QwikMD — Integrative Molecular Dynamics Toolkit for Novices and Experts

João V. Ribeiro, Rafael C. Bernardi, Till Rudack, John E. Stone, James C. Phillips, Peter L. Freddolino, Klaus Schulten
2016 Scientific Reports  
The proper functioning of biomolecules in living cells requires them to assume particular structures and to undergo conformational changes. Both biomolecular structure and motion can be studied using a wide variety of techniques, but none offers the level of detail as do molecular dynamics (MD) simulations. Integrating two widely used modeling programs, namely NAMD and VMD, we have created a robust, user-friendly software, QwikMD, which enables novices and experts alike to address biomedically
more » ... elevant questions, where often only molecular dynamics simulations can provide answers. Performing both simple and advanced MD simulations interactively, QwikMD automates as many steps as necessary for preparing, carrying out, and analyzing simulations while checking for common errors and enabling reproducibility. QwikMD meets also the needs of experts in the field, increasing the efficiency and quality of their work by carrying out tedious or repetitive tasks while enabling easy control of every step. Whether carrying out simulations within the live view mode on a small laptop or performing complex and large simulations on supercomputers or Cloud computers, QwikMD uses the same steps and user interface. QwikMD is freely available by download on group and personal computers. It is also available on the cloud at Amazon Web Services. Nearly 40 years ago, in what was considered the first step of molecular dynamics (MD) simulations applied to biological systems, the dynamics of a folded globular protein (bovine pancreatic trypsin inhibitor) was studied by solving the equations of motion for the atoms with an empirical potential energy function 1 . The results suggested that the protein interior is fluid-like, a result of the diffusional character of local atom motions 2 . Since then, computer simulations of biomolecular systems have grown rapidly, passing from simulating small proteins in vacuum to simulating huge protein complexes in solvent/lipid environments 3 . Building on structural data from diverse experimental sources, today's MD simulations permit the exploration of biological processes in unparalleled detail 4 . The exponential growth of MD-based investigations is clear when the number of publications indexed at Thomson Reuters' Web of Science that contain the topic "molecular dynamics" is checked. Figure 1A shows that about 35,000 studies employing MD were published in 2015. But what exactly is MD and how can it be used? Proteins are not rigid bodies as even the first simulations, and a few experiments before them, were able to show 2 . Dynamics is clearly important for protein function, and discovering the mechanisms underlying function depends on a complete understanding of biological molecules, including their dynamics. Even long before the first protein MD was carried out, Richard Feynman remarked 5 : "Certainly no subject or field is making more progress on so many fronts at the present moment than biology, and if we were to name the most powerful assumption of all, which leads one on and on in an attempt to understand life, it is that all things are made of atoms, and that everything that living things do can be understood in terms of the jigglings and wigglings of atoms". Employing computers and based on a variety of experimental information, MD assists investigation of atomic motion as does no other methodology. The trajectories of molecules are determined mostly by numerically solving Newton's equations of motion for a system of interacting particles, the molecular atoms. The forces between the atoms and their potential energies are calculated from interatomic potentials in molecular mechanics force fields that are, in effect, huge data bases of molecular properties. Developed as a simple method in the late 1950's, MD algorithms evolved greatly, especially for the study of biological systems. All-atom MD simulations, employing classical mechanics, allowed the study of a broad range of biological systems, from small molecules such as anesthetics 6 or small peptides 7 , to very large protein complexes such as the ribosome 8 , chemoreceptor arrays 9 , or virus capsids 10,11 . Hybrid classical/quantum MD simulations allowed the study of enzymatic activity 12 , catalysis 13 , and biological membranes 14,15 . All of this was made possible by the development of a multitude of algorithms and MD computer programs closely coupled to ongoing advances in parallel computing and computing hardware. The role of these programs is clearly increasing every year as they become more complex, allowing them to be applied to a plethora of different scientific questions. This importance is reflected by the fact that all these programs are gaining users and, therefore, more citations every year (see Fig. 1B ). Most MD protocols are well established and can be applied to a wide range of investigations, while new protocols continue to be developed, pushing the limitations of MD simulations. Conformational sampling, very long simulations (in the millisecond regime), very large systems (with protein complexes in the megadalton scale), and resolving large structures in real-space are some of these limitations 4,16,17 . However, even though most protocols are well established, the necessity of understanding the underlying details of these protocols led MD to be viewed for a long time as a demanding technique only accessible to a user with deep knowledge in computational methodologies. Even though the advances in software and hardware in the last decade allowed MD to become accessible to a broader group of scientists, for most of them the necessity of computational methodology knowledge is still an obstacle. Unfortunately, the learning curve to acquire such knowledge is not smooth, requiring years to learn the details of some MD protocols. A major concern is that, while some novice MD users might feel discouraged by the difficulties, others might end up committing mistakes that are hard to be detected. As with many other techniques, MD is very sensitive to the parameters set by the user, and a mistake in even a simple step might lead to a questionable result. Simple mistakes made while preparing a simulation can easily result in errors in both carrying out and analyzing simulations. However, it is possible to identify a "safe" set of parameters for the vast majority of MD applications. Thus, MD, like any other experimental technique, is turning from a research to a routine laboratory technique, one that could be offered as a laboratory kit, but such a kit has not been made available yet. Actually, such kit has been already partially presented by a few commercial programs, e.g., MOE and Discovery Studio -BIOVIA. The idea behind these programs, which include modeling, docking, MD and many other computational biology techniques, is to assist users in employing computational biology through automating tedious steps. However, besides being not freely available for research, these programs may lack some of the MD capabilities that some experts need for their work, especially because MD is not the main focus of these programs. Other interfaces, such as CHARMM-GUI 18 , allow a user to prepare simulations, but do not assist in execution and analysis. Taking advantage of the fact that our group develops two widely employed computational tools for structural biology, namely the simulation program NAMD 19 and the set-up and analysis program VMD 20 , we developed
doi:10.1038/srep26536 pmid:27216779 pmcid:PMC4877583 fatcat:464rg7kcyzbgpcctssflgjerza