Installation¶
RepliSage needs Python 3.10 or newer and is tested on Linux. The 3D stage uses OpenMM, which is much faster on a GPU, while the Monte Carlo runs on the CPU.
From PyPI¶
pip install pyRepliSage
replisage -h
This installs the replisage command. The package is imported as RepliSage.
If replisage -h shows old options (for example --region_start), an older installed copy is shadowing the repository. Reinstall with pip install -e . in the repository root, or run python -m src.run from there.
From source¶
git clone https://github.com/SFGLab/RepliSage
cd RepliSage
pip install -r requirements.txt # dependencies
pip install -e . # the replisage command, editable
Without installing, you can also run it from the repository root: python -m src.run -h.
A clean environment (recommended)¶
conda create -n replisage python=3.11
conda activate replisage
pip install pyRepliSage
Optional parts¶
| You want | Install |
|---|---|
The pyvista movies (movie_3d_sister.mp4, movie_3d_compartment.mp4) |
pip install "pyRepliSage[viz]" (the ffmpeg encoder comes with imageio-ffmpeg) |
| 3D on a GPU | a CUDA or OpenCL OpenMM build, then --platform CUDA (or OpenCL) |
| To run the tests | pip install "pyRepliSage[test]", then pytest (fast tests) or python run_tests.py (all checks with one summary) |
| To build these docs | pip install "pyRepliSage[docs]", then mkdocs serve |
The interactive viewer_3d.html needs nothing extra: it is one self-contained file that opens in any browser.
Check that it works¶
replisage -h # lists every flag
replisage --bedpe-file loops.bedpe --region chr14:80835000-98674700 --n-steps 5000
The second command is a short loop-extrusion run. It writes a folder with report.html that you can open in a browser.
Docker¶
A Dockerfile is in the repository: docker build -t replisage .
Troubleshooting¶
| Symptom | Cause and fix |
|---|---|
| The first Monte Carlo run pauses for a few seconds | numba compiles the energy functions once; later runs start at once |
No template found for residue in the 3D stage |
the force field file does not match the structure; keep the default FORCEFIELD_PATH |
pyBigWig fails to install |
install it from conda-forge: conda install -c conda-forge pybigwig |
| The movie is skipped | install pyvista; use --skip-movies to skip the movies |
| GPU not used | check --platform, and that your OpenMM sees the device: python -m openmm.testInstallation |
Next: the command line and examples.