Visium HD 3'

[ktpolanski]

Hello,

10X has released Visium HD 3', and actively advertises the platform's VDJ capabilities when combined with long read sequencing. However, the technology comes with proprietary spatial barcode/UMI processing that is carried out under the hood by Spaceranger, and the end result of the prescribed long read processing is a BAM with back-propagated CB/UB tags. The CBs in particular come in the form of spot names rather than barcodes.

Is there some way to shoehorn this into MiXCR?

[mizraelson]

I think the easiest workaround at the moment is to convert the BAM file to FASTQ, taking the barcodes from the tags and splicing them with the read sequence. Based on the description, they should contain the uncorrected barcode nucleotide sequences:
• CR - Visium barcode sequence as reported by the sequencer
• CY - Visium barcode read quality (Phred scores as reported by the sequencer)
• UR - Visium UMI sequence as reported by the sequencer
• UY - Visium UMI read quality (Phred scores as reported by the sequencer)

If you create a script that produces a FASTQ file with the structure:

CB-UMI-Read

then you can run MiXCR on that data.

[ktpolanski]

That's fair. I'm afraid that the CR/UR sequences may not be of consistent lengths, as per 10X tech support there's no consistent R1 structure anymore. However, I can in principle just generate fake cell/barcode "reads" and tell MiXCR to process them. How do I tell it to perform no cell barcode/UMI correction? That would make it the easiest to generate fake ones.

[ktpolanski]

I pulled up a prior query and I see mention of --dont-correct-tag-type. I also see mention of me having shoehorned stuff into a generic preset with --tag-pattern "^(UMI:N{28})\^(R2:*)", which would be fine enough when combined with bespoke downstream post-processing (also conveniently dumped in that thread). So it's something at least!

[mizraelson]

Inconsistent length is not an issue. You can add anchor points when you join barcodes and then use a pattern like:

^(CELL:N{10:15})atgcatgc(UMI:N{10:15})(R1:*)\^(R2:*)

In this case CELL barcode and UMI can be anything between 10 to 15bp as long as the are separated by atgcatgc.

You can disable correction if you run the commands separately:

mixcr align ...
mixcr refineTagsAndSort --dont-correct-tag-type Cell --dont-correct-tag-type Molecule  ...
mixcr assemble ...
mixcr export clones ...

[ktpolanski]

This does kind of loop back around to the problem from the earlier discussion, where the generic pacbio didn't support cell tags (which I'm pretty sure would also translate to general ONT, which this one will be). Or has that changed since?