using alevin-fry in Nextflow pipeline nf-core/scrnaseq?

[Khajidu]

I heard alevin-fry would soon replace alevin in the Nextflow pipeline nf-core/scrnaseq, but how can I use it in the meantime? I have data that look like 10XV1 data (1 index file and 2 read files for each read pair) for info (I cannot share them unfortunately).

[rob-p]

Hi @Khajidu,

Thanks for bringing this up. What organism are you working in? Is your data a standard 10x v1 library? If so, we can demonstrate the mock commands with another public dataset that should then work with your data. @DongzeHE, could we please put together a small set of commands that will run some 10x v1 data through the alevin-fry pipeline?

Thanks!
Rob

[Khajidu]

I'm working on catshark brain samples, the data has been sequenced with 10XV3 technology, but is organized in 3 files per read pair as described above, as if it were 10XV1 data. Treating it as V3 data gives no mappings.

[DongzeHE]

Unfortunately, alevin and alevin-fry both are based on pufferfish, so if you get no mappings with alevin, I would not think running alevin-fry will make any difference. Instead, do you think that there is any chance that the reference transcriptome used for building salmon index has something wrong?

If possible, could you please share the commands you used for building your salmon index?

[rob-p]

So, just for some context @DongzeHE — these other tools were run through the nf-core scrnaseq pipeline. So you can find the exact commands used there.

It's possible mapping issues are due to the reference or data quality etc., but 0 is a very low mapping rate. If the library preparation is, in fact, non-standard though, that would perhaps explain what is going on.

[Khajidu]

In fact, I get a non-zero mapping rate... over 0 sequences (I work from a genome and a GFF and it gives me 50k sequences, but it seems that none of them is being mapped on)

[rob-p]

Right, sorry — what I mean by mapping rate here is the number of sequenced reads from your experiment that are assigned to some target transcript. However, in all of the single-cell pipelines, it's possible to have a non-zero mapping rate, but still not have valid cells identified due to the way valid cell barcodes are determined, filtered, etc.

[Khajidu]

oh that would be why... I'll be looking into this tomorrow.