TOAST：a method for identifying Tad bOundaries based on graph Auto-encoders and clustering

1. before running, you need to install the following package or tools.

Pytorch 1.7.1, Python 3.6.12, Numpy 1.19.5, Matplotlib 3.3.3, Cooltools 0.4.1, Fanc 0.9.24, Jupyter 1.0.0, Pandas 1.1.5, pyBigWig 0.3.18, seaborn 0.11.1.

intall conda and create the virtual environment.

download the Anaconda3-2020.07-Linux-x86_64.sh file from https://repo.anaconda.com/archive/

bash Anaconda3-2020.07-Linux-x86_64.sh

conda create --name pytorch1.7.1 python=3.7

intall packages using pip

pip install flask

pip install numpy

pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu

pip install matplotlib

pip install seaborn

pip install hdbscan

2. run the web server version

When the environment is prepared, you can downlaod the TOAST-webserver.zip, and just run the webserver.py using the following code:

conda activate pytorch1.7.1

python webserver.py

visit the web linker: http://127.0.0.1:5000

3. prepare data

you can download the simulated Hi-C contact from paper "Forcato, M. et al. Comparison of computational methods for Hi-C data analysis. Nature methods 14, 679-685 (2017)."

you can download the real Hi-C data of GM12878 cell line with GEO ID GSE63525 is provided by the Rao Lab.

For .hic format data, you can dump data using juicer_tools.js and get the NxN matrix, where N is the number of equal-sized regions of a chromosome.

in this paper, we use the NxN format matrix as input.

4. Content of folders

call codes for identifying, ploting, and evaluating TADs.

5. Input matrix file format:

The input to CASPIAN is a tab seperated N by N intra-chromosomal contact matrix derived from Hi-C data, where N is the number of equal-sized regions of a chromosome.

6. output

the output is the TAD file at the merge file.

running

enter the /src folder, you can do the following process: (1)toast.py: run to get and plot TADs. you can run it by justing changing the input and output folder.

(2)plot_heatmaps.py: plot heatmaps for all TADs

(3)evaulate_TAD.py:evalute TADs

(4)evaulate_TAD-simnarity.py: compare the similarity between TADs.

(5)evaluate-achor-count.py:compute the anchor count for Chip-seq factors.

(6)plot_TAD_markers.py: plot the CHip-seq tracks.

(7)plot-man-figures.py: plot figures in this paper.