Dealing with lots of uncertainties - identifiability, model structure, experimental artefacts and model discrepancy – in predictions from ion channel models

Today our field continues to use the same approach to voltage-clamp experimental design that Hodgkin & Huxley used in the 1940s, in terms of designs that enable model parameter values to be estimated manually from graph paper. We have been developing short, high-information voltage clamp protocols to characterise ion currents. I will discuss various rationales for designs that consider parameter identifiability, model selection and minimising experimental artefacts. We then use computational optimisation to fit simple mathematical models for hERG/IKr to the resulting experimental currents, and use it to predict the results of conventional voltage clamp protocols and physiological action potential clamps. Our latest work also uses an additional mathematical model to account for patch clamp artefacts to consolidate information from different patch clamp recordings more reliably, as well as an ensemble-of-experimental-designs approach to get a handle on the likely size of model discrepancy.