Closed ATpoint closed 3 years ago
Hi Alexander
Thank you posting this bug. There was indeed a problem with how the coefficients were named when using a batch indicator and the glmGamPoi method.
I have fixed this in the develop branch, which you can install via remotes::install_github("ChristophH/sctransform@develop")
Can confirm that this works just fine, thank you for the prompt reply. Might I suggest that you add a note in the README until the next release since I know at least one other person stumbling over this.
@ChristophH After loading v0.3.2 from CRAN, exact same code as above I see:
Calculating cell attributes from input UMI matrix: log_umi
Error in is.nan(rel_attr) :
default method not implemented for type 'list'
Works fine when going back toremotes::install_github("ChristophH/sctransform@develop", ref = "af8ba07")
. Any idea what this is?
Ah, sorry did not realize this was already pointed out in https://github.com/ChristophH/sctransform/issues/88
Hi Christoph,
running
sctransform:.vst()
with methodglmGamPoi
produces only NAs for residual_mean and residual_variance in case thatcattr
andbatch_var
were specified.It is a 54295 x 10116 experiment with two batches. Log and sessionInfo below, I do not have a second dataset here on my machine right now, but this seems to be specific when using the glmGamPoi method, all other methods produced the resudial_variance column as expected. Can you comment on this, something I do wrong on my end?