GROMACS 2025.0 with PLUMED 2.10.1 - User Guide

Cluster: XLence (UNIMI Dipartimento di Scienze Farmacologiche e Biomolecolari) Date: September 30, 2026 Installation: /sw/gromacs/2025.0-avx2/ (one build for login and compute nodes)

📦 Version Information

GROMACS: 2025.0, patched with PLUMED 2.10.1 PLUMED: 2.10.1, loaded at run time from the plumed module CUDA: 12.5 (embedded via RUNPATH, no module needed), code for the RTX 2080 Ti (sm_75) MPI: OpenMPI 4.1.6 (the system mpirun) SIMD: AVX2 (256-bit), on all nodes FFTW: 3.3.10, built by GROMACS with AVX2

One build for all nodes

GROMACS 2025.0 has a single build that runs on the login nodes and on node1-5/ngs. There is no architecture detection any more.

🚀 Loading the Module

module load gromacs          # gromacs/2025.0; also loads plumed (plumed/2.10.1)

# Verify
gmx_mpi --version | head -3
echo $GROMACS_ROOT

The module can be loaded together with the Python analysis environment: module load md/26 gromacs/2025.0.

Note: a .tpr file written by GROMACS 2025.0 cannot be read by older GROMACS versions. Prepare and run it with the same version.

💻 Main Executable

gmx_mpi - GROMACS with MPI support (use this for all simulations)

All GROMACS tools are accessed via subcommands:

gmx_mpi mdrun      # Run MD simulation
gmx_mpi grompp     # Prepare MD run
gmx_mpi pdb2gmx    # Convert PDB to GROMACS topology
gmx_mpi editconf   # Edit configuration
gmx_mpi solvate    # Solvate system
gmx_mpi genion     # Add ions

📖 Usage Examples

1. System Preparation

module load gromacs

# Convert PDB to GROMACS format
gmx_mpi pdb2gmx -f protein.pdb -o conf.gro -p topol.top -ff amber99sb-ildn -water tip3p

# Define box
gmx_mpi editconf -f conf.gro -o box.gro -c -d 1.0 -bt cubic

# Solvate
gmx_mpi solvate -cp box.gro -cs spc216.gro -o solvated.gro -p topol.top

# Add ions (neutralize)
gmx_mpi grompp -f ions.mdp -c solvated.gro -p topol.top -o ions.tpr
gmx_mpi genion -s ions.tpr -o ionized.gro -p topol.top -pname NA -nname CL -neutral

2. Basic GPU Simulation

#!/bin/bash
#SBATCH --job-name=gromacs
#SBATCH --nodes=1
#SBATCH --ntasks=1
#SBATCH --cpus-per-task=4
#SBATCH --gres=gpu:1
#SBATCH --time=24:00:00

module purge
module load gromacs

# Non-bonded and PME on the GPU, OpenMP threads as many as the allocated cores
# (mdrun also moves the update to the GPU when the system allows it)
GPU="-nb gpu -pme gpu -ntomp $SLURM_CPUS_PER_TASK"

# Energy minimization
gmx_mpi grompp -f em.mdp -c ionized.gro -p topol.top -o em.tpr
gmx_mpi mdrun -v -deffnm em -ntomp $SLURM_CPUS_PER_TASK

# NVT equilibration
gmx_mpi grompp -f nvt.mdp -c em.gro -r em.gro -p topol.top -o nvt.tpr
gmx_mpi mdrun -v -deffnm nvt $GPU

# NPT equilibration
gmx_mpi grompp -f npt.mdp -c nvt.gro -r nvt.gro -t nvt.cpt -p topol.top -o npt.tpr
gmx_mpi mdrun -v -deffnm npt $GPU

# Production
gmx_mpi grompp -f md.mdp -c npt.gro -t npt.cpt -p topol.top -o md.tpr
gmx_mpi mdrun -v -deffnm md $GPU

3. PLUMED-Enhanced MD

GROMACS is patched with PLUMED 2.10.1 and loads the PLUMED kernel at run time from PLUMED_KERNEL, which the plumed module sets (module load gromacs loads it):

#!/bin/bash
#SBATCH --gres=gpu:1

module purge
module load gromacs

# Create plumed.dat
cat > plumed.dat << 'EOF'
# Distance collective variable
d1: DISTANCE ATOMS=1,10
PRINT ARG=d1 FILE=COLVAR STRIDE=100
EOF

# Run with PLUMED
gmx_mpi mdrun -v -deffnm md -plumed plumed.dat

4. Multiple Simulations and Replica Exchange

#SBATCH --ntasks=2
#SBATCH --gres=gpu:1

module load gromacs

# One simulation per directory (each with its topol.tpr, and plumed.dat if used)
mpirun -np 2 gmx_mpi mdrun -multidir rep0 rep1 -plumed plumed.dat
# Replica exchange every 500 steps
mpirun -np 2 gmx_mpi mdrun -multidir rep0 rep1 -replex 500 -plumed plumed.dat

With -multidir, PLUMED writes one output per replica (COLVAR.0, COLVAR.1, ...).

Important: all simulations must have the same nsteps. With different numbers of steps, GROMACS 2025.0 with PLUMED hangs at the end of the shortest one. With -replex, PLUMED can print collective variables but not apply a bias (not yet supported by the PLUMED patch for 2025.0).

🔧 Example MDP Files

Energy Minimization (em.mdp)

integrator  = steep
emtol       = 1000.0
emstep      = 0.01
nsteps      = 50000

cutoff-scheme = Verlet
nstlist     = 10
coulombtype = PME
rcoulomb    = 1.0
rvdw        = 1.0
pbc         = xyz

NVT Equilibration (nvt.mdp)

integrator  = md
dt          = 0.002
nsteps      = 50000     ; 100 ps

nstxout     = 5000
nstvout     = 5000
nstenergy   = 500
nstlog      = 500

continuation = no
constraint_algorithm = lincs
constraints = h-bonds

cutoff-scheme = Verlet
nstlist     = 10

coulombtype = PME
rcoulomb    = 1.0
rvdw        = 1.0

tcoupl      = V-rescale
tc-grps     = Protein Non-Protein
tau_t       = 0.1  0.1
ref_t       = 300  300

pcoupl      = no
pbc         = xyz

Production (md.mdp)

integrator  = md
dt          = 0.002
nsteps      = 5000000   ; 10 ns

nstxout     = 0
nstvout     = 0
nstfout     = 0
nstxout-compressed = 5000
compressed-x-grps  = System
nstenergy   = 5000
nstlog      = 5000

continuation = yes
constraint_algorithm = lincs
constraints = h-bonds

cutoff-scheme = Verlet
nstlist     = 10

coulombtype = PME
rcoulomb    = 1.0
rvdw        = 1.0

tcoupl      = V-rescale
tc-grps     = Protein Non-Protein
tau_t       = 0.1  0.1
ref_t       = 300  300

pcoupl      = Parrinello-Rahman
pcoupltype  = isotropic
tau_p       = 2.0
ref_p       = 1.0
compressibility = 4.5e-5

pbc         = xyz
DispCorr    = EnerPres

The neighbor list (rlist, nstlist) is tuned by mdrun itself with the Verlet scheme; ns_type no longer exists.

📊 Analysis Examples

module load gromacs

# RMSD
echo "4 4" | gmx_mpi rms -s md.tpr -f md.xtc -o rmsd.xvg -tu ns

# RMSF
echo "4" | gmx_mpi rmsf -s md.tpr -f md.xtc -o rmsf.xvg -res

# Radius of gyration
echo "4" | gmx_mpi gyrate -s md.tpr -f md.xtc -o gyrate.xvg

# Hydrogen bonds
echo "1 1" | gmx_mpi hbond -s md.tpr -f md.xtc -num hbnum.xvg

For Python analysis (MDAnalysis, MDTraj, ...): module load md/26, which can be loaded together with gromacs/2025.0.

🐛 Troubleshooting

"Could not find GPU"

Ensure #SBATCH --gres=gpu:1 is set
Check with nvidia-smi

"You are trying to use plumed, but it is not available. Check your PLUMED_KERNEL environment variable."

No plumed module is loaded: module load gromacs loads it, unless it was unloaded afterwards

"reading tpx file (...) version 137 with version 133 program"

The .tpr was written by GROMACS 2025.0 and read by an older GROMACS: use gmx_mpi from gromacs/2025.0

mdrun hangs at the end of a `-multidir` run

The simulations have different nsteps (see "Multiple Simulations" above): give them the same nsteps

"Fatal error: number of coordinates in coordinate file does not match topology"

Rebuild TPR: gmx_mpi grompp -f file.mdp -c file.gro -p topol.top -o file.tpr

📚 Documentation

Local: - Build, validation and choices: /sw/gromacs/2025.0-avx2/README-xlence.md - PLUMED on the cluster: /sw/plumed/2.10.1/README-xlence.md

Online: - GROMACS Manual: https://manual.gromacs.org/2025.0/ - Tutorials: https://tutorials.gromacs.org/ - PLUMED Manual: https://www.plumed.org/doc-v2.10/user-doc/html/

📝 Citation

Abraham, M.J., et al. GROMACS: High performance molecular simulations through multi-level parallelism from laptops to supercomputers. SoftwareX 1-2, 19-25 (2015)

For PLUMED:

Tribello, G.A., et al. PLUMED 2: New feathers for an old bird. Computer Physics Communications 185, 604 (2014)

🔄 Version History

Date	Version	Changes
2025-10-09	2024.3 + PLUMED 2.9	AVX2 and AVX-512 builds
2026-09-30	2025.0 + PLUMED 2.10.1	One AVX2 build for all nodes, independent of md/26; 2024.3 and md/26's own GROMACS removed

For questions contact: Cluster administrators Installation: September 30, 2026 Installed by: Claude Code + Uliano Guerrini