Meet-EU Team SA2, Topic A : Prediction of TADs
Based on results of the benchmark done on TADs prediction tools (Dali and Blanchette, 2015), we designed a CNN in order to discriminate TADs border, given a (1,33,33) windows from HiC data. In the aim of reconstructing full TADs, we also try to predict the length associated with a given border.
- numpy == 1.19.5 (https://numpy.org/doc/stable/index.html)
- scipy == 1.7.1 (https://scipy.org/)
- tensorflow-gpu == 2.6.1 (https://www.tensorflow.org/install)
- HiCtoolbox (available in the repository)
Download model.zip file, unzip it at the location of your choice.
In the predict.py file, variable "HiCfilename" (line 10), enter the path of your HIC map. You may want to change the name of the output file, this can be done line 41
Note: in order to use a CPU instead of a GPU, please change the shape format order from (1,33,33) to (33,33,1) as CPU does not support channel first.
Maxime Christophe, Antoine Szatkownik, Wiam Mansouri
HiCtoolbox module was developed and share by Léopold Caron and Julien Mozziconacci