Here,

[C] is the concentration, Selleck INCB018424 and there are two parameters: [IC50], the half-maximal inhibitory concentration; and the Hill coefficient n. In previous work ( Beattie et al., 2013 and Elkins et al., 2013) we found little benefit, if not just additional uncertainty, in considering the Hill coefficients from these data sources; so in this study we assume that n = 1, and fit IC50 values only. We use three of the latest human ventricular action potential models — ten Tusscher and Panfilov (2006), Grandi, Pasqualini, and Bers (2010), and O’Hara, Virág, Varró, and Rudy (2011). These models were chosen as they are all candidates for use in in-silico action potential modelling for cardiac safety, and it will be valuable to examine the consistency of their predictions. The ten Tusscher and Panfilov (2006) Selleck Linsitinib model contains a limited number of differential equations (17) and outer membrane currents (12), and is a refinement of the ten Tusscher, Noble, Noble, and Panfilov (2004) model. The model was developed to provide realistic conduction velocity restitution and to integrate the latest human data at the time. It has been very widely used for a range of studies

and has proved robust: making good predictions in a number of situations. The Grandi model is a human-tailoring of the Shannon, Wang, Puglisi, Weber, and Bers (2004) rabbit model, which features detailed calcium handling. It aimed to improve the balance of repolarizing potassium currents, and to capture reverse-rate dependence of IKr block. This model is more complex than ten Tusscher, with 14 outer-membrane currents many of which are divided into two for the cleft and bulk sarcolemmal spaces. There are a correspondingly below larger number of equations (39). The O’Hara model is a more recent human ventricular model, much of it was built ‘from scratch’ using data from human hearts. The O’Hara et al. (2011) paper shows improved APD dependence on pacing

rate in this model relative to the others. This model has 41 differential equations, again there are 14 types of outer membrane currents, including late sodium. Having been parameterised by different datasets, these models may represent some of the underlying variation between cells, locations in the heart, or indeed individuals, that could be reflected in the variation observed in the TQT study. In Fig. 2 we show basic properties of these models, in terms of response to blockade of certain ion channels, at steady 1 Hz pacing.1Fig. 2 highlights some differences between model behaviours. On the top row we see that the O’Hara model responds more dramatically to block of IKr than the other models: the action potential becomes markedly prolonged, and at 100% IKr block the cell fails to repolarise and remains at depolarised potentials. In contrast, the ten Tusscher model shows a large prolongation under IKs block, whereas the other models show little response.