Kieran R Campbell

Thanks for the input @LTLA, we'll go for this option then

Interesting question and I'm not sure we tested this fully. There'll be a trade-off in terms of the more markers the better for assignment, but if the data you're assigning...

Are you making sure to only use marker genes as input to the function? (i.e. not tx-wide)?

Okay this looks like the same issue we're having fun with over in #52 which I think is a windows-specific can't locate correct version of tensorflow as per here https://github.com/rstudio/tensorflow/issues/95...

Hi @robertamezquita Yes `SingleCellExperiment`s are supported as the main input. Please see the [vignette](https://github.com/Irrationone/cellassign/blob/master/vignettes/introduction-to-cellassign.Rmd) or simply look at the _input format_ section in `?cellassign` Kieran

One big comment before investigating: the input should be marker genes only - it looks like you're inputting 19371 genes - are these really all markers? See e.g. the basic...

Hi @robertamezquita It's on the to-do list, and should be priority now. I'd been putting it off because of the tensorflow integration, which I'm not sure will play well with...

Thanks Aaron, this could be a lifesaver. The python/tensorflow experience has not been great. I'll have a look into this when back from holiday in a couple of weeks.

Hey @LTLA Just made a version of cellassign that uses basilisk, seems to work nicely from my limited testing. What's the overall plan for the package? Is it going into...

Looks like 2 problems here. Size factors not being carried over because I suspect `sizeFactors(sce)` is NULL. You should be able to recompute them by calling `sce