Abstract
The intercalated disk (ID) is a structurally heterogeneous junctional complex essential for synchronized cardiac conduction and contraction. Previous computational models have investigated the influence of ID structure on cardiac conduction. However, most have relied on oversimplified geometries and uniformly distributed ion channels, limiting their ability to capture nanoscale heterogeneity. In this study, we expand our previous finite element mesh framework to produce a more physiologically realistic representation of the ID, incorporating spatially heterogeneous gap junctions and multiple ion and ionic current dynamics. We systematically quantify the impact of key structural and electrophysiological features on conduction by generating a comprehensive library of 384 ID mesh configurations and simulating tissue-level conduction for both strong and reduced gap junctional coupling. Further, we employed a multilayer perceptron neural network approach to quantify gradient-based sensitivity analysis, enabling a systematic quantification of the relative influence of geometric and nanostructural factors on ID and cleft dynamics, as well as tissue-level conduction across multiple regimes. In particular, sensitivity analysis revealed that gap junctional coupling, cleft geometry, and nanostructure heterogeneity are the dominant determinants of cleft potential, sodium current synchronization, and conduction velocity. We identify that membrane separation of the ID interplicate and plicate regions can exhibit context-dependent influences on conduction, either enhancing or slowing, depending on gap junctional coupling. Collectively, these findings highlight the regime-dependent roles of ID ultrastructure and establish a quantitative framework that links nanoscale ID morphology to tissue-scale cardiac conduction.