Assessing the potential for novel compounds to increase pro-arrhythmic risk is a high priority for pharmaceutical companies. A recent initiative led by the US Food & Drug Administration aims to streamline this process and provide more accurate predictions based on in-vitro technologies rather than clinical trials. The Comprehensive in-vitro Pro-arrhythmia Assay (CiPA) [1] aims to link high-throughput automated ion channel patch-clamp screening in expression systems for multiple ion channel targets; stem-cell derived myocyte assays; and mathematical models of electrophysiology to understand whether results are consistent and extrapolate to the adult human situation [2]. In previous work we showed how information on multiple ion channels could improve prediction of human clinical risk [3]. Recently we have been studying how uncertainty in ion channel screening [4], and choice of mathematical model, can influence simulation predictions. Here we show our latest results examining the variability that is inherent in ion channel screening, and how this might be quantified. Once the variability is described by a probability distribution it can be propagated through simulations of electrophysiological activity to provide a probability distribution of outputs. We demonstrate how this propagation works, and how a simpler model (or emulator) can be used to speed up this process. We demonstrate the effect of this uncertainty using a dataset of ion channel interactions for reference (marketed) compounds of known risk [5], and show how this affects the conclusions we can draw about how dangerous a novel compound is likely to be.