Biophysical Society Thematic Meeting | Canterbury 2023

Towards a More Perfect Union: Multi-Scale Models of Muscle and Their Experimental Validation

Monday Speaker Abstracts

MODEL BUILDING CHALLENGES AND THE TIME-SCALE PROBLEM IN THE COMPUTATIONAL STUDY OF MUSCLE BIOPHYSICS. Steven D. Schwartz ; 1 The University of Arizona, Chemistry and Biochemistry, Tucson, AZ, USA The molecular machines of muscle present some standard and some unique challenges to theoretical investigation. The first challenge is encountered when there is no extant atomic structure. We faced this years ago in the cardiac thin filament when only fragments and the troponin core had any published structures. Docking and molecular minimization allowed us to create models of fully solvated thin filament. New cryo-em structures will help to alleviate such problems, but this technology is unable to find average structures and range of motion for mobile regions. In this case our collaboration with Jil Tardiff’s group has proved transformative. Given a structure, however; the challenges of developing usable computational approaches remain. The number of atoms in a fully solvated model of just the cardiac thin filament is in the multiple millions. Because an atomistic simulation is required to accurately evolve the state for the fastest timescale in the system accessible times are limited to nanosecond to microsecond times even on the fastest GPU accelerated processors – clearly far less than any biologic timescale. Our group has developed and applied enhanced sampling methods to overcome the timescale problem – examples of this are transition path sampling for reactions such as the hydrolysis of ATP by myosin and metadynamics for conformational transitions. We will present such methods and results. A final approach we have employed is to use short time simulations (nanosecond) to compare wildtype to mutant thin filament complexes. This is done with the understanding that the overall structure of the thin filament must be stable. Thus, relatively short timescale simulations are expected to allow inference of the impact of mutation. Lately we have coupled this approach with machine learning algorithms to make predictions of pathogenicity of variants of unknown significance.

31