CondenSimAdapter 1.0.4

An automated workflow for protein condensate simulations, covering the main stages from coarse-grained (CG) to all-atom (AA)

CondenSimAdapter

An automated workflow for protein condensate simulations, covering the main stages from coarse-grained (CG) to all-atom (AA).

Installation

Quick Start (Recommended)

Step 1: Create environment and install heavy dependencies

conda create -n condensim python=3.11 -y
conda activate condensim

# PyTorch + DGL (CUDA 12.x)
pip install "torch>=2.4,<2.5" --index-url https://download.pytorch.org/whl/cu124
pip install dgl -f https://data.dgl.ai/wheels/torch-2.4/cu124/repo.html

# OpenMM with CUDA 12
pip install "openmm[cuda12]>=8.2"

For other setups:

• CUDA 13.x: replace cu124 with cu13 and use openmm[cuda13]
• AMD GPU: pip install "openmm[hip6]>=8.2" or openmm[hip7]
• CPU only: use --index-url https://download.pytorch.org/whl/cpu and plain openmm>=8.2

Step 2: Install CondenSimAdapter

pip install CondenSimAdapter

Step 3: Download neural network models (~180 MB)

adapter models download

Models are hosted on GitHub Releases and cached locally.

Step 4: Verify

adapter --help
adapter info
adapter models status

Development Installation

git clone https://github.com/hanlab-computChem/CondenSimAdapter.git
cd CondenSimAdapter

# Install heavy deps (Step 1 above), then:
pip install -e ".[ml,openmm,dev]"

# Download models
adapter models download

# Run tests
pytest tests/unit -v

Usage

Typical Workflow

# 1. Create project configuration
adapter init my_project --topol cubic -c FUS:10

# 2. Run CG simulation
adapter cg -f config.yaml

# 3. Backmap to all-atom
adapter backmap -i output_CG -f config.yaml

# 4. Energy minimization
adapter minimize -i output_backmap -f config.yaml

# 5. Generate production run scripts
adapter to_run -f config.yaml

CLI Commands

CORE COMMANDS:
    cg               Run coarse-grained simulation
    backmap          Backmap CG structure to all-atom representation
    minimize         Energy minimization with AMBER/CHARMM force fields
    to_run           Generate production run scripts for minimize output

UTILITY COMMANDS:
    init             Initialize a new configuration template
    droplet-density  Estimate protein density in droplet geometry
    info             Display system and environment information
    forcefield       Manage custom all-atom force fields
    models           Manage neural network models for backmapping

Requirements

• Python >= 3.10, < 3.12 (3.11 recommended)
• Linux x86_64
• CUDA >= 12.x capable GPU (optional, CPU mode works too)
• GROMACS >= 2023 (for topology preparation)

Package Structure

Component	Size	Notes
Source code + data	~3 MB	Installed via pip
Neural network models	~235 MB	Downloaded via adapter models download
PyTorch + DGL + OpenMM	~2 GB	Installed via pip (Step 1)

Links

• PyPI: https://pypi.org/project/CondenSimAdapter/
• GitHub: https://github.com/hanlab-computChem/CondenSimAdapter
• Issues: https://github.com/hanlab-computChem/CondenSimAdapter/issues

License

GPL-3.0