GILoop
GILoop is a deep learning model for detecting CTCF-mediated loops on Hi-C contact maps.
New: GILoop now supports .cool input
Installation:
conda create -n GIL python=3.8
conda activate GIL
pip install -r requirements.txt
After running the code segment above, please download models and data from GILoop_assets and replace the ones in the local directory.
Usage:
You can either use .cool files as input, or use the output format of Juicer dump as input.
Cooler input:
python demo.py
API usage is self-documented in demo.py. This script is the full pipeline including data pre-processing (sequencing depth alignment, normalization, and expected vector calculation, etc.), and GILoop sampling and training. The preprocessing overhead could be heavy and could lead to a large running time.
Juicer dump text file as input:
python demo_from_processed.py
In this case, please make sure the text files are KR-normalized O/E matrices, and the source and the target are of similar number of reads. The script only carries out GILoop patching and training. No preprocessing is applied to the data.