Deconvoluting the Electrophysiological Signatures of Myocardial Ischemia using a Validated Machine Learning Framework

Area of Science:

• Computational biology
• Cardiovascular physiology
• Machine learning in medicine

Background:

• Myocardial ischemia involves complex changes like hyperkalemia, acidosis, and ATP depletion, altering cardiomyocyte electrophysiology.
• Discerning the individual impact of these factors on action potential (AP) changes is challenging.
• This study investigates machine learning's ability to differentiate these distinct ischemic patterns within a single AP.

Purpose of the Study:

• To develop and validate a machine learning model capable of deconvoluting distinct ischemic drivers affecting cardiomyocyte electrophysiology.
• To predict key physiological parameters like extracellular potassium ([K+]o) and intracellular pH (pHi) from action potentials.
• To assess the generalizability and robustness of the machine learning model across different computational frameworks.

Main Methods:

• A multi-target regression model was developed using the Luo-Rudy (1991) computational model of a ventricular cardiomyocyte.
• The model was trained to predict [K+]o and pHi under simulated ischemic conditions.
• The model's performance was evaluated on a test set and generalized to the Ten Tusscher (2006) framework.

Main Results:

• The model achieved high accuracy, with low mean squared errors for predicting [K+]o and pHi.
• The model accurately predicted [K+]o and pHi when applied to the Ten Tusscher (2006) model, demonstrating robustness.
• Feature importance analysis identified resting membrane potential (RMP) as the key predictor for [K+]o and action potential duration (APD) for pHi.

Conclusions:

• The developed machine learning approach successfully distinguishes individual ischemic drivers (hyperkalemia, acidosis) from action potential changes.
• This method offers potential for in silico drug screening and mechanistic analysis of myocardial ischemia.
• The findings highlight the distinct electrophysiological signatures associated with different ischemic factors.