Computational modelling of ion channel kinetics under drug intervention on cardiac myocytes

Thesis

The pumping of the human heart is driven by the rhythmic contraction of the heart's muscle cells, the cardiomyocytes. The synchronised contraction of these cardiomyocytes relies on the dynamic flow of ions across the membrane. These ionic flows change the voltage potential across the cell membrane, which forms the action potential. The current conducting through the human Ether-à-go-go-Related Gene (hERG) ion channel repolarises the membrane potential. The inhibition of this current prolongs the action potential and increases the proarrhythmic risk. Since the hERG channel is highly susceptible to blocking by drugs, the degree of this blockage is crucial in assessing the proarrhythmic risk of drugs. However, the drug-binding process and its inhibitory effect on the hERG current are complex and poorly understood. Many researchers have developed mathematical models to capture the channel behaviour under drug blockage and to provide proarrhythmic risk predictions of drugs, but these developments have not led to satisfactory prediction outcomes. The Li model of drug-binding within the hERG channel, developed by research teams in the FDA, incorporated drug-trapping and improved the proarrhythmic risk predictions. This formulation performs as a benchmark for risk prediction models under the CiPA paradigm. It is therefore essential to ensure that the predictions of the Li drug-binding formulation are robust and reliable.

We first perform a sensitivity analysis on the Li drug-binding formulation to investigate the contribution of each parameter to the proarrhythmic risk predicted by the model. We then change the underlying cell models of the Li drug-binding formulation and repeat the analysis. The lack of contribution of the governing parameters of this formulation to model prediction suggests that these parameters cannot be accurately elucidated from experimental data. This is investigated by an identifiability analysis. Utilising the identifiability analysis results, we simplify the formulation by removing the non-identifiable parameter and compare the risk categorisation of the simplified formulations to the Li drug-binding formulation. We show that the simplified formulation is able to provide the same risk classification as the Li drug-binding formulation for drugs from a reference dataset. We end by discussing the importance of this type of in-depth analysis and model evaluation in ensuring the credibility of model predictions.

Authors: Farm, HJ

• DOI: 10.5287/ora-mv9rvgrmb
• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English
• Keywords: drug binding; drug trapping; cardiac drug safety; cardiac modelling; parameter identifiability; hERG channel; parameter sensitivity
• Subjects: Mathematical models; Computational biology