All data (including example data) producing gene count matrix with 0s

[IzzyJRussell]

Hi, we have been implementing scPipe into a scRNAseq analysis and have been unable to produce a gene count matrix with any expressed genes. Therefore we have run the example data (copying the code directly from the tutorial in the documentation) to check for issues with scPipe instead of the data input.

We found that when running the example code exactly the gene_count.csv produces a table full of 0s implying that there is no expression from the detected genes. Is this an intended result for the example data? If this was not intended what would be causing this issue?

[IzzyJRussell]

I believe it may be linked to this issue from 2018?

[Shians]

Hello Izzy,

Thanks for your issue report. The example isn't meant to produce a table full of 0s. It's unclear what kind of bug is causing this and I am looking into it.

Kind regards,
Shian

[Shians]

Sorry @IzzyJRussell for the delayed response, I've been travelling for the past couple of weeks. I've narrowed the issue down to the fact that sc_demultiplex(file.path(data_dir, "out.map.bam"), data_dir, barcode_annotation_fn,has_UMI=FALSE) has has_UMI=FALSE set.

@LuyiTian will have to confirm. But I think setting has_UMI=FALSE uses the read names as UMI and causes each UMI to only have count 1 which is then filtered out by UMI_correct1. Setting has_UMI=TRUE produces counts for the example, but I'm not sure why it was set to FALSE in the first place.

[IzzyJRussell]

Thanks for getting back to me, we determined that in this instance using the Rsubread featurecounting function would be more suitable for our reanalysis - I wish you luck resolving the issue.

[jdumdie]

In the tutorial, it's recommended for full-length protocols like SMART-seq, to set has_UMI=FALSE in the sc_demultiplex function. What do you recommend for SMART-seq protocols that have used UMIs? Is there a work-around to get the read counts out? I am just getting a matrix of zeros, as above.