MultiQC: add sourmash module (compare + gather)

[nekrut]

What

Adds a sourmash module to the IUC MultiQC wrapper, exposing MultiQC's native
sourmash module (present in MultiQC core since 1.x but not previously selectable
in this wrapper). Implemented as a new sourmash_plugin.xml following the
template_plugin.xml pattern, plus the usual five wiring points in multiqc.xml.

Two submodules are supported:

• compare — MultiQC's sourmash/compare module searches for *.labels.txt
  and then loads the matrix from the same path minus .labels.txt
  (numpy.load). The plugin therefore symlinks the matrix produced by
  sourmash compare -o NAME and its accompanying NAME.labels.txt into the
  software dir under a shared basename, so both are discovered and the
  clustered sample-similarity heatmap is rendered.
• gather — symlinks sourmash gather CSV(s) for the sourmash/gather
  submodule (content-matched header).

Why

sourmash compare produces an N×N genome/sample similarity matrix; surfacing its
clustered heatmap inside a MultiQC report is a natural QC summary for
multi-genome / multi-sample panels. Until now Galaxy users had to embed the PNG
via custom/image content; this wires up the real interactive MultiQC module.

Changes

• New tools/multiqc/sourmash_plugin.xml (@SOURMASH_COMMAND@ token + sourmash_form + sourmash_test).
• tools/multiqc/multiqc.xml: <import>, command #elif, software <option>, <when>, and <expand macro="sourmash_test"/>.
• tools/multiqc/macros.xml: @VERSION_SUFFIX@ 1 → 2.
• New test-data: sourmash_compare_matrix + sourmash_compare.labels.txt (real sourmash compare output over the existing-style tiny k=21 signatures) and sourmash_gather.csv.

Testing

• planemo lint tools/multiqc/multiqc.xml → passes (no errors/warnings).
• Verified locally with MultiQC 1.35: the staging layout makes the compare
  module find both matrix + labels and the gather module parse the CSV; the
  report contains sourmash-compare-heatmap and Sample similarity
  (the test's <assert_contents>).

🤖 Generated with Claude Code

[bgruening]

Why do we have repeat here with min=1?

[bgruening]

Is one enough but wyou can provide 2 different one? What happens if you provide two gathers?

[nekrut]

This <repeat name="output" min="1"> mirrors the established pattern used by the other MultiQC plugins in this tool (bismark, deeptools, fastqc, gatk, picard, rseqc, samtools, star, vcftools all use the identical <repeat name="output" ... min="1">). min="1" just requires that a sourmash result block contributes at least one output — otherwise nothing would be staged for that block.

The repeat is there precisely so you can provide more than one output and/or mix types: e.g. one compare plus one gather, or several gathers. Each repeat entry is uniquely indexed in the staged filename, so there are no collisions:

• compare: compare_<i>_<j> + compare_<i>_<j>.labels.txt
• gather: gather_<i>_<j>_<k>_<identifier>.csv

I verified two gather entries end-to-end: they stage as gather_0_0_0_*.csv and gather_0_1_0_*.csv, and MultiQC reports Found 2 gather results and renders both. So providing two gathers (or two compares, or any mix) works as expected.

Note also that within a single gather entry the input param already has multiple="true", so a user can either add multiple files to one entry or add multiple repeat entries — both are handled.

[bgruening]

Lgtm! I'm just wondering about the repeat, and what happens if a user just provides one, or multiple times the same. Should ir have a max=2?

[nekrut]

Thanks @bgruening! On the max question: I'd prefer to leave the repeat uncapped (no max), to stay consistent with the other MultiQC plugins here — none of bismark/deeptools/fastqc/gatk/picard/rseqc/samtools/star/vcftools put a max on their output repeat. A cap of 2 would be a bit arbitrary: a realistic use is, say, one compare plus several gather CSVs from different queries in the same MultiQC report, so an upper bound would get in the way.

On the edge cases you raised, I verified both locally:

• Just one output: works (that's the common case; min="1" guarantees at least one).
• The same output provided multiple times: also fine — each repeat entry is indexed in the staged filename (gather_<i>_<j>_<k>_<id>.csv, compare_<i>_<j>), so even providing the identical file twice stages as gather_0_0_0_* and gather_0_1_0_* with no collision. MultiQC reports Found 2 gather results and renders the report without error.

So there's no crash/collision risk that a max would be protecting against. Happy to add max if you feel strongly, but my preference is to keep it consistent with the sibling plugins.

[SaimMomin12]

Sourmash is WIP here: #7491

I will push it to get is merged in coming days