Gradient-based parameter optimization method to determine membrane ionic current composition in human induced...

Area of Science:

• Cardiology
• Computational Biology
• Stem Cell Biology

Background:

• Human induced pluripotent stem cell-derived cardiomyocytes (hiPSC-CMs) exhibit heterogeneous action potentials (APs) due to varying membrane ionic current expression.
• Accurate determination of ionic channel function is crucial for understanding and modeling cardiac electrophysiology in hiPSC-CMs.
• Previous research suggests that identical AP configurations might arise from different combinations of ionic currents, posing challenges for inverse problem-solving.

Purpose of the Study:

• To develop and computationally validate a method for determining whole-cell ionic conductance (G_x) patterns from individual spontaneous AP configurations in hiPSC-CMs.
• To assess the feasibility and reliability of a gradient-based optimization method for estimating G_x values.
• To investigate the potential for multiple solutions (local minima) in the inverse problem of G_x estimation.

Main Methods:

• Generation of 'cell-specific models' by fitting a baseline model's AP output to experimental APs from hiPSC-CMs.
• Randomization of G_x values for 4-6 major ionic currents within a ±5-15% range to create initial parameter sets.
• Application of a gradient-based optimization algorithm to recover G_x values by minimizing the mean square error (MSE) between target and model APs.
• Global search space analysis by randomizing G_x values over a 0.1-10 fold range to identify potential local minima in MSE.

Main Results:

• The gradient-based optimization method demonstrated progressive convergence of randomized G_x populations towards the original cell-specific model values as MSE decreased.
• Mapping the MSE across the global parameter space revealed a single global minimum, indicating the absence of other significant local minima.
• The computational approach successfully recovered accurate ionic conductance values, confirming the uniqueness of the solution within the tested parameter space.

Conclusions:

• The developed gradient-based optimization method is a feasible and reliable tool for determining ionic conductance patterns from hiPSC-CM action potentials.
• The study confirms that for the tested models, the inverse problem of G_x estimation has a unique solution, mitigating concerns about non-identifiability.
• This method provides a robust approach for characterizing ionic current expression in hiPSC-CMs, advancing their utility in disease modeling and drug screening.