Installation Guide

This page provides detailed instructions for installing DROPPS (Distributed Rapid Operation Platform for Phase-separation Simulations). DROPPS is implemented in Python and uses OpenMM, numpy, gsd, and MDAnalysis as its computational backend.

Installation consists of the following steps:

  1. Preparing a Python environment

  2. Installing required dependencies

  3. Installing DROPPS from source

  4. Verifying the installation

DROPPS requires Python 3.10+, CUDA (optional), and a working compiler toolchain if OpenMM/GSD are compiled from source.

Quick start

DROPPS is distributed using python wheals package. It is recommended to install DROPPS using a conda virtual environment by the following very simple commands.

On Linux (preferably CentOS Stream 7/8/9) or MacOS (preferably M1/M2/M3/M4 Mac) without GPU support or with OpenCL Support

conda create -n DROPPS python=3.12.11
conda activate DROPPS
pip install dropps-0.2.0-py3-none-any.whl

On high-performance Computing Cluster with GPU Support

conda create -n DROPPS python=3.12
conda install -c conda-forge openmm cudatoolkit
pip install dropps-0.2.0-py3-none-any.whl

Manual installation

Prerequisites

Before installing DROPPS, ensure you have the following dependencies:

  • Python 3.10 or newer

  • pip ≥ 22.0

  • openMM ≥ 8.1

  • OpenCL toolkit (optional, for GPU acceleration on AMD cards or Apple silicons)

  • NVIDIA CUDA toolkit (optional, for GPU acceleration)

On Linux (Debian/Ubuntu):

sudo apt update
sudo apt install -y build-essential cmake git python3-dev python3-venv

Creating a Python Environment

It is recommended to install DROPPS inside a dedicated conda environment:

conda create -n dropps python=3.12 -y
conda activate dropps

Or using venv:

python3 -m venv dropps-env
source dropps-env/bin/activate

Installing Dependencies

DROPPS uses the following libraries:

  • OpenMM (MD engine)

  • GSD (If gsd file support is enabled)

  • MDAnalysis (trajectory manipulation)

  • numpy, scipy, h5py, tqdm

Install OpenMM:

conda install -c conda-forge openmm

Install GSD:

conda install -c conda-forge gsd

Install MDAnalysis:

pip install MDAnalysis

Install Python utilities:

pip install numpy scipy h5py tqdm

Installing DROPPS

DROPPS is currently under development and distributed via GitHub.

Clone the repository:

git clone https://github.com/tymworld/DROPPS.git
cd DROPPS

Install using pip:

pip install dist/dropps-0.2.0-py3-none-any.whl

Or install in editable (developer) mode:

pip install -e dist/dropps-0.2.0-py3-none-any.whl

Configuring GPU Support

DROPPS can run on:

  • NVIDIA GPUs (CUDA)

  • AMD GPUs / Apple silicons (OpenCL)

  • CPU only (OpenMM Reference platform)

To check that OpenMM sees your GPU:

python -m openmm.testInstallation

To verify HOOMD GPU backend:

python - << EOF
import hoomd
print(hoomd.device.GPU())
EOF

Environment Variables

Optional environment variables for performance tuning:

export OPENMM_CPU_THREADS=8
export OMP_NUM_THREADS=8
export HOOMD_DISABLE_JIT=1

Verifying the Installation

Run the built-in test script:

dps -h

Run a simple LLPS simulation example:

dps pdb2dps -s FF -r 1 -oc FF -op OC -on OC
dps genmesh -f FF.pdb -p FF.itp -n 1 -mesh 1 1 1 -oc box -op box
cp /PATH_TO_DPS/share/template/md.mdp ./
dps grompp -f box.pdb -p box.top -m md.mdp -o FF.tpr
dps mdrun -s FF.tpr -o FF

If the simulation starts successfully, the installation is complete.

Troubleshooting

OpenMM CUDA not detected

  • Check CUDA installation

  • Ensure correct conda channel priority

  • Ensure your GPU is supported by OpenMM

HOOMD “No GPU detected”

  • Update NVIDIA drivers

  • Try installing HOOMD from conda-forge only

MDAnalysis trajectory loading errors

  • Ensure packages are up-to-date:

    pip install --upgrade MDAnalysis h5py

If problems persist, please open an issue on the DROPPS GitHub page.

Note

DROPPS is actively developed. Installation instructions may change frequently as new modules and plugins are added.