Closed dbdimitrov closed 2 months ago
Hi @dbdimitrov, do you confirm that condition_key
is "condition"
?
I also get a NaN pvalue for the AAED1 gene. This is a priori not a bug: pydeseq2 filters out p-values based on Cooks outliers (cf the docs).
You can turn this off by setting cooks_filter=False
when initialising a DeseqStats
object.
That being said, it seems like the ctdata
anndata has floats as counts, whereas ints are expected. Converting counts to ints in the piece of code you provided changes the results:
With floats:
Converting to ints:
I'll dig a bit more into this but I think that we should raise an error (or at least a warning) in case counts are not ints (and / or maybe cast to int when possible).
Hi @BorisMuzellec,
Yes, indeed condition_key
was "condition" - sorry.
Thanks a lot for the clarification!
Hi,
Thanks for developing this package and for improving it.
Describe the bug I noticed a bug where while all other stats are calculated, some p-values are assigned to NaN.
To Reproduce small adata: https://drive.google.com/file/d/1wwZDcsEVZD0ldBrtP63vqA1zTPpM05Q2/view?usp=drive_link
code
pydeseq2 version:
Expected behavior Throw a warning? Is it expected to have NaNs here? The counts look like there is some variance, etc.
Desktop (please complete the following information):
Exception Thrown when using the problematic gene alone
The first guess on the deviance function returned a nan. This could be a boundary problem and should be reported.
Let me know if I can provide any further info.