New module: lsa/cosine (calculate cosine similarity) #10002

miguelrosell · 2026-02-12T13:03:20Z

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

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 · 2026-02-12T16:46:29Z

modules/nf-core/cosimflow/main.nf

+    container "${ workflow.containerEngine == 'singularity' && !task.ext.singularity_pull_docker_container ?
+        'https://depot.galaxyproject.org/singularity/mulled-v2-ad9dd5f398966bf899ae05f8e7c54d0d94018089:00bd7e543b7135f299839496732948679f2e3073-0' :
+        'biocontainers/mulled-v2-ad9dd5f398966bf899ae05f8e7c54d0d94018089:00bd7e543b7135f299839496732948679f2e3073-0' }"


Suggested change

container "${ workflow.containerEngine == 'singularity' && !task.ext.singularity_pull_docker_container ?

'https://depot.galaxyproject.org/singularity/mulled-v2-ad9dd5f398966bf899ae05f8e7c54d0d94018089:00bd7e543b7135f299839496732948679f2e3073-0' :

'biocontainers/mulled-v2-ad9dd5f398966bf899ae05f8e7c54d0d94018089:00bd7e543b7135f299839496732948679f2e3073-0' }"

container "${ workflow.containerEngine == 'singularity' && !task.ext.singularity_pull_docker_container ?

'https://community-cr-prod.seqera.io/docker/registry/v2/blobs/sha256/4d/4d94f159b95315adf8bf54fdc9db88db10a5aef72dca6245dd163b91e9e0437e/data' :

'community.wave.seqera.io/library/r-base_r-lsa_r-pheatmap_r-optparse_pruned:901156bc11e60b28' }"

I made you some containers via Wave/ Seqera Containers. These are easier to maintain for multi-dependency modules.

pinin4fjords · 2026-02-12T16:53:04Z

modules/nf-core/cosimflow/main.nf

+    when:
+    task.ext.when == null || task.ext.when
+
+    script:


Use template 'cosimflow.R' with the script in templates/cosimflow.R instead of inline R, keeps things readable.

I'm going to attach a suggested version of your script content to use in that way.

pinin4fjords · 2026-02-12T17:41:35Z

modules/nf-core/cosimflow/environment.yml

+name: cosimflow
+channels:
+  - conda-forge
+  - bioconda
+  - defaults
+dependencies:
+  - r-base
+  - r-lsa
+  - r-pheatmap
+  - r-optparse
+  - r-readr
+  - r-readxl
+  - r-dplyr


Suggested change

name: cosimflow

channels:

- conda-forge

- bioconda

- defaults

dependencies:

- r-base

- r-lsa

- r-pheatmap

- r-optparse

- r-readr

- r-readxl

- r-dplyr

---

channels:

- conda-forge

- bioconda

dependencies:

- conda-forge::r-base=4.5.2

- conda-forge::r-lsa=0.73.4

- conda-forge::r-pheatmap=1.0.13

- conda-forge::r-optparse=1.7.5

- conda-forge::r-readr=2.1.6

Pin versions with channel prefixes: conda-forge::r-lsa=0.73.4 not just r-lsa

Remove defaults channel - nf-core only uses conda-forge and bioconda

pinin4fjords · 2026-02-12T17:43:57Z

modules/nf-core/cosimflow/main.nf

+    )
+
+    # AQUÍ PASAMOS LOS ARGUMENTOS DE NEXTFLOW DIRECTAMENTE
+    args_list <- c("--input", "$expression_matrix", "--out_prefix", "$prefix")


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 · 2026-02-12T17:44:27Z

modules/nf-core/cosimflow/main.nf

+
+    # ---- Load table ----
+    if (grepl("\\\\.xlsx?\$", opt\$input, ignore.case=TRUE)) {
+        df <- as.data.frame(read_excel(opt\$input))


Does it really need Excel input? Let's stick with CSV if poss please

pinin4fjords · 2026-02-12T17:49:00Z

modules/nf-core/lsa/cosine/tests/main.nf.test

+            // 3. Verificamos que termine con éxito
+            assert process.success
+
+            // 4. Verificamos que se generen los archivos


again, with apologies, English would be better

I am so sorry! I will keep everything in English in the future. Thanks for correcting it all!

pinin4fjords · 2026-02-12T17:51:09Z

modules/nf-core/lsa/cosine/tests/main.nf.test

+            assert process.success
+
+            // 4. Verificamos que se generen los archivos
+            assert process.out.matrix


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() },

pinin4fjords · 2026-02-12T17:51:52Z

modules/nf-core/lsa/cosine/tests/main.nf.test

+            assert process.out.versions
+        }
+
+    }


Also need a stub test (once you've added a stub to the main.nf)

pinin4fjords · 2026-02-12T17:52:22Z

modules/nf-core/cosimflow/main.nf

+        paste0("    r-pheatmap: ", packageVersion("pheatmap"))
+    ), "versions.yml")
+    """
+}


You'll also need a 'stub' - take a look at some other modules if you're not sure what I mean

pinin4fjords · 2026-02-12T18:01:53Z

modules/nf-core/cosimflow/main.nf

@@ -0,0 +1,114 @@
+process COSIMFLOW {


I think this is principally a wrapper around lsa::cosine, there isn't any tool called 'cosimflow' installed here, so by convention this module should really be LSA_COSINE, stored in a file at modules/nf-core/lsa/cosine

That fits with conventions here for e.g. deseq2/differential.

…shots

…odules into nf-core-cosimflow

miguelrosell · 2026-02-13T01:56:27Z

@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!

miguelrosell and others added 2 commits February 12, 2026 13:51

New module: cosimflow for cosine similarity

1f1e708

Merge branch 'master' into nf-core-cosimflow

006e0d1

pinin4fjords requested changes Feb 12, 2026

View reviewed changes

miguelrosell added 4 commits February 13, 2026 01:22

Refactor module: Rename to lsa/cosine, update deps, add stub and snap…

7c11956

…shots

Merge branch 'nf-core-cosimflow' of https://github.com/miguelrosell/m…

06144cd

…odules into nf-core-cosimflow

Fix: remove duplicate tests folder inside module directory

3294995

Fix: remove 'def' from script variables so that template can access them

a1dca67

miguelrosell changed the title ~~New module: cosimflow for cosine similarity~~ New module: lsa/cosine (calculate cosine similarity) Feb 13, 2026

Fix: removing Spanish comment

a52a2b6

New module: lsa/cosine (calculate cosine similarity) #10002

Are you sure you want to change the base?

New module: lsa/cosine (calculate cosine similarity) #10002

Conversation

miguelrosell commented Feb 12, 2026 • edited by ewels Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

pinin4fjords left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

pinin4fjords Feb 12, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

miguelrosell commented Feb 13, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

miguelrosell commented Feb 12, 2026 •

edited by ewels

Loading

pinin4fjords Feb 12, 2026 •

edited

Loading