SFGLab / gStripe

Graph-based architectural stripe caller.

gStripe

Graph-based stripe caller that uses chromatin loop data.

Requirements:

• Python 3.10
• Operating system: Linux (Ubuntu 18.04 or higher, recommended) or Windows 10 (or higher).

Installation

1. Optional, but recommended: Create and activate a python envrionment using pyenv, Conda or Mamba.
2. Install gStripe, depending on the chosen source:
   • Installing directly from github: run pip install git+https://github.com/SFGLab/gStripe.git.
   • Installing from a zip file: run pip install gStripe-v1.0.zip in the directory containing the downloaded file. The name of the .zip archive might differ, depending on the download source and version number.

To verify, that the installation proceeded correctly, you can do the following:

1. Run python -m gstripe.gstripe --help: you should see the help message. If not, check if the correct environment has been activated.
2. Run gStripe on an example file provided in examples/basic_test.bedpe in the github repository or in the .zip archive, using the following command: python -m gstripe.gstripe basic_test.bedpe .. The console output should end with [INFO] main(0.06s): All done. (timing may vary) and two new files should be created: ./basic_test.bedpe.gstripes_raw.tsv (results) and basic_test.bedpe.gstripe.log (log).

Usage

Run python -m gstripe.gstripe input_loops_file.bedpe output_directory to call the stripes using loops from input_loops_file.bedpe and place the results (input_loops_file.bedpe.gstripes_raw.tsv by default) and the log file in output_directory. This results in saving the candidate stripes to the output directory in a .tsv file. They should then be filtered by the user.

It is recommended to use the default values of all parameters specified in usage options (python -m gstripe.gstripe --help).

Use --fix_bin_start in case of binned data, where adjacent bins would overlap (i.e. when one anchor end is "15000" and the start of an adjacent anchor is also "15000"). Recommended for HiChIP.

In case of problems with multiprocessing, use --max_workers=1