Support additional simpleaf chemistries and orientations

[jeremymsimon]

Description of feature

Currently, there doesn't seem to be a way to specify all the 10x read chemistries supported by simpleaf here:
https://github.com/COMBINE-lab/simpleaf/blob/main/resources/chemistries.json

In particular, I'd like to run nf-core/scrna via simpleaf on 10x 5' v2/v3 chemistry, which doesn't seem to be currently supported in this workflow (from what I can tell). Can we either enable the specification of any supported simpleaf chemistry on the above list?

Alternatively, can we create a parameter to pass additional custom options to simpleaf a la extra_salmon_quant_args in the rnaseq workflow? If we went that route, we could pass the -d, --expected-ori parameter to get the orientation correct for 5' chemistry, as noted in my issue here COMBINE-lab/alevin-fry#118 and discussed elsewhere

Thanks!

[grst]

Any argument to --protocol that is not interpreted by the scrnaseq pipeline will be passed in verbatim to the underlying tool. So it should be possible to specify any of the chemistries from that JSON, or even the geometry strings.

LMK if you have issues with that approach.

[jeremymsimon]

Oh, interesting! I had no idea it worked that way. Will give this a try and report back, but if this indeed works it would be helpful to note in the documentation that this is possible

[jeremymsimon]

Hi @grst so I think this works in principle, but in practice there's an issue with the simpleaf quant step that I suspect needs an edit.

If I run 10x 5' v3 data via --aligner simpleaf --protocol 10xv3-5p, it attempts to run simpleaf quant with the following command:

simpleaf quant \
      --t2g-map t2g_3col.tsv \
      --chemistry 10xv3-5p \
      --index index \
      --reads1 Pool176_1_S1_L001_R1_001.fastq.gz,Pool176_1_S1_L002_R1_001.fastq.gz \
      --reads2 Pool176_1_S1_L001_R2_001.fastq.gz,Pool176_1_S1_L002_R2_001.fastq.gz \
      --resolution cr-like \
      --output simpleaf_quant \
      --threads 6 \
      --anndata-out \
      --unfiltered-pl [] \

which subsequently fails with the error The provided path [] does not exist as a regular file.. I think it's because there is no whitelist provided via the protocols.json for 10xv3-5p

I saw some chatter on the nf-core slack awhile back when @FelixKrueger first implemented this (related to #263) so I wondered whether I could override this --unfiltered-pl parameter by passing --knee via my nextflow.config file, but this then instead complains saying the argument '--unfiltered-pl [<UNFILTERED_PL>]' cannot be used with '--knee' because it seemingly doesn't replace --unfiltered-pl but rather it simply passes --knee in addition.

If I manually remove the --unfiltered-pl [] from the simpleaf quant step in the work directory .command.sh and replace it with --knee and run this via bash .command.run, it seems to proceed fine

Everything also "works" ie populates --unfiltered-pl 10x_V3_barcode_whitelist.txt.gz appropriately if I run this via --protocol 10XV3 since it grabs the whitelist specified in the nf-core json file, but I'm not sure this will process my 5' data correctly (according to https://github.com/COMBINE-lab/simpleaf/blob/main/resources/chemistries.json it appears to reference different whitelists)

So overall I think there could be two interrelated issues here: one is that changing the knee/whitelist specification for simpleaf doesn't seem to work (at least in my hands) since it's not replacing the default unfiltered-pl option. The other is that when a chemistry is not in the protocols.json there needs to be a way to automatically or manually fill in the appropriate whitelist or switch to one of the other barcode selection methods. I suspect I could pass an expected number of cells via the samplesheet to circumvent this, but for my samples this number is unknown so I haven't tested it

Let me know if I've missed anything here- happy to test and provide feedback if helpful

Thanks!