search

Hughes-Genome-Group / deepC

A neural network framework for predicting the Hi-C chromatin interactions from megabase scale DNA sequence
GNU General Public License v3.0
33 stars 12 forks source link

Prediction of interactions without being skeleton transformed #8

Open ahmedabbas811 opened 5 months ago

ahmedabbas811 commented 5 months ago

Hi all,

I understand that the training was done using skeleton transformed Hi-C, so the outputs are also skeleton transformed.

How to get the output predictions without being skeleton transformed? Is there a way to do that?

Thanks

rschwess commented 1 month ago

Hi,

I'm afraid we never implemented a back transformation.

In theory, one could record the average interaction strength per skeleton bin and then try to align the predicted score back to interactions using that but this is purely a suggestion.

But we never tried that.