To the Editor: Molecular dynamics (MD) simulations are a wellestablished technique to investigate time-resolved motions of biological macromolecules at atomic resolution1. Methodological advances, continued software optimization and hardware acceleration have broadened the applicability of MD simulations with respect to feasible system size, runtime and quality2. Coupled to these advances, availability of vast cloud storage is enabling the creation of MD trajectory databases3. However, accessing, viewing and sharing MD trajectories is restricted by large file sizes and the need for specialized software (e.g., VMD, Chimera, Schrodinger Suite Products), which greatly limits the audience to which the MD data are available. In light of increasingly interdisciplinary research and remote collaborations, it is desirable to make the atom trajectories of MD simulations widely available to facilitate interactive exploration and collaborative visual analysis as well as to promote discussions. While some tools exist for analysis or visualization of trajectories, none of these offers a straightforward and easy solution for sharing and viewing MD trajectories online (for a comparison, see Supplementary Note 1). Here, we present MDsrv, a tool to stream MD trajectories and show them interactively within web browsers without requiring advanced knowledge in specialized MD software (Fig. 1). MDsrv is available as a software package (via PyPi and conda, Supplementary Software) that can be run locally or deployed to a dedicated web server to make data from MD simulations accessible to a wide audience of researchers, which helps to facilitate collaboration between computational and experimental researchers (see Supplementary Note 1). For interactive and remote exploration of trajectories, we used client-server architecture to create a web-based platform. MDsrv can be used for viewing or serving MD simulations. The latter usage offers two modes of operation—one is an easy-to-use command-line tool for local service (Supplementary Fig. 1a); the other deploys the tool on a server to provide a dedicated streaming service (see Supplementary Notes 2–4). MDsrv supports structures, topologies and trajectories from common MD packages including Amber, Gromacs, NAMD, Tinker or Desmond. To aid visualization and analysis of MD trajectories, a number of processing steps can be performed on the structure and trajectory data (Supplementary Table 1). When analyzing MD simulations, the focus is generally on the internal motions of the macromolecule rather than on the diffusion movement in aqueous solution or lipid bilayers. To display unprocessed data from nascent simulations, on-the-fly superposition to a reference structure, handling of simulations with periodic boundary conditions and secondary structure assignment can be performed. By applying these calculations only to displayed frames, no scalability issues arise when viewing large structures and very long trajectories. Finally, to reduce the amount of transferred data, coordinate frames can be retrieved individually, and atoms from solvent or other, nonrelevant molecules can be filtered out before transferring a frame. Scalable molecular graphics for the MDsrv web application are provided by the NGL Viewer4. Accelerated 3D graphics are enabled by WebGL, a standard built directly into web browsers without requiring installation of plugins. NGL supports a wide Autodesk for supplying cardboard VR devices at international conferences. We thank the Canadian Institute for Advanced Research and the Canada Excellence Research Chairs program (both to O.P.E.) for financial support to attend and present at international conferences. O.P.E. holds the Anne and Max Tanenbaum Chair in Neuroscience.
[1]
Ferran Sanz,et al.
Membrane cholesterol access into a G-protein-coupled receptor
,
2017,
Nature Communications.
[2]
Alexander S. Rose,et al.
NGL Viewer: a web application for molecular visualization
,
2015,
Nucleic Acids Res..
[3]
Allison Doerr,et al.
A database of dynamics
,
2010,
Nature Methods.
[4]
Michael Levitt,et al.
Birth and future of multiscale modeling for macromolecular systems (Nobel Lecture).
,
2014,
Angewandte Chemie.
[5]
J. P. Grossman,et al.
Biomolecular simulation: a computational microscope for molecular biology.
,
2012,
Annual review of biophysics.