[Q] How to test significance of change between pileups?

[AudreyBaguette]

State the question
I generated a set of pileups, all using the same list of regions, but with different Hi-C files. How do I test if the resulting pileups are significantly different from one another?

What have you tried?

• I tried finding an existing method, but didn't find any
• Methods form a paper suggested separating the pileup to get each snip, computing one pileup value per snip (ex the mean of the middle 3x3 pixels), then comparing the distribution of snip values between the two Hi-C files with a t-test. Cooltools permits to find the values for each snip, but not coolpup

Additional context
While the method stated above, using cooltools, could work for pileups using a set of punctual regions (like CTCF peaks), cooltools isn't able to take as input a set of ranges (like compartments). It'd like to generate pileups for various Hi-C maps (with different coverage) then see if the pileups I obtain are significantly different from each other considering the changes in coverage.

[efriman]

There's no statistical tests implemented in coolpup.py, but you can extract contact frequency distributions along the "stripes" (store_stripes=True or --stripe). I'm not sure if this is what you mean but you can see an example of how to get the values here: https://github.com/efriman/Friman_etal_ULI/blob/main/pileups/TSS_quartiles_score_stripes.ipynb