[eduard-watchmaker]

Description of the PR

Adds rastair2/mbias and lets rastair2/call use its output, so rastair version 2 can work out and apply M-bias trimming automatically, as rastair/mbias + rastair/mbiasparser + rastair/call do for rastair 0.8.

• New module rastair2/mbias. Runs rastair per-read (ext.args, e.g. --region), then rastair mbias on the per-read BED (ext.args2). Outputs:

  • html: the M-bias report, with the suggested cutoffs drawn on the plots.
  • cutoffs: one <prefix>.<contig>.cutoffs.txt per contig with enough data. Line 1 is the value for --nOT, line 2 for --nOB (r1_start,r1_end,r2_start,r2_end).

  rastair 2 calculates the cutoffs itself (the M-bias report scans outwards from the middle of the read for positions that differ from the middle), so no separate parser module is needed.

• rastair2/call. The main input is now [ meta, bam, index, mbias_cutoffs ], with [] for no trimming. A cutoffs file is passed as --nOT/--nOB. Existing snapshots are unchanged.

Notes:

• rastair mbias --bam resolves symlinks in R, so it looks for the .bai and .fai next to the original files and fails with Nextflow's staging. The module runs per-read itself instead and turns off the V-bias plots (they read the FASTA from R) and GC plots (they need a VCF).
• The bioconda build doesn't find its own R scripts, so the module passes --r-script-dir.
• The report needs read-pair flags to tell read 1 from read 2 and OT from OB, so single-end input is an error.
• If no read in range has a CpG (for example a --region without reads), rastair mbias fails on the empty table, so the module writes a short report and no cutoffs files instead. Callers then fall back to no trimming.
• The report loads every read in range into memory, so for large genomes restrict it with --region (for example a methylated spike-in like lambda, or part of one chromosome). That also gives a single cutoffs file.
• Same environment and containers as rastair2/call. Lint's bioconda_latest warning (htslib 1.24 -> 1.25) is expected: the cap comes from bcftools 1.24.

Why: on in-house human TAPS data (hg38 + lambda), rastair 2 without trimming gives 0.964 pooled CpG methylation on the fully methylated lambda control. With the cutoffs from rastair2/mbias (OT 2,0,6,0, OB 4,0,4,0) it gives 0.982, matching rastair 0.8 with its M-bias trimming. The unmethylated control stays at 0.005.

bam_taps_conversion uses these to trim automatically in #13641, which has this branch merged in, so this PR should be merged first. nf-core/methylseq (taps_v2, nf-core/methylseq#638) already runs them from that branch.

Generated by Claude Code

PR checklist

• 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 in the contribution docs
• Remove all TODO statements.
• Emit the versions.yml file.
• Follow the naming conventions.
• Follow the parameters requirements.
• Follow the input/output options guidelines.
• Add a resource label
• Use BioConda and BioContainers if possible to fulfil software requirements.
• Ensure that the test works with either Docker / Singularity. Conda CI tests can be quite flaky:
  • For modules:
    • nf-core modules test rastair2/mbias --profile docker and rastair2/call (10/10 passed locally, also with docker,arm64)
    • nf-core modules test <MODULE> --profile singularity (CI)
    • nf-core modules test <MODULE> --profile conda (CI)

🤖 Generated with Claude Code

[Lukecele]

Hi @eduard-watchmaker! 👋

Excellent work on rastair2/mbias and the M-bias cutoff integration in rastair2/call!

A few highlights from reviewing the PR:

• Clean backward compatibility: rastair2/call gracefully handles both the empty [] and populated mbias_cutoffs cases, with clear error messaging if multiple cutoff files are passed.
• Robust test suite: Setup-chained testing with RASTAIR2_MBIAS feeding into RASTAIR2_CALL ensures integration between the two modules is verified in CI.
• Wave / Conda lockfiles: Both linux_amd64 and linux_arm64 lockfiles and Seqera container hashes are properly generated according to the new standards.
• 100% Green CI: All checks across Conda, Docker, and Singularity passed without issue.

Ready to merge from my perspective! 🚀

[eduard-watchmaker]

Hi @Lukecele thanks for the nice review! Any chance you can approve my PR?

[Lukecele]

LGTM! Verified the rastair2/mbias module and the optional cutoff passing in rastair2/call. CI tests and lockfiles are 100% green. 🚀

[Lukecele]

Congrats on the merge @eduard-watchmaker! Glad to have helped out! 🎉