search

XiaoTaoWang / NeoLoopFinder

A computation framework for genome-wide detection of enhancer-hijacking events from chromatin interaction data in re-arranged genomes
Other
60 stars 16 forks source link

CNV adjustment #26

Open DittmanC opened 3 years ago

DittmanC commented 3 years ago

I have adjusted the CNV using calculate-cnv, segment-cnv, and correct-cnv. After that, I was trying to see in the triangular plot of the cool file at the same region. Basically, it calculated chr5 at ~110mb region has the lost of one copy. But why after I did the adjustment based on the CNV profiling, I see the whole blank regions at ~110mb instead of intensifying the signal intensity due to copy umber loss? chr5_CNV

XiaoTaoWang commented 3 years ago

Hi,

This is really interesting! I think one reason that might induce the artificial blank regions is that those regions were somehow filtered out at the beginning of the normalization procedure. To make the test easier, I just released a new version (0.2.4.post1) that makes the parameters of the CNV normalization tunable. Can you upgrade your NeoLoopFinder to this version (pip install -U neoloop) and increase the value of the "--mad-max" parameter when you run correct-cnv? (also make sure to specify "-f" to overwrite the existing "sweight" column in your cool file).

Another way to avoid this is running correct-cnv on higher-resolution matrices (in your case, perhaps 50K/25K/10K, ...), as insufficient number of valid bins with specific copy number values might also introduce noises into the normalization procedure.

Let me know if these solutions work!

Xiaotao

DittmanC commented 3 years ago

just updated and tried again. In the segment-cnv 90mb-115mb region should be losing one copy. I have compared --mad-max 0 with cnv adjustment and without cnv adjustment. They look the same.

Screenshot 2021-10-31 at 12 10 30 AM

While I have also tried --mad-max 100 and --mad-max 0. They also look the same.

Screenshot 2021-10-31 at 11 27 14 AM

Why I cannot see the adjustment because of the lost of copy? Is the correct-cnv will only remove the CNV region instead of compensate the weaken of signal due to copy number loss?