Introduction

In this session, we will learn about coarse-grained (CG) molecular dynamics (MD) simulations. Coarse-graining refers to making a simplified model of a molecular system, e.g., reducing groups of atoms to point masses (``beads''). As a result, systems too large and processes too slow for all-atom MD simulations with current computing resources still can be studied using CG models. Of course, this comes at a price of reduced accuracy.

This tutorial presents one CG method that has been quite successful in a number of applications, termed shape-based coarse-graining (SBCG; see Arkhipov et al., Structure, 14:1767, 2006; Biophys. J., 95:2806, 2008). In this method, a small number of CG beads are used to represent overall shapes of proteins or lipid membranes, with typical ratio of 200-500 atoms per bead. The tutorial introduces tools for SBCG modeling that are provided in VMD as plugins (http://www.ks.uiuc.edu/Research/vmd/plugins/cgtools/).

Figure: Atomic structure of a BAR domain dimer. Left, charged residues are shown in blue (positive) and red (negative). Right, a BAR domain bending negatively charged membrane in an all-atom MD simulation.

For exercises, we will use proteins called BAR domains (Peter et al., Science, 303:495, 2004). BAR domains are $\alpha$-helical bundles capable of forming homodimers, featuring a high density of positively charged residues on their curved surface. Accordingly, they can bind to and bend negatively charged membrane, which makes them key players in membrane remodeling processes in cells. In experiments, multiple BAR domains are often observed acting in concert, forming regular lattices on the membrane surface, which enhances membrane bending. Due to the large size of a lattice involving multiple BAR domains, and long time scales needed to observe membrane bending, all-atom MD falls short of revealing many important characteristics of the process, and one has to employ CG approaches, such as SBCG (Arkhipov et al., Biophys. J., 95:2806, 2008).

Since CG approaches employ greatly simplified models of molecules, one should be very careful in applying them. The SBCG method represents shapes of large molecules consisting of thousands of atoms with a small number of beads, typically 10 to 50. Furthermore, the arrangement of the beads is tuned to reproduce the shape of the molecule, such as that available from an X-ray crystal structure; assemblies of beads representing individual molecules behave as near-rigid, although elastic, bodies. Thus, processes where significant changes in structures of individual molecules may happen, or where fine atomic-level interactions are important, are not good candidates for SBCG studies.

The SBCG model has been designed to permit handling of SBCG structures and simulations in the same way as it is done for all-atom structures and MD simulations. That is, operating in an SBCG representation, one uses VMD and NAMD without any changes in comparison with the all-atom case, and works with the same file types as for all-atom modeling, such as PSF and PDB for structures, and topology, parameter, and configuration files for simulations (see VMD and NAMD tutorials, http://www.ks.uiuc.edu/Training/Tutorials/). However, these SBCG PSF, PDB, parameter and topology files first need to be created according to the all-atom model that one desires to coarse-grain. In this tutorial, we will learn how to use SBCG plugins of VMD to build such files, and to refine parameters of the SBCG models for subsequent simulations.

One should keep in mind that with the sizes of systems that are typically addressed using SBCG models, limitations of PDB/PSF file formats (how large a number the file can hold in the coordinate, mass, and charge fields) may easily become an issue. For example, the coordinate field in PDB files only allows values less then $10,000$Å, or 1$\mu$m. If a larger system is considered, in cases when PSF/PDB files are necessary one should scale (diminish) coordinates to reduce the system size, while for actual NAMD simulations or for work with VMD one should use binary file formats, such as those of .coor or .dcd formats, which do not have size limitations. In this tutorial, we will not have such a problem as we consider a relatively small system for the exercises.