Deep learning model of myofilament cooperative activation and cross-bridge cycling in cardiac muscle

Cardiac muscle contraction is driven by the cross-bridge cycle, where myosin heads generate force by cyclically attaching to and pulling on actin filaments using energy from ATP. Modeling this process is central to understanding cardiac sarcomere mechanics. In this study, we developed supervised machine learning (ML) models using artificial neural networks (ANNs) to simulate cross-bridge cycling and muscle behavior under isosarcometric, isometric, and isotonic conditions. Trained on synthetic data, the ANN captured nonlinear dependencies among calcium concentration, stiffness, sarcomere length, temperature, and force output. Error analysis through histograms and unity-line scatterplots validated prediction accuracy and identified underfitting and overfitting patterns. Comparisons across ANN architectures showed how hidden layer complexity affects model generalization. The present deep learning models accurately reproduced key physiological behaviors, including steady-state force–Ca2+ relations, sarcomere length changes, and force–velocity relations, and matched theoretical results. This work demonstrates the potential of ML tools to enhance cardiac muscle modeling and exploit existing experimental datasets for improved prediction of cardiac muscle diseases.