Discovering Cardiac Action Potential Model Equations Using Sparse Identification of Nonlinear Dynamics.

Area of Science:

• Computational Biology
• Mathematical Modeling
• Cardiac Electrophysiology

Background:

• Cardiac action potential (AP) models are crucial for understanding heart function.
• Identifying accurate equations and parameters for AP models remains a significant challenge.
• Existing models often struggle to precisely match experimental data.

Purpose of the Study:

• To evaluate the effectiveness of the Sparse Identification of Nonlinear Dynamics (SINDy) approach for modeling cardiac action potentials.
• To assess SINDy's ability to identify differential equations and parameters from synthetic AP data.
• To determine SINDy's performance with various cardiac-specific models and complex dynamics.

Main Methods:

• Utilized SINDy, a sparse regression technique, to identify differential equation models from synthetic AP data.
• Applied SINDy to two-variable polynomial models, including the FitzHugh-Nagumo (FHN) model and its cardiac variants.
• Tested SINDy's capability in fitting data with time-dependent stimulus currents and alternans dynamics.

Main Results:

• SINDy successfully reproduced the underlying equations for all tested FHN-based cardiac models.
• Cardiac variants showed higher sensitivity to parameter choices and optimizer settings compared to the baseline FHN model.
• SINDy demonstrated proficiency in identifying models even with the inclusion of time-varying stimulus currents.

Conclusions:

• SINDy is a promising data-driven method for developing accurate and parsimonious cardiac AP models.
• The approach offers a robust framework for matching differential equations to experimental cardiac electrophysiology data.
• SINDy provides a valuable tool for balancing model complexity and predictive accuracy in cardiac modeling.