hemberg-lab / SC3

A tool for the unsupervised clustering of cells from single cell RNA-Seq experiments
http://bioconductor.org/packages/SC3
GNU General Public License v3.0
119 stars 55 forks source link

No 'dimnames' attribute for array #34

Closed JRWK closed 7 years ago

JRWK commented 7 years ago

Running SC3 on a dataset, post QC and normalisation. The clustering completes, followed by calc_biology, but upon attempting to plot_de_genes or plot_markers, I get the following error:

Error in dataset[names(de_genes), , drop = FALSE] : no 'dimnames' attribute for array

The fData of the dataset is properly updated with markers_clusts, markers_padj, and de_padj columns, which leads me to believe that it might be a plotting error?

wikiselev commented 7 years ago

Hi, thanks for letting me know about this issue! Could you please share your SCESet (after running SC3) with me (vk6@sanger.ac.uk)? And also could you please send me the exact command you run, that produces an error? Additionally, I will also need an output of

library(SC3)
sessionInfo()

and

source("https://bioconductor.org/biocLite.R")

Many thanks, Cheers, Vlad

wikiselev commented 7 years ago

Hi John,

It looks like the information about sample names and gene names is lost in your SCESet object. For example, if you run:

exprs(regression_scran_sum)[1:3, 1:3]

you will see that neither columns nor rows have any names. This is a problem which shouldn't happen with default scater functions unless your overwrite the exprs slot manually.

Looking at your object name, I suppose you have performed scran normalisation, which had overwritten the exprs slot in your object, and the new matrix does not contain any names, therefore SC3 plotting functions don't work. In this case, this would be a scran bug. Or maybe lack of scater checks...

Could you please find out at which step of your analysis colnames and rownames of exprs(regression_scran_sum) disappear and then let me know? I have both developers of scater and scran here on campus and quickly contact them.

Cheers, Vlad