Molecular modeling and simulation have come to play an indispensable role in both interpreting experimental results as well as advancing the discovery and understanding of biological phenomena. Historically, the main challenges in applying the principles of molecular modeling to systems of biological relevance have been the poor scaling of these techniques and lack of computational power. The multiscale challenge, moreover, is a problem of dimensionality in both length and time scales. For example, the most advanced all-atom molecular simulations are on the order of 2-3 million atoms with a simulation length of up to 100 ns – far short of time and length scales investigated in most biological experiments. As many biological phenomena take place on disparate orders of magnitude, it is of the utmost importance to adopt a multiscale approach. This poster is a presentation of recent research efforts to address both these challenges with application to the study of the actin cytoskeleton.

The structure, properties and function of the cytoskeleton are determined in large part by the matrix of actin filaments, fibers and bundles that constitute the makeup of cytoskeletal architecture. Recently it has been determined that the directed assembly and growth of the network of actin filaments are responsible for cell motility. A key component of this network is the Arp2/3 complex, whose major function is to serve both as a nucleation site for new actin filaments and as an anchor/branch point between existing filaments. We have performed molecular dynamics (MD) simulations of the isolated Arp2/3 complex and probed the interaction between Arp2/3 and full actin filaments. In turn, results obtained from these simulations have been translated into coarse-grained (CG) models that are able to propagate atomic-level details to meso- and larger length scales.

The second part of this poster presents an investigation of the folding of an important alpha-helical domain within the G-actin protein. Such computational investigations present unique challenges given that many protein folding events occur on far longer timescales than those accessible by traditional molecular simulations. Transformation between the folded and unfolded states requires the crossing of a free-energy barrier that is much larger than kBT, and therefore extremely unlikely to be observed during a classical MD simulation. To address these challenges, we have used the method of metadynamics, which is able to bias a simulation along a set of collective variables that offer a reduced description of the process of interest. Using metadynamics allows one to faithfully reconstruct the free-energy surface for the folding event, and to obtain, among other things, estimates of the free-energy differences between the folded and unfolded state as well as an estimate of the free-energy barrier for folding.