Tools for detecting SVs in Hi-C using machine learning.

Installation and usage

1. Create new pip virtual environment

python -m venv hict_venv

2. Activate new pip virtual environment Unix/MacOS:

source hict_venv/bin/activate

Windows:

hict_venv\Scripts\activate

3. Install wheel file

pip install hict_patterns-0.1-py3-none-any.whl

4. Check installation

hict_patterns -h

5. Unzip weight.zip to the working directory from which you are going to execute the rest of the commands Done!

Script parameters

Main script of project including all modules together can be run as console tool.

hict_patterns file_path [--search_in_1k] [-B BATCH_SIZE] [--device DEVICE] 

Required input

--file_path or first argument

Path to HiC file - .mcool format, should have 50Kb, 10Kb, 5Kb resolitions and 1kb resolution if --search_in_1k option used. If haven't file with this resolutions use cooler_zoomify.

Optional input

--search_in_1k

Whether to perform or not detection on 1Kb resolution. Default is not.

-B BATCH_SIZE

Size of data batch processed simultaneously by neural network, larger size reduces time of work but requires more RAM and VRAM. Default is 512.

Output

Output file is a table in .csv format. It consist from 3 columns. First two is whole genome range coordinates of structural variation. Third is identified сlass.

Example run and data

You could test work on sample file

hict_patterns dong_colluzzii.mcool  --search_in_1k --device auto

You should find 3 structural variations in the result.csv file