AMBER 26 (pmemd 26 + AmberTools 26) - User Guide

Cluster: XLence (UNIMI Dipartimento di Scienze Farmacologiche e Biomolecolari) Date: September 30, 2026 Installation: /sw/amber/26/ (one build for login and compute nodes)

📦 Version Information

Amber (pmemd): 26, update 3 AmberTools: 26, update 3 PLUMED: loaded at run time from the plumed module (2.10.1) CUDA: 12.5 (embedded via RUNPATH, no module needed) GPU code: sm_75 (RTX 2080 Ti on the compute nodes, RTX 2060 SUPER on xlence2) MPI: OpenMPI 4.1.6 Python: 3.12, Amber's own Miniforge in $AMBERHOME/miniconda

One build for all nodes

Amber 26 has a single build (AVX2, tuned for the compute nodes' Xeon W-2155) that runs on the login nodes and on node1-5/ngs. There is no architecture detection any more.

🚀 Loading the Module

module load amber            # amber/26, the default

# With PLUMED
module load amber plumed     # plumed/2.10.1, the default

# Verify
echo $AMBERHOME              # /sw/amber/26

No miniforge3 module is needed. The Python tools (ParmEd, pytraj, MMPBSA.py, pdb4amber, ...) use Amber's own Python. For your own scripts that import them, use amber.python instead of python3.

Note: with a conda environment of your own active, amber.conda acts on that environment, not on Amber's. Deactivate it before using amber.conda.

💻 Main Executables

MD Engines

Executable	Description
`pmemd.cuda`	GPU-accelerated MD (single precision, single GPU)
`pmemd.cuda_DPFP`	GPU MD in double precision (for checks)
`pmemd.cuda.MPI`	Multi-GPU MD, replica exchange
`pmemd.MPI`	CPU parallel MD (MPI, at least 2 ranks)
`pmemd`	Serial CPU MD
`sander`	Classical MD engine (older, more features), QM/MM
`sander.quick.cuda`	QM/MM with the QUICK QM engine on the GPU

Preparation Tools

Tool	Description
`tleap`	Topology and coordinate file preparation
`antechamber`	Small molecule parameterization
`parmchk2`	Parameter file checker
`pdb4amber`	PDB file preparation for AMBER
`packmol-memgen`	Membrane system builder
`proprep`	Interactive protein preparation (user data in `~/.proprep`)

Analysis Tools

Tool	Description
`cpptraj`	Trajectory analysis (VERY powerful); `cpptraj.OMP` with OpenMP
`ambpdb`	PDB file manipulation

Python Tools

Tool	Description
`parmed`	Parameter/topology editor
`pytraj`	Python interface to cpptraj
`MMPBSA.py`	Free energy calculations (MM-PBSA/GBSA)

📖 Usage Examples

1. Basic GPU Simulation (Single Node)

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

module purge
module load amber

# Minimization
pmemd.cuda -O -i min.in -p system.prmtop -c system.inpcrd -o min.out -r min.rst -inf min.info

# Heating
pmemd.cuda -O -i heat.in -p system.prmtop -c min.rst -o heat.out -r heat.rst -x heat.nc

# Production
pmemd.cuda -O -i prod.in -p system.prmtop -c heat.rst -o prod.out -r prod.rst -x prod.nc

pmemd.cuda runs on the GPU: one CPU core is enough.

2. Multi-GPU Simulation

#!/bin/bash
#SBATCH --job-name=amber_multigpu
#SBATCH --nodes=1
#SBATCH --ntasks=2
#SBATCH --cpus-per-task=1
#SBATCH --gres=gpu:2
#SBATCH --time=48:00:00

module purge
module load amber

# One MPI rank per GPU
mpirun -np 2 pmemd.cuda.MPI -O -i prod.in -p system.prmtop -c heat.rst -o prod.out -r prod.rst -x prod.nc

Amber recommends one simulation per GPU (https://ambermd.org/GPUSupport.php): pmemd.cuda.MPI is meant mainly for replica exchange and multi-replica runs.

3. PLUMED-Enhanced MD

PLUMED is loaded at run time from the plumed module. It is enabled in the &cntrl namelist of the MD input, not on the command line:

#!/bin/bash
#SBATCH --job-name=amber_plumed
#SBATCH --nodes=1
#SBATCH --ntasks=1
#SBATCH --gres=gpu:1
#SBATCH --time=24:00:00

module purge
module load amber plumed

# Collective variables
cat > plumed.dat << 'EOF'
d1: DISTANCE ATOMS=1,10
PRINT ARG=d1 FILE=COLVAR STRIDE=100
EOF

# In prod.in, inside &cntrl:  plumed=1, plumedfile='plumed.dat',
pmemd.cuda -O -i prod.in -p system.prmtop -c input.rst -o prod.out -r prod.rst -x prod.nc

The same works with pmemd, pmemd.MPI and sander. At start-up the program prints +++ PLUMED_KERNEL="/sw/plumed/2.10.1/lib/libplumedKernel.so" +++; without module load plumed, a run with plumed=1 stops (see Troubleshooting).

Note: pmemd and sander have no -plumed command-line option; with it they stop at start-up.

4. Trajectory Analysis with cpptraj

module load amber

# Interactive mode
cpptraj

# Batch mode
cat > analysis.cpptraj << 'EOF'
parm system.prmtop
trajin prod.nc
autoimage
rms first @CA
atomicfluct out rmsf.dat @CA byres
radgyr out rgyr.dat @CA,C,N,O
run
EOF

cpptraj -i analysis.cpptraj

5. Python Analysis with pytraj

module load amber

amber.python << 'EOF'
import pytraj as pt

# Load trajectory
traj = pt.load('prod.nc', 'system.prmtop')

# Calculate RMSD
rmsd = pt.rmsd(traj, mask='@CA')
print(f"Average RMSD: {rmsd.mean():.2f} Å")

# Calculate radius of gyration
rg = pt.radgyr(traj, mask='@CA,C,N,O')
print(f"Average Rg: {rg.mean():.2f} Å")
EOF

6. Free Energy with MMPBSA.py

module load amber

# Create input file
cat > mmpbsa.in << 'EOF'
&general
  startframe=100, endframe=500, interval=5,
/
&gb
  igb=5, saltcon=0.150,
/
EOF

# Run MM-GBSA calculation
MMPBSA.py -O -i mmpbsa.in -o FINAL_RESULTS.dat -sp solvated.prmtop -cp complex.prmtop -rp receptor.prmtop -lp ligand.prmtop -y prod.nc

🔧 System Preparation Workflow

1. Prepare PDB File

module load amber

# Clean PDB file
pdb4amber -i input.pdb -o clean.pdb --dry

# Add hydrogens with reduce
pdb4amber -i input.pdb -o clean.pdb --reduce

2. Create Topology with tleap

tleap -f - << 'EOF'
source leaprc.protein.ff19SB
source leaprc.water.tip3p

# Load structure
mol = loadpdb clean.pdb

# Add counterions
addIons mol Na+ 0
addIons mol Cl- 0

# Solvate
solvatebox mol TIP3PBOX 12.0

# Save topology and coordinates
saveamberparm mol system.prmtop system.inpcrd
savepdb mol system.pdb

quit
EOF

3. Parameterize Small Molecules

# For organic molecules
antechamber -i ligand.pdb -fi pdb -o ligand.mol2 -fo mol2 -c bcc -s 2

# Generate frcmod file
parmchk2 -i ligand.mol2 -f mol2 -o ligand.frcmod

# Load in tleap
tleap -f - << 'EOF'
source leaprc.gaff2
LIG = loadmol2 ligand.mol2
loadamberparams ligand.frcmod
saveamberparm LIG ligand.prmtop ligand.inpcrd
quit
EOF

🧪 Example Input Files

Minimization (min.in)

Minimization
 &cntrl
  imin=1, maxcyc=5000, ncyc=2500,
  cut=10.0, ntb=1,
  ntpr=100, ntwx=0,
 /

Heating (heat.in)

Heating: 0K -> 300K
 &cntrl
  imin=0, ntx=1, irest=0,
  nstlim=25000, dt=0.002,
  ntf=2, ntc=2,
  tempi=0.0, temp0=300.0,
  ntpr=500, ntwx=500, ntwr=5000,
  cut=10.0, ntb=1,
  ntt=3, gamma_ln=2.0,
  nmropt=1,
 /
 &wt type='TEMP0', istep1=0, istep2=25000,
     value1=0.0, value2=300.0 /
 &wt type='END' /

Production (prod.in)

Production MD
 &cntrl
  imin=0, ntx=5, irest=1,
  nstlim=5000000, dt=0.002,
  ntf=2, ntc=2,
  temp0=300.0,
  ntpr=5000, ntwx=5000, ntwr=50000,
  cut=10.0, ntb=2, ntp=1, taup=2.0,
  ntt=3, gamma_ln=2.0,
 /

🐛 Troubleshooting

"CUDA is not supported on this GPU"

Cause: Trying to run on incompatible GPU or no GPU available Solution: Check nvidia-smi, request GPU with #SBATCH --gres=gpu:1

"Unit cell is not initialized"

Cause: Using ntb=1 or ntb=2 without periodic box Solution: Either solvate system or use ntb=0, cut=999.0 for vacuum

"+++ ERROR: You are trying to use an invalid plumed object. +++"

Cause: plumed=1 in &cntrl without the plumed module, so PLUMED_KERNEL is not set Solution: module load amber plumed

"MPI version of PMEMD must be used with 2 or more processors!"

Cause: pmemd.MPI started with one rank Solution: mpirun -np 2 or more (and #SBATCH --ntasks to match), or use pmemd

Python: "ImportError: ... libcrypto.so.3: version `OPENSSL_3.3.0' not found" when importing pytraj

Cause: The module's PYTHONPATH makes another Python (python3, the one of md/26, or your own conda environment) import Amber's modules, which are built for Amber's Python and its libraries Solution: Run your scripts with amber.python

"could not open file"

Cause: Missing input files or wrong paths Solution: Check all files exist: ls -lh *.prmtop *.inpcrd *.in

📚 Documentation and Support

Local Documentation: - Build, local patches and validation: /sw/amber/26/README-xlence.md - PLUMED on the cluster: /sw/plumed/2.10.1/README-xlence.md

Online Resources: - AMBER Homepage: https://ambermd.org - Manuals (Amber 26, AmberTools 26): https://ambermd.org/Manuals.php - Tutorials: https://ambermd.org/tutorials/ - Mailing List: http://lists.ambermd.org/

PLUMED Resources: - PLUMED Manual: https://www.plumed.org/doc - PLUMED Masterclass: https://www.plumed.org/masterclass

📝 Citation

For how to cite Amber 26 and AmberTools 26, see https://ambermd.org/CiteAmber.php

For PLUMED:

Tribello, G.A., Bonomi, M., Branduardi, D., Camilloni, C., Bussi, G. PLUMED 2: New feathers for an old bird Computer Physics Communications 185, 604 (2014)

🔄 Version History

Date	Version	Changes
2025-10-12	24 + AmberTools 25	Initial installation with PLUMED 2.9 integration (dynamic)
2026-09-30	26 + AmberTools 26	One AVX2 build for all nodes, Amber's own Python, PLUMED at run time (plumed/2.10.1). Amber 24 (`amber/24`) and Amber 25 (in `md/26`) removed

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