Can a Reduced Dimension Interface Model be More Computationally Effective than a Docking Algorithm?

Abstract

Actomyosin dynamics are central to elemental cellular processes, yet the precise molecular mechanisms by which actin and myosin transform ATP's chemical energy into mechanical work remain a mystery. Further advances in understanding the actomyosin machine require computational, molecular-level models of the structural dynamics of myosin and its actin complexes, which correlate multi-domain structural changes with force/movement measurements and biochemical kinetics, to serve as adjuncts to molecular-level experiments.Existing modeling approaches, e.g., molecular dynamics which provides atomic-scale simulations of crystal structures, statistical mechanics which provides molecule-size extrapolations of experimental results, lack either the computational speed or the structural resolution necessary to capture the details necessary to reveal critical dynamics of supramolecular assemblies like actomyosin. Novel modeling approaches using targeted coarse-graining of high-resolution crystal structures and interaction potential fields may offer unified, mechanically rigorous, yet computationally efficient method to model supramolecular assemblies, as well as to develop model adjuncts for experiments.While a core component in any such model is adequately depicting protein-protein binding-domain dynamics, key structural transitions cannot be defined by static crystal structures, as they involve dynamically disordered states. As a result, development of reasonable, inter-molecular potential functions requires estimations of protein-protein interactions, which could be and typically are obtained from docking algorithms like ZDOCK/RDOCK or AutoDock that, depending on the specific proteins, can be computationally expensive. In this work, the conjecture that a Reduced Dimension Interface Model (RDIM), which is constructed using multipole potential expansions, is sufficiently accurate, yet computationally more efficient_and perhaps even better suited for supramolecular assembly dynamics modeling_than standard docking algorithms. Here, the relative efficacy, merits and demerits of RDIMs for estimating protein-protein interactions examined in the context of two key binding domains, namely myosin:actin and G-actin:G-actin.