New module: modkit/entropy

[sahuno]

PR checklist

• This comment contains a description of changes (with reason).
• Tests added (stub + real, both --cpg mode).
• New tool follows module conventions.
• Versions broadcast via topic: versions.
• Naming/parameter/I-O conventions followed.
• Resource label set (process_high).
• BioConda + BioContainers used.
• `nf-core modules lint modkit/entropy` — 50/0/0.
• `nf-test test --profile conda` — 2/2 passed.

Summary

Adds a new nf-core module wrapping `modkit entropy`, which computes methylation entropy over genomic windows from one or more mod-BAMs. Entropy is a per-window measure of the diversity of methylation patterns across reads covering the window and is complementary to mean methylation.

The module supports modkit's file-vs-directory output dichotomy: when `--regions` is supplied (either via `ext.args` or the third input channel), modkit writes per-region BED / bedgraph / TSV into a directory (emitted via `regions_bed`, `bedgraph`, `tsv` outputs); otherwise a single genome-wide BED is produced (emitted via `bed`).

Accepts multiple input BAMs (passed as repeated `--in-bam`) for multi-sample entropy.

Why

`modkit entropy` is the canonical tool for quantifying methylation heterogeneity in ONT methylation data and is not currently in nf-core/modules. It complements the existing `modkit/pileup` (mean methylation) for heterogeneity-aware analyses.

Test data

Uses the existing `test.sorted.phased.bam` from nf-core/test-datasets (modules branch, `genomics/homo_sapiens/nanopore/bam/`). No new test data required.

🤖 Generated with Claude Code

[itrujnara]

Hi, thanks for the high-quality submission. I have left a handful of comments.

[github-actions]

This PR has been tagged as awaiting-changes or awaiting-feedback by an @nf-core/modules contributor. Remove stale label or add a comment if it is still useful.

[sahuno]

Thanks for the review @itrujnara — all three points addressed.

1. Stub now simulates the --regions output. modkit entropy --regions
writes a directory containing <prefix>_regions.bed and
<prefix>_windows.bedgraph instead of the single genome-wide BED, so the stub
branches on whether regions is present and creates the matching layout.

While confirming the filenames against modkit 0.6.1 I found the tsv output
channel never fires — modkit entropy --regions emits only the _regions.bed
and _windows.bedgraph files, no TSV. Removed that emit and corrected the
meta.yml description, which claimed a .tsv was produced.

2. Added tests that use the regions input — one real and one stub. The
real one asserts the exact output filenames, that bed is not emitted when
regions is passed, and that the region_name column echoes the interval
from the input BED.

3. Yes — now checking file contents. modkit entropy has a --header
flag, so the test config passes --cpg --header and the tests assert on the
header line and the row structure:

• genome-wide BED and per-window bedgraph:
  #chrom start end entropy strand num_reads
• per-region BED: chrom start end region_name mean_entropy ...
• every data row has all six columns and a parseable numeric entropy value

Still snapshotting only the versions topic for the non-stub tests: the entropy
column is not byte-deterministic across CPU counts (I get 0.69360054 vs
0.6936004 for the same window between runs with different thread counts), so
an md5 snapshot would be flaky in CI.