GROMACS

What it simulates

GROMACS runs classical molecular dynamics of proteins, lipids, nucleic acids and solvent. It integrates Newton's equations on a force field with bonded terms, Lennard-Jones and Coulomb interactions. Long-range electrostatics use particle-mesh Ewald (PME).

The input to gmx mdrun is a single binary run input file, .tpr. gmx grompp builds it from three text inputs: an .mdp file with the settings, a coordinate file (.gro) and a topology (.top). Output is a log (md.log), an energy file (md.edr), a compressed trajectory (md.xtc), a full-precision trajectory (md.trr) and a checkpoint (md.cpt).

The key assumption is a fixed-charge, pairwise force field. Bonds do not break, and protonation states are set in the topology.

When to use it

• Solvated proteins, membranes and nucleic acids, with the AMBER, CHARMM, OPLS or GROMOS force fields.
• Free-energy and long-timescale work on CPU or GPU, where speed matters.

Alternatives on biosimhub: OpenMM takes a PDB and a Python script, so it suits a quick solvated run without writing .mdp and .top files. LAMMPS fits non-biomolecular systems and custom potentials.

The biosimhub example runs a prebuilt .tpr. Build one on your own machine with gmx grompp.

Install it yourself

GROMACS 2026.3 from conda-forge. A dry run resolved gromacs 2026.3 nompi_h2d14928_100:

conda install -c conda-forge gromacs=2026.3

The program is gmx. A minimal workflow from coordinates and a topology:

gmx grompp -f md.mdp -c conf.gro -p topol.top -o md.tpr
gmx mdrun -deffnm md

-deffnm md names every output md.*. -nsteps N overrides the step count stored in the .tpr. The recipe also passes -ntmpi 1 -ntomp 2 -nb cpu -pin off -noconfout. In the GROMACS docs, -ntmpi sets thread-MPI ranks, -ntomp sets OpenMP threads per rank, -nb cpu runs short-range non-bonded work on the CPU, and -pin off stops mdrun from binding threads to cores. In the example run, -noconfout left out the final confout.gro.

Reading the output

The example water.tpr holds 1530 atoms: 510 water molecules (SOL, OPLS atom types) in a 2.5 nm cubic box. gmx dump -s water.tpr shows leap-frog md, dt 0.002 ps, 5000 steps, PME, 0.9 nm cutoffs, a V-rescale thermostat at 300 K (tau-t 0.1 ps), and no pressure coupling. So 5000 steps is 10 ps.

• md.log (28078 bytes) prints an energy block every 500 steps (nstlog).
• md.edr (10512 bytes) holds energies every 100 steps. gmx energy -f md.edr reads it.
• md.xtc (59084 bytes) holds 11 frames, one every 500 steps.
• mdrun_stdout.txt is empty (0 bytes). mdrun writes its progress to stderr, and the log has the content.

gmx energy on this run reports:

• Step 0: potential energy +4571 kJ/mol, temperature 1374 K, pressure 1.1e5 bar. The starting structure is not relaxed.
• 2 ps: -20856 kJ/mol and 322.5 K. 10 ps: -20868 kJ/mol and 299.6 K.
• The whole-run average temperature is 338 K, which includes the hot start. Average only after equilibration.

Your own runs

dr() { docker run --rm --user "$(id -u):$(id -g)" -e HOME=/tmp -v "$PWD/in:/work/in:ro" -v "$PWD/out:/work/out" -w /work/out "$@"; }

1. A longer run: dr biosimhub/gromacs:2026.3 gmx mdrun -s /work/in/tpr.tpr -deffnm md -nsteps 25000 -ntmpi 1 -ntomp 2 -nb cpu -pin off -noconfout. That is 50 ps. gmx energy -b 10 (skip the first 10 ps) gave potential energy -20993.2 kJ/mol (error estimate 25, RMSD 136.8) and temperature 299.72 K. md.xtc grew to 272980 bytes and md.edr to 48912.
2. Your own tiny system. gmx solvate -cs spc216.gro -box 2.5 2.5 2.5 -p topol.top -o box.gro gave 510 water molecules, 1530 atoms. The topology used oplsaa.ff/forcefield.itp and oplsaa.ff/tip3p.itp. Steepest-descent minimisation (integrator = steep, emtol = 1000) converged in 7 steps. A 5000-step MD with gen-vel = yes, gen-temp = 300, gen-seed = 7 and constraints = h-bonds averaged 301.5 K over the last 5 ps (error estimate 2.4 K). Minimising first removes the 1374 K start of the example.
3. Run the resulting md.tpr with the commands above, or upload it to the biosimhub page.

Pitfalls

• Averages that include equilibration are wrong. The example's 338 K average sits 38 K above its 300 K target. Use gmx energy -b <ps>.
• -nsteps only changes the length. Time step, thermostat and cutoffs live in the .tpr, so change them in the .mdp and run grompp again.
• Minimise before MD. An unrelaxed box starts at +4571 kJ/mol and 1374 K.
• With -noconfout there is no final structure file. Use md.cpt to continue, or drop the flag.
• On a shared machine -pin off and a fixed -ntmpi/-ntomp stop mdrun from grabbing cores that belong to other jobs.

Further reading

• mdrun performance options: https://manual.gromacs.org/current/user-guide/mdrun-performance.html
• gmx mdrun: https://manual.gromacs.org/current/onlinehelp/gmx-mdrun.html
• mdp options: https://manual.gromacs.org/current/user-guide/mdp-options.html
• Install guide: https://manual.gromacs.org/current/install-guide/index.html
• Source: https://github.com/gromacs/gromacs
• Van Der Spoel D et al. (2005) J Comput Chem 26:1701-1718. https://doi.org/10.1002/jcc.20291