evaluation of the success of mathematical models in predicting later safety test results has been lacking. In this work we present the results of a trial undertaken at GlaxoSmithKline R&D, in which we evaluate mathematical model predictions of the results of a rabbit ventricular wedge experiment, for hundreds of compounds. We fit dose–response curves from IonWorks and PatchXPress for INa, ICaL, IKr and IKs, enabling us to calculate a ‘scaling factor’ for the maximum level of each of these currents upon application of a given compound concentration. We then scale the mathematical models of these currents accordingly in a simulated rabbit myocyte, and examine the predicted prolongation of the cellular action potential, comparing this with prolongation of the experimentally recorded QT interval. We present the open-source computational tools that we have developed to perform this study, and the sensitivity and specificity of the computational model outputs. The results suggest that the replacement of some animal tissue experiments with simulation is possible in the near future.