Add process/script to classify with singler

[sjspielman]

Stacked on #176
Closes #166

This PR adds the R script and nextflow process to classify samples with SingleR. While working on this, I realized we actually haven't been as comprehensively testing as we maybe want to be (#177), so I'm opening this PR with SCPCP000004 added as project to run through in the stub profile so we can ensure this really is passing CI. I will remove that once this passes to request review.

[sjspielman]

The stub workflow passed with SCPCP000004 included at this commit: ae98ecf. See run: https://github.com/AlexsLemonade/OpenScPCA-nf/actions/runs/17305059525/job/49125753586

So, I removed SCPCP000004 from the stub projects since it passed and can now request review.

[allyhawkins]

This LGTM, but I will note that we should be testing this with actual data too and not just with the stub in CI. I might run this through with the simulated option for data just to be sure that the scripts also run as expected.

[sjspielman]

This LGTM, but I will note that we should be testing this with actual data too and not just with the stub in CI.

I confess I was waiting until the end for a big debug, but yeah, fair.... Spam incoming 🙃
edit - i'm going to try running locally first before running via triggering the GHA, so maybe no spam :)

[sjspielman]

Coming along nicely with testing here, until I hit this roadblock EOD Friday at the classification step and ran away from it for the weekend! This error is...confusing:

Caused by:
  No such variable: file -- Check script '/home/ec2-user/.nextflow/assets/AlexsLemonade/OpenScPCA-nf/./modules/cell-type-neuroblastoma-04/main.nf' at line: 103


Source block:
  """
  for file in ${library_files}; do
    classify-singler.R \
      --sce_file ${file} \
      --singler_model_file ${singler_model} \
      --singler_output_tsv \$(basename \${file%.rds}_singler.tsv.gz) \
      --threads ${task.cpus}
  done
  """

The error is complaining about the file loop variable. I did a bunch of local testing to see what might be going on here, and the library files don't seem to be staged. Only if I do something like echo $library_files (directly use the library_files variable in the script block) do the files get staged and no error is thrown. After much, much staring, I can't find any syntax differences between this and, e.g. doublet-detection or scimilarity module code, in terms of reading in the libraries_ch and looping over the files.

Anyways, I'm stumped! I've just submitted a batch run with scimilarity turned on (to confirm these files are being staged - they definitely are!) and with a publishDir for the classify process as well as an emit.. I don't think this will help, but I as mentioned, am stumped! https://cloud.seqera.io/orgs/CCDL/workspaces/OpenScPCA/watch/KkqiAAa4Dm3Pn/v2/tasks

Something else I want to mention, the train_singler process takes an unbelievable amount of memory - seemingly (?) way more memory than I experience when running it locally. It gets bumped to mem 48/64 to get a successful run. Something to keep tabs on?

[jashapiro]

The error is complaining about the file loop variable

Since that is a bash variable, you need a backslash:

--sce_file \${file} \

[sjspielman]

And we're back on track 😄 https://cloud.seqera.io/orgs/CCDL/workspaces/OpenScPCA/watch/3eSN9IU2ATPmtz/v2/tasks

[allyhawkins]

LGTM, thanks for testing!