open2c / coolpuppy

A versatile tool to perform pile-up analysis on Hi-C data in .cool format.
MIT License
78 stars 12 forks source link

[Q]When the new version performs rescaling pileup analysis, there is a problem with the ignore_diags parameter #124

Closed WANGchuang715 closed 1 year ago

WANGchuang715 commented 1 year ago

I noticed that when the new version performs rescaling pileup analysis, the same ignore_diags parameter presents different results, and the new version does not go to the masked diagonal area.However, the new version of On-diagonal pileup analysis does not have this problem. A1 TAD domains pileup (16) ![A1 TAD domains pileup (17)](https://user-images.githubusercontent.com/67625036/217697467-a18d5f16-ed48-45c6-930b-d1f08d5 A1 TAD boundary pileup (4) 44b25.png)

I'm not sure if it will have an effect on the result, because I need to calculate domain_score, or do you have any suggestion for ignore_diags parameter?

Phlya commented 1 year ago

Could you please clarify exactly what you did, what you expected and what the problem is?

efriman commented 1 year ago

I think I found out what the issue was and it should be fixed. Please try again after updating using the latest github version

WANGchuang715 commented 1 year ago

I think I found out what the issue was and it should be fixed. Please try again after updating using the latest github version Thank you for your modification, the display is now normal, and the domain_score has also changed. image Different ignore_diags parameters will produce different scores, do you have any suggestions for this parameter setting?

Phlya commented 1 year ago

Great to hear! Thank you for fixing this @efriman

In general in most cases ignore_diags=2 is a good choice. But sometimes it's a little data dependent, in particular for rescaling.

WANGchuang715 commented 1 year ago

Great to hear! Thank you for fixing this @efriman

In general in most cases ignore_diags=2 is a good choice. But sometimes it's a little data dependent, in particular for rescaling.

Thanks for your advice.