New module: lsa/cosine (calculate cosine similarity)

[miguelrosell]

Hi @ewels and @pinin4fjords!

Following up on our chat in the proposals channel, here’s the PR for the cosimflow module. It’s a tool for calculating pairwise cosine similarity from expression matrices (outputting both the matrix and a heatmap).

I think it could be a handy addition for the rnaseq pipeline, maybe as a downstream QC step to see how samples cluster together beyond the usual PCA.

PR checklist:

Closes: xxx

• This comment contains a description of changes (with reason).
• If you've fixed a bug or added code that should be tested, add tests!
• If you've added a new tool - have you followed the module conventions.
• If necessary, include test data in your PR (Created a small dummy CSV on-the-fly in the test script).
• Remove all TODO statements.
• Broadcast software version numbers to topic: versions.
• Follow the naming conventions.
• Follow the parameters requirements.
• Follow the input/output options guidelines.
• Add a resource label (process_medium).
• Use BioConda and BioContainers.

Tests performed locally:

• nf-core modules test cosimflow --profile conda

Quick note: I ended up putting the R script directly inside the main.nf script block. It was the cleanest way I found to handle the dollar signs and variable escaping for R/Nextflow without everything blowing up.

[pinin4fjords]

Thanks for the PR!

There's a few things to do to make this fit in with repository standards, and I think there's some complexity to be stripped.

In particular, since there's no actual software called 'cosimflow', I think this needs to be named like lsa/cosine in my view, since it's principally a wrapper around that function. Or possibly it could be a script under custom/.

I've simplified the script a bit to apply some of my suggestions for you:

pr10002_cosimflow_template.R.zip

[pinin4fjords]

The script constructs c("--input", "$expression_matrix", "--out_prefix", "$prefix") then parses them back into opt$input and opt$out_prefix - getting the same values it started with. Only task.ext.args actually needs parsing. Feed $args directly to optparse for configurable options (--method, --min_gene_mean), and use Nextflow template variables for input/prefix. See attached template.

[pinin4fjords]

if the outputs are deterministic, they should be snapshotted, if not you should snapshot file names at least, see e.g. how deseq2/differential module does it:

                { assert snapshot(
                        process.out.results,
                        ...
                        file(process.out.dispersion_plot_png[0][1]).name,
                        ...
                    ).match() },

[miguelrosell]

@pinin4fjords Thank you for all of corrections, comments and suggestions, I really appreciate it, for your patience and the detailed help!!

I hope I have applied everything you requested:

-Renamed it to lsa/cosine.

-I switched the input to CSV-only (removed readxl dependency) as suggested.

-I updated meta.yml (license/maintainers) and fixed the variable scope in main.nf so the R template works correctly.

• And the tests and stubs are now locally passing, with snapshots included.

• And I have fixed the non-English issue :)

If there's anything that I have missed or that needs a second look at, I'm all ears!

Again, thank you!

[pinin4fjords]

Much better!

[miguelrosell]

Hi @pinin4fjords!

I applied both changes:

• I took out readxl and dplyr from environment.yml
• I switched to dynamic versions in the stub.

Like always, I'm always here and willing to change anything I might have missed if necessary.
Thank you!! :)))

[miguelrosell]

Hi @pinin4fjords! I just updated the branch with master to resolve the out-of-date status.

Looks like the workflows are currently awaiting approval to run. Is the next step approving if the checks pass or am i missing something?

Have a good day and thank youu!!

[pinin4fjords]

Looks like the workflows are currently awaiting approval to run. Is the next step approving if the checks pass or am i missing something?

If you go on the nf-core Slack to the #github-invitations channel you can get yourself added to the org so that workflows will run for you!

[miguelrosell]

Hello, @pinin4fjords! I managed to fix all the schema requirements for the meta.yml and aligned the process name as well. I think all linting and tests pass with 0 errors now!
Ready to merge whenever you are.

Have a good Friday!