swappedDrops on STARSolo output

[mbatiuk]

Hi,

Is there any chance swappedDrops could be run on STARSolo output?

Currently STARSolo outputs barcodes.tsv, features.tsv and matrix.mtx but no h5 files.
While swappedDrops requires h5 file

Thanks

[LTLA]

tl;dr No.

The long answer is that you can use removeSwappedDrops() to run the same algorithm on any molecule-level information, regardless of format. The emphasis here is molecule-level; you need the UMI, assigned gene and assigned cell for each individual transcript molecule. This is what is returned by Cellranger in their molecule information HDF5 file.

It is not sufficient to run this on the count matrix, which aggregates the molecules into counts at the gene level. This discards information about individual molecules, preventing swappedDrops() from making a decision about whether a particular molecule is a swapping artifact. I'm not familiar enough with Starsolo to know whether molecule-level stats are generated.

[mbatiuk]

OK, thanks for your response. STARSolo can output BAM/SAM files but not HDF5

I will ask STAR developers if there is any way to get HDF5 or maybe add this functionality

[LTLA]

Just so we're clear: the HDF5 file format is not a requirement. The requirement is just to get molecule-level information. Cellranger happens to store this in a HDF5 format, but processing pipelines are free to use whatever format they want; as long as you can get the relevant pieces of information into R, you can call removeSwappedDrops() on it. Of course, if the pipeline produced a Cellranger-style HDF5 file, this would be easiest for everyone, but I won't presume to dictate formats to others.

[mbatiuk]

OK, I corrected question to STAR developers to be more broad than HDF5

On the other hand, could BAM/SAM files provide this info?

OR if I understand correctly, a lot happens during read count/collapsing UMI barcodes and correcting for UMI barcode errors that STAR performs so SAM is not a good source of molecule info

[LTLA]

On the other hand, could BAM/SAM files provide this info?

Possibly. I've seen other cases where the processing pipeline deposits the identity of the assigned gene and cell barcodes into the SAM tags. One could also store the UMI sequence and indicate whether a particular read is a duplicate for a molecule.

In practice, custom BAM tags are a pain to work with; this would need to be handled on a case-by-case basis.