PyWGCNA

Question

[yoshifumimiya]

I am using pyWGCNA to analyze 37 samples x 50000 genes.

I got the following error in analyseWGCNA().

What should I do? Please give me your opinion.

[LauferVA]

please read about reproducible examples (reprex) and then alter your post. E.g., here: https://reprex.tidyverse.org/

[yoshifumimiya]

Thank you for your comment.

I apologize for the inadequacy of my question. I am having trouble understanding how to use "reprex" or "pyreprex". I am using Google colaboratory for python analysis. I have copied the more detailed code and output to Github and will attach it.

pyWGCNA question

Any valuable feedback from you would be greatly appreciated. Best regards.

[LauferVA]

Very nice!!! I will work on this at some time today when I have time. Until then, I can offer this:

the function that is throwing the error is not a part of the pyWGCNA package.

Instead, pyWGCNA calls Seaborn, and Seaborn calls matplotlib.

Now, within matplotlib, a function in _base.py() calls shrunk_to_aspect(), which ultimately throws the error. The best way forward is to google that exact error message. Doing that will show you a couple threads of people having the same problem in different scenarios: Sometimes you can study those scenarios and find a workaround.

Consider this thread. In their case, geopandas calls matplotlib, not seaborn, but the solution could be the same or similar to your case.

Will look more when have time.

[yoshifumimiya]

Thanks for your response!!

   if box_aspect <= 0 or fig_aspect <= 0:
    --> 545             raise ValueError("'box_aspect' and 'fig_aspect' must be positive")

    <Figure size 4500x0 with 2 Axes>

I believe box_aspect is the cause of the error.

import matplotlib.pyplot as plt
fig1, ax = plt.subplots()
ax.set_box_aspect(1)

According to Google, the above code was added as a similar error countermeasure. However, since pyWGCNA is a library, it could not be incorporated from the outside.

[LauferVA]

I believe box_aspect is the cause of the error.

Yes - I think it is, because the fig_aspect variable is set to 1 by default (transforms.py, line 522).

it could not be incorporated from the outside

That's true, BUT you could temporarily alter the the shrunk_to_aspect() function within transforms.py, or perhaps one of the functions in _base.py() file within the matplotlib package, by writing:

vi /usr/local/lib/python3.10/dist-packages/matplotlib/transforms.py

you could the hardcode the box_aspect variable, for instance, then change it back afterwards.

[yoshifumimiya]

Thank you for your valuable input!!

I will try your suggestion.

[LauferVA]

This is definitely not one of the easier / quicker debugging problems you come across - it is pretty difficult.

please feel free to continue posting updates if you get stuck.

VAL

[LauferVA]

Yoshi - Were you able to figure this out?

[yoshifumimiya]

Hi, VAL! I'm still trying to figure this out. Will be working on it this weekend.

I will report back with or without progress.

[yoshifumimiya]

This is definitely not one of the easier / quicker debugging problems you come across - it is pretty difficult.

Your thoughts are correct.

I have rewritten the code in transforms.py but noticed that it is not reflected.

PyWGCNA_question_2

I am thinking of incorporating "box_aspect = 1" if there is a way to reflect this..,