Biophysical Society Thematic Meeting | Hamburg 2022

Biophysics at the Dawn of Exascale Computers

Poster Abstracts

21-POS Board 21 A SPATIAL DECOMPOSITION APPROACH TO MARKOV MODELING Tim Hempel 1,2 ; Frank Noé 1,2,3 ; 1 Freie Universität Berlin, Department of Mathematics & Computer Science, Berlin, Germany Modeling the dynamics of ever-larger biological systems is becoming increasingly relevant to our understanding of molecular cell biology. The combination of molecular dynamics (MD) simulations with classical Markov state models (MSMs) or related deep learning techniques such as VAMPnets have proven important tools in that regard, providing access to simplified models of molecular kinetics. Despite their successes, these approaches are inherently limited by the size of modeled systems. As larger macromolecular complexes come with an increased number of weakly coupled subsystems, the number of combinatorial global states grows exponentially, hampering our efforts to sample all distinct global state transitions. Here, we leverage weak couplings between subsystems to estimate a global kinetic model without requiring the sampling of all combinatorial global states. First, this approach is implemented with classical MSMs, yielding a method we term independent Markov decomposition (IMD). Using the example of empirical few-state MSMs of ion channel models, we show that IMD can reproduce experimental conductance measurements with a major reduction in sampling when compared with standard MSMs. Second, IMD is combined with VAMPnets, forming an end-to-end deep learning framework that automatically decomposes a system into subdomains, while simultaneously learning individual subsystem Markov models (iVAMPnets). We show that iVAMPnets can successfully model high-dimensional MD data for the synaptotagmin C2A domain, even when the sampling is insufficient for standard VAMPnets. Both methods, IMD and iVAMPnets, strive to provide data-efficient and easily interpretable models of molecular kinetics that are applicable to highly complex biological systems. 2 Freie Universität Berlin, Department of Physics, Berlin, Germany 3 Rice University, Department of Chemistry, Houston, TX, USA

94