The recently released 2022 version of GROMACS comes with some exciting new features and improvements. Perhaps the most exciting is the introduction of the CP2K-GROMACS interface. CP2K is a quantum chemistry package that when combined with existing GROMACS simulation techniques, such as energy minimization and classical MD, will allow multi-scale Quantum Mechanics/Molecular Mechanics (QM/MM) simulations to be performed in fully periodic systems. If you would like more information on using the GROMACS+CP2K interface for QM/MM, there is an on-demand course available via BioExcel with complementary tutorials on the BioExcel GitHub along with a webinar by Dmitry Morozov.
Non-equilibrium pulling simulations are performed by applying forces or constraints between one or more pairs of groups of atoms. A new pull coordinate type, the transformation pull coordinate, has been added to permit mathematical transformation of a pull coordinate. This is simply a user-defined mathematical formula added to the .mdp file as a string. It allows the user to define complex reaction coordinates via non-linear transformations by, for example, combinations of multiple pull coordinates, and it can enhance sampling by combination with the Accelerated Weight Histogram Method.
Increased control over alchemical transformation pathways is now possible owing to the newly formulated soft-core non-bonded interactions in free-energy calculations. There are now two schemes available to choose from: the Beutler et al. [1] and the newly implemented Gapsys et al. [2] soft-core functions. The benefit of the new method is that it circumvents the occasional issue with the Beutler model that, for some combinations of Coulomb and Lennard-Jones parameters, unwanted nonbonded potential energy minima may arise.
Existing tools have also been improved: within gmx trjconv, the frame closest to the specified time with -dump is always written; and only selected atoms are written to TNG files when requested. The migration of gmx msd to the trajectoryanalysis framework not only creates a 20% decrease in execution time, but also comes with a new -maxtau option which limits the time delta value between frames when calculating MSDs. This allows users to constrain the sampling to only useful values, greatly benefiting those with large systems that would otherwise receive out-of-memory errors and slow execution times. Finally, gmx lie and gmx chi are now easier to use.
In addition to the new features, as with each new version of the software, there are several performance improvements. All of the improvements can be found in the new version release notes; however, a few notable examples include replica-exchange molecular dynamics is now able to run on GPU; free-energy kernels have been accelerated using SIMD and so free-energy calculations are now up to three times faster when running on GPU; the gmx grompp command now runs 20-50% faster owing to improvements in the loops in the parameter- and atom-lookup code; and parallelization in GPU-enabled runs using CUDA will, by default, use direct GPU communication.
Finally, a Spotify playlist, GROMACS: “Jiggling atoms with music” based on the GROMACS reminds you… quotes is available. Why not take a listen next time you are running your simulations? To continue finding out more about the new release, catch up on the latest BioExcel webinar where Paul Bauer talks about some of these new changes including the new CP2K interface, new pulling simulation abilities, and some performance improvements and hardware support. Our next webinar features Szilárd Páll as he talks about improvements in the GROMACS heterogeneous parallelization, so stay tuned for that.
1 T.C. Beutler, A.E. Mark, R.C. van Schaik, P.R. Greber, and W.F. van Gunsteren, “Avoiding singularities and numerical instabilities in free energy calculations based on molecular simulations,” Chem. Phys. Lett., 222 529–539 (1994).
2 V. Gapsys, D. Seeliger, and B.L. de Groot, “New Soft-Core Potential Function for Molecular Dynamics Based Alchemical Free Energy Calculations”, J. Chem. Theor. Comput., 8 2373-2382 (2012).
