hclust2-1.0.0 :: test failure

[EricDeveaud]

Hello,

I just installed hclust2 from tagged release 1.0.0, in order to be abble to generate heatmaps form metaphlan outputs.

in order to test it I ran the test as described here: https://github.com/SegataLab/hclust2/blob/master/examples/HMP-MetaPhlAn/run.sh using the provided HMP.species.txt file

ang got the same error than: #3
and also: biobakery forum #1732 #1705

any insight ?

NB

python/3.8.1
cycler==0.10.0
hclust2==1.0.0
kiwisolver==1.3.1
matplotlib==3.4.2
numpy==1.21.0
pandas==1.3.0
Pillow==8.3.0
pyparsing==2.4.7
python-dateutil==2.8.1
pytz==2021.1
scipy==1.7.0
six==1.16.0

regards

Eric

[EricDeveaud]

NB in order to check I change all the 0.0 in HMP.species.txt to random values
this "fixed" the only finite values

this yield to the next error which is:

  File "/opt/gensoft/exe/hclust2/1.0.0/venv/lib/python3.8/site-packages/hclust2-1.0.0-py3.8.egg/EGG-INFO/scripts/hclust2.py", line 74, in inverse
AttributeError: module 'matplotlib.cbook' has no attribute 'iterable'

regards

Eric

[saras224]

Hey @EricDeveaud
Does anybody know how hclust2 determine the --ftop n features based on --fperc.. in the example they have used --fperc value 99 but I want to know the logic behind that. Why specifically 99? why not 100? How to decide the percentile value?

Thanks

[EricDeveaud]

see: https://forum.biobakery.org/t/hclust2-py-error-distance-matrix-finite-values/1732/2