Closed BenxiaHu closed 1 year ago
Hello, by default it's the mean of the central 3x3 square of the pileup, which would typically be observed/expected enrichment, and indicative of average dot strength. Hope this helps.
thanks. when I ran coolpup.py
coolpup.py 10000.mcool::/resolutions/10000 loop_coolpuppy.txt --features_format bedpe --expected 10000_expected.tsv -o 10kb_cooluppy
do you think observed/expected enrichment is log(observed/expected) value or not?
The plots is probably displayed in log scale, but the values are not.
thanks. how to show log(observed/expected)? add this paramter --scale log
the scale in the right of the heatmap is too large (0.083-12). looks weird.
The colour is log scaled, that's why it looks like this. In the CLI you can't do it, sorry, only using python API.
okay. I am trying to use API
ImportError: cannot import name 'plotpup' from 'coolpuppy' (/python38/lib/python3.8/site-packages/coolpuppy/init.py)
I can import coolpuppy, but not from coolpuppy import plotpup. do you know what causes such issue?
Maybe you are using old coolpuppy?
Could you please post the full error?
I just upgraded the coolpuppy to 1.0.0
here is the new error. plotpup.py --input_pups _significant_Short_10kb_cooluppy significant_Median_10kb_cooluppy \
significant_Long_10kb_cooluppy --rownames Short Median Long --colnames test \ --output significant_diff_10kb_cooluppy_mustache.pdf --scale log
ValueError: Multiple pileups for one of the conditions, ensure unique correspondence for each col/row combination
Yes, you need to specify what goes along rows vs columns. For example, I think you need --rows features
.
thanks. that works.
I have another question, given that I split the chromatin loops into 3 classes based on the loop length, do you think that I need to --rescale
to rescale pileups to the same shape and size.
based on my knowledge, the long-range loops should show lower contact frequencey compared to short-range loops.
but my results look weird. I am not sure whether the coolpup.py causes this phenomenon
is there any paramter to do fold enrichment of the center piexl over the nearby bins?
Try --norm_corners 3
, for example.
it works. but the value is > 2. do you think the result makes sense?
Well, without seeing the result, sounds good, makes sense. You have >2-fold enrichment of O/E interactions relative to background.
Makes sense to me!
but usually the fold enrichment is greater than 1 but less than 2.
It really depends on your data, resolution, regions you chose... For specific situations I had enrichment >10.
hello, thanks for developing the coolpuppy tool. it is very helpful. I have one question: what does APA score mean? would you like to explain it? Best,