Moltemplate: A Tool for Coarse-Grained Modeling of Complex Biological Matter and Soft Condensed Matter Physics

Abstract

Coarse-grained models have long been considered indispensable tools in the investigation of biomolecular dynamics and assembly. However, the process of simulating such models is arduous because unconventional force fields and particle attributes are often needed, and some systems are not in thermal equilibrium. Although modern molecular dynamics programs are highly adaptable, software designed for preparing all-atom simulations typically makes restrictive assumptions about the nature of the particles and the forces acting on them. Consequently, the use of coarse-grained models has remained challenging. Moltemplate is a file format for storing coarse-grained molecular models and the forces that act on them, as well as a program that converts moltemplate files into input files for LAMMPS, a popular molecular dynamics engine. Moltemplate has broad scope and an emphasis on generality. It accommodates new kinds of forces as they are developed for LAMMPS, making moltemplate a popular tool with thousands of users in computational chemistry, materials science, and structural biology. To demonstrate its wide functionality, we provide examples of using moltemplate to prepare simulations of fluids using many-body forces, coarse-grained organic semiconductors, and the motor-driven supercoiling and condensation of an entire bacterial chromosome.

Keywords: LAMMPS; coarse-grained simulation; molecular dynamics; molecular modeling.

Figure 1.. Coarse-grained physics-based models of chromosomes and organelles.

a) Coarse-grained DNA represented with three particles every 42 bp, including a dummy particle to represent the local superhelical state. A rotating motor (white) applies a constant torque to 4 particles. 219 lines of text files were required to implement this example, including 79 lines of moltemplate files, 43 lines to invoke the polymer generator and insert a twist motor, 32 lines of minimization and run protocols, and a short Python script to generate tabulated potentials. b) Predicted conformation of the entire genome of Caulobacter crescentus (4Mbp) in the absence of DNA-binding proteins, created by relaxing, twisting, and compressing a circular polymer that was originally stretched while confined in a tube of radius 320nm. Bottle-brush-like supercoils form as a result of maintaining the polymer at constant torsional tension. This example was implemented using 397 lines of text. c) Detail of a large, highly-branched plectonemic supercoil (10kbp). d)-e) Simulating the formation of a lipid bilayer using the MARTINI force field. This example contains 300 lipids, 6000 waters, and requires 220 lines of text. f) A liposome with protein inclusions containing 120 proteins, 65000 lipids, and implemented with 544 lines of text (including PACKMOL files). PACKMOL was used to randomize the molecular positions, and moltemplate was used to assemble the LAMMPS simulation files. Files for these examples can be downloaded at http://doi.org/10.5281/zenodo.4392267. Systems visualized using VMD and TopoTools.

Figure 2.. Files needed to prepare and run a LAMMPS simulation with moltemplate

a) The “propane.lt” file contains the definition of a coarse-grained “Propane” molecule containing 3 particles, 2 bonds, and one angular spring. b) A similar file defining a coarse-grained “Water” molecule includes a text block (“params_sw.txt”) describing parameters for its more complicated (many-body) force-field. c) These molecule objects can be used as building blocks to create text files describing more complex systems. Here, two “new” commands create a water-hydrocarbon mixture containing 4^3 = 64 Propanes and 12^3 = 1728 Waters. Text enclosed in each “write(FILENAME)” statement will be appended to the generated file (eg. “Data Bonds”) each time a copy of the molecule is created, and the counter variables ($atom:, $bond:, $angle:) will be replaced by integers and incremented. However, type variables beginning with @ are not incremented. Interactions between water and propane are also specified. LAMMPS files generated by moltemplate are shown in e), with rectangles enclosing the portion of text generated by each molecule copy. d) Command used to run moltemplate. The optional “-atomstyle” argument customizes particle attributes. e) Files generated by moltemplate. Note: Coordinates are modified by the move() commands in part c). “Data” files are concatenated together by moltemplate and renamed “system.data”. f) Command used to run LAMMPS g) The “run.in” file is a LAMMPS file containing links to the files generated by moltemplate as well as LAMMPS-specific run-time settings.

Figure 3.. Integrated pipeline for building models of entire bacterial cells.

3D structures for the ~500 types of proteins in a mycoplasma proteome are curated in the online tool Mesoscope and used to create interactive draft models in cellPACKgpu. Moltemplate, with the utility program cellPack2moltemplate, then converts molecular location and orientation information into a LAMMPS input file, to perform a coarse-grained simulation that eliminates steric clashes and generates more realistic 3D models. Final models are interactively explored with cellPACKgpu. Molecules with steric clashes are shown in red in both model images.