streamd

StreaMD: a tool to perform high-throughput automated molecular dynamics simulations

StreaMD provides an end-to-end molecular dynamics workflow that takes a PDB structure as input and automatically performs system preparation, equilibration, production and continuation runs, and analysis, producing XTC trajectories together with ready-to-use plots and CSV outputs.

Features:

• Run multiple simultaneous molecular dynamics simulations

• Run multiple replicas of the same system for multiple complexes in a single command

• Simulation for different systems:

  • Protein in Water;
  • Protein - Ligand;
  • Protein - Cofactor (multiple);
  • Protein - Ligand - Cofactor (multiple);

• Simulations of boron-containing molecules using Gaussian software

• Simulations of ligand-binding metalloproteins with MCPB.py

• Distributed computing using dask library

• Running parallel simulations on multiple servers

• Extending the time of MD simulations

• Continuing interrupted MD simulations

• Restarting interrupted MD preparation by invoking the same command

• Implemented tools for end-state free energy calculations (gmx_MMPBSA) and protein–ligand interaction analysis (ProLIF)

• Support for customized .mdp files

• Interactive trajectory convergence analysis for multiple complexes

• GPU support

Quick start

Download the Conda environment YAML file (CPU-only or GPU):

• CPU-only

wget -O env.yml https://raw.githubusercontent.com/ci-lab-cz/streamd/refs/heads/master/env.yml

• GPU

wget -O env_gpu.yml https://raw.githubusercontent.com/ci-lab-cz/streamd/refs/heads/master/env_gpu.yml

# Create environment (choose CPU-only or GPU)
conda env create --file env.yml -n md          # or env_gpu.yml on GPU-capable hosts
conda activate md

# Install
pip install streamd
# or latest main branch
pip install git+https://github.com/ci-lab-cz/streamd.git

Minimal protein-ligand run (1 ns)

run_md -p protein.pdb -l ligand.mol --md_time 1

Protein - multiple ligands multiple replicas runs (1 ns)

run_md -p protein.pdb -l ligands.sdf --md_time 1 --replicas 3 --seed 1024

Extend successfully finished simulations

run_md --wdir_to_continue md_files/md_run/protein_H_HIS_ligand_*/ --md_time 10

GPU-accelerated simulations

run_md -p protein_HIS.pdb -l ligand.mol --md_time 1 --device gpu --ncpu 32

Change simulation box

run_md -p protein.pdb -l ligand.mol --md_time 1 --box_type dodecahedron --box_padding_nm 1.2

Use a specific force field

run_md -p protein.pdb --md_time 1 --protein_forcefield amber99sb-ildn

Pass the --protein_forcefield value that matches the directory name of the desired pdb2gmx .ff bundle in your GROMACS installation.

MM-PBSA/MM-GBSA calculation support

run_gbsa --wdir_to_run md_files/md_run/protein_H_HIS_ligand_1 md_files/md_run/protein_H_HIS_ligand_2 -c 128 -m mmpbsa.in

This functionality is based on the gmx_MMPBSA tool

ProLIF (Protein-Ligand Interaction Fingerprints) Analysis

run_prolif --wdir_to_run md_files/md_run/protein_H_HIS_ligand_1 md_files/md_run/protein_H_HIS_ligand_2 -c 128 -s 5

This functionality is based on the ProLIF tool

More examples can be found in the documentation

Documentation

https://streamd.readthedocs.io/

Maintainers can find the automated PyPI release procedure in docs/releasing.md.

Ready-to-use containers (Apptainer)

Pre-built .sif images are available (CPU and GPU) in the Zenodo record

• CPU image

wget -O streamd_cpu.sif https://zenodo.org/records/18176058/files/streamd_cpu.sif

• GPU image

wget -O streamd_gpu.sif https://zenodo.org/records/18176058/files/streamd_gpu.sif

Run apptainer:
CPU: apptainer run --cleanenv streamd_cpu.sif run_md --help
GPU: apptainer run --nv --cleanenv streamd_gpu.sif run_md --help

The provided .sif images are intended for Apptainer on Linux/HPC systems. GPU usage requires an NVIDIA GPU node and launching with --nv and run_md --device gpu.

License

MIT

Citation

Ivanova A, Mokshyna O, Polishchuk P.
StreaMD: the toolkit for high-throughput molecular dynamics simulations.
J. Cheminf. 2024, 16 (1), 123.
https://doi.org/10.1186/s13321-024-00918-w