Detailed models of the cardiomyocyte action potential (AP) have been used as the basis for multi-scale investigations into the healthy heart, arrhythmias, cardiomyopathies, and the effects of drugs or genetic mutations. Recently, AP models have been used predictively, for example to estimate drug effects and assist in risk stratification. If AP models could be personalized, such studies could be repeated for individual patients and form the basis for diagnosis and tailored treatment.
For example, models could be personalized by incorporating differences due to gender and age, the effects of known mutations/polymorphisms, or changes in ionic balance due to disease. Stem-cells, coerced into becoming cardiomyocyte-like `hiPSC-CMs', could be created from patient donor-cells so that personalized electrophysiological measurements could be made. But to make confident, clinically useful predictions, there are still major difficulties to overcome.
Using a recent case study on IKr, we discuss how biological variability must be separated from uncertainty and measurement error, and how novel voltage-clamp protocols - developed through close experimenter-modeler interaction - can help. Secondly, we show how gathering a large electrophysiological data set can reveal a diversity that, at first glance, far exceeds the small changes typically introduced when models are tailored. Thirdly, using results from an open online resource (The Cardiac Electrophysiology Web Lab), we show how published models show a similarly wide range of predictions, making it hard to choose the appropriate model to personalize. Finally, we discuss how current models are not species-specific, let alone patient-specific, and we propose community-driven strategies to explore the full extent of this problem and ultimately address it.
By discussing these four cases in the context of the wider personalized medicine debate, we demonstrate key challenges for the use of personalized AP models and the crucial role of collaboration, open software, and shared data resources.