Sensitivity and Uncertainty Analysis of Two Human Atrial Cardiac Cell Models Using Gaussian Process Emulators

Area of Science:

• Computational biology
• Cardiac electrophysiology
• Mathematical modeling

Background:

• Biophysically detailed cardiac cell models are essential for understanding cardiac myocyte function.
• Model complexity and non-linearity make parameter-output relationships difficult to ascertain.
• Quantifying parameter uncertainty and variability requires extensive model evaluations.

Purpose of the Study:

• To demonstrate the utility of Gaussian process emulators for systematic sensitivity and uncertainty analysis of cardiac cell models.
• To quantitatively analyze the Courtemanche and Maleckar human atrial action potential models.
• To compare mechanisms between different cardiac cell models.

Main Methods:

• Gaussian process emulators were trained on existing cardiac cell model data.
• Variance-based sensitivity indices were calculated to identify influential parameters.
• A two-stage analysis approach was employed, refining emulators with sensitive parameters.
• Comparison of model mechanisms was facilitated by second-stage sensitivity indices.

Main Results:

• Gaussian process emulators effectively captured action potential and calcium transient dynamics.
• Sensitivity analysis revealed minimal interaction effects between model parameters.
• The relationship between L-type Ca2+ current and action potential plateau was quantified.
• Opposite effects of ultra-rapid K+ channel conductance on action potential duration were predicted and confirmed.

Conclusions:

• Gaussian process emulators are effective tools for sensitivity and uncertainty analysis in biophysically detailed cardiac cell models.
• This approach enables quantitative comparison of mechanisms across different models.
• The findings facilitate a deeper understanding of cardiac electrophysiology and model behavior.