Uncertainty quantification and sensitivity analysis of cardiac cell models using Gaussian process emulators

by Professor Richard Clayton
Summary: Cardiac cells are electrically excitable, and the biophysical mechanisms underlying electrical activation and recovery can be represented by systems of couple, stiff and nonlinear ordinary differential equations. These models have become valuable tools, which are used to study normal and abnormal electrical activation in cardiac tissue as well as the response of cardiac cells to drug interventions. However, the models embed large numbers of parameters that are assigned single values based on experimental data, which leads to two problems. First, the electrical activation and recovery in real cardiac cells varies from one beat to the next, and from one cell to another. Second, the models are highly detailed, and so the sensitivity of model outputs to changes in parameters is difficult to establish without running large numbers of simulations. We have addressed these problems by using Gaussian processes to represent the outputs of cardiac cell models as a function of the model parameters. This approach allows model parameters to be assigned a distribution rather than a fixed value, and enables a systematic sensitivity analysis to be undertaken.