search

dozmorovlab / HiCcompare

Joint normalization of two Hi-C matrices, visualization and detection of differential chromatin interactions. See multiHiCcompare for the analysis of multiple Hi-C matrices
https://dozmorovlab.github.io/HiCcompare/
Other
18 stars 3 forks source link

Loess normalization of balanced data #10

Closed Phlya closed 5 years ago

Phlya commented 6 years ago

Hi. Do you think it's wrong to use balanced (iteratively corrected) data in loess normalization? Why I am asking, is currently HiCcompare doesn't detect some true differences in my datasets (i.e. "obvious" differences by eye, and validated biologically). Maybe there are some other ways to tweak the algorithm, but I can see that balancing the data improves fold change for some of them a lot... So is there a reason not to do this?

jstansfield0 commented 6 years ago

I would recommend playing with the filtering before trying to use ICE'd data. You can use the filter_params() function to get a better idea of what level of filtering you should be using.

As for using the ICE'd data, yes it will be accepted by the functions and should run fine. But you might be over normalizing the data.

Phlya commented 6 years ago

OK, thanks. Using filter_params is slightly challenging because I run everything on the cluster... Maybe I can save the object and then load it locally? I'll give it a go.

jstansfield0 commented 6 years ago

Try saving the hic.table for a single chromosome as an object after you normalize it and then load it locally to use filter_params.