DeepHicIntegrator

This tool permits the integration of a Hi-C matrix with one or several histone marks by interpolating in the latent space of an Autoencoder.

Installation

Clone the repository

git clone https://github.com/kabhel/DeepHicIntegrator.git
cd DeepHicIntegrator

Requirements

1. A linux distribution.
2. Python3 and the following python packages : tensorflow, keras, docopt, schema, pandas, numpy, scipy, matplotlib, sklearn, cooler, hic2cool and m2r (for Sphinx).

pip3 install -r requirements.txt

1. A Hi-C matrix in .hic file format.

Please, download the GSE63525 HUVEC genome in order to run the toy example.

wget -i ftp://ftp.ncbi.nlm.nih.gov/geo/series/GSE63nnn/GSE63525/suppl/GSE63525_HUVEC_combined_30.hic.gz
gunzip GSE63525_HUVEC_combined_30.hic.gz

1. One or several histone marks in 2D dimension.

Run the program

Toy example

./deep_hic_integrator data/hic_matrix/GSE63525_HUVEC_combined_30.hic data/histone_marks/100K/

Get help

Usage:
    ./deep_hic_integrator <HIC_FILE> <HM_PATH> [--resolution INT]
                                               [--chr_train INT]
                                               [--chr_test INT]
                                               [--hist_mark_train STR]
                                               [--square_side INT]
                                               [--epochs INT]
                                               [--batch_size INT]
                                               [--encoder STR]
                                               [--decoder STR]
                                               [--output PATH]
                                               [--help]

Arguments:
    <HIC_FILE>                          Path of the Hi-C matrix file (.hic format)
    <HM_PATH>                           Path of the repository containing the histone mark files

Options:
    -r, INT, --resolution INT           Resolution representing the number of pair-ended reads
                                        spanning between a pair of bins. [default: 25000]
    -a INT, --chr_train INT             Chromosome used to train the autoencoder [default: 1]
    -t INT, --chr_test INT              Chromosome used to test the autoencoder [default: 20]
    -m STR, --hist_mark_train STR       Name of the histone mark used to train the autoencoder
                                        [default: h3k4me3]
    -n INT, --square_side INT           Size N*N of a sub-matrix [default: 20]
    -p INT, --epochs INT                Number of epochs for the training [default: 50]
    -b INT, --batch_size INT            Batch size for the training [default: 64]
    -e STR, --encoder STR               Trained encoder model (H5 format) [default: None]
    -d STR, --decoder STR               Trained decoder model (H5 format) [default: None]
    -o PATH, --output PATH              Output path [default: results/]
    -h, --help                          Show this

Documentation

The documentation is generated with Sphinx and built on ReadTheDocs.

Author

Hélène Kabbech : Bioinformatics master student intern at the Medical Center University of Goettingen (Germany)

License

This project is licensed under the GNU License.