[eduard-watchmaker]

This PR adds rastair version 2 as an optional caller in bam_taps_conversion, for nf-core/methylseq's TAPS path, with automatic M-bias trimming. rastair 0.8 stays the default.

Depends on #13678 (new rastair2/mbias; rastair2/call takes its cutoffs). That branch is merged in here, so this diff shows its files until #13678 is merged; please merge #13678 first. nf-core/methylseq (taps_v2, nf-core/methylseq#638) already uses this subworkflow from this branch.

• New input use_rastair2 (boolean).
  • false runs the existing rastair 0.8 chain unchanged: rastair/mbias → rastair/mbiasparser → rastair/call → rastair/methylkit. The existing test's outputs are byte-identical; its snapshot only gains the new, empty channels.
  • true runs rastair version 2:
    • rastair2/mbias on paired-end samples, to find how many bases to ignore at each read end. Its report needs read-pair flags, so single-end samples are called without trimming.
    • rastair2/call with each sample's cutoffs as --nOT/--nOB (methylation/SNV VCF plus a CpG BED).
    • bcftools/stats on the VCF for MultiQC.
• New input trims (list). [] uses the M-bias estimate (rastair 0.8: rastair/mbias + rastair/mbiasparser; rastair version 2: rastair2/mbias). [ nOT, nOB ] (each r1_start,r1_end,r2_start,r2_end) uses those bases to ignore for both callers and skips the M-bias steps, so pipelines can expose the trims as parameters. Invalid values are an error. If the estimate can't be made (single-end samples for rastair2, or no reads with CpGs in range), nothing is trimmed.
• Choosing the contig for M-bias: pipelines should restrict RASTAIR2_MBIAS with --region in ext.args (for example a methylated spike-in like lambda), which also keeps memory down on large genomes. If mbias reports several contigs, the subworkflow uses the longest one.
• Why trim: on in-house human TAPS data, rastair version 2 gives 0.964 pooled CpG methylation on the fully methylated lambda control without trimming and 0.982 with the mbias cutoffs, matching rastair 0.8 with its M-bias trimming.
• New outputs: vcf, vcf_tbi, bed, bed_tbi, bcftools_stats, mbias_html and mbias_cutoffs. Whichever caller isn't run leaves its outputs empty.
• Shared reference: with rastair2, all samples share one reference, whether the pipeline passes it once (value channel) or once per sample (as methylseq does).
• Interface change: these are new fifth and sixth inputs. No other component in this repo calls bam_taps_conversion; nf-core/methylseq will pass false or its --taps_caller choice.
• Licence: rastair version 2 is free for non-commercial use only (University of Oxford). The meta.yml says so; pipelines should gate it behind an explicit opt-in.

Tests: the existing rastair 0.8 test; rastair2 with two samples and one reference as a value channel (checks the cutoffs reach call); a paired-end plus a single-end sample (only the paired-end one is trimmed); manual trims for each caller (no M-bias steps, trims reach call); invalid trims fails; and a rastair2 stub with a per-sample reference.

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!
• Follow the naming conventions.
• Follow the input/output options guidelines.
• Ensure that the test works with either Docker / Singularity. Conda CI tests can be quite flaky:
  • For subworkflows:
    • nf-core subworkflows test bam_taps_conversion --profile docker (7 tests)
    • nf-core subworkflows test bam_taps_conversion --profile singularity (CI)
    • nf-core subworkflows test bam_taps_conversion --profile conda (CI)

Lint: 0 failures, 0 warnings.

🤖 Generated with Claude Code

[adamrtalbot]

Almost there, these are fairly minor nits, tbh:

• Blocking: use lowercase channel and named closure parameters in the new tests.
• Not blocking: put a bare assert workflow.success before assertAll(...), and write grouped inputs in test names as [ bam, bai ].
• The subworkflow logic looks good. combine(...).first() handles both reference styles, and the rastair 0.8 path is unchanged.

While you are here (not blocking):

• tests/main.nf.test:28-40 also uses Channel.

[github-actions]

This PR exceeds the recommended size of 1000 lines. Please make sure you are NOT addressing multiple issues with one PR. Note this PR might be rejected due to its size.

[eduard-watchmaker]

Thanks! All points applied: bare assert workflow.success before assertAll, and [ bam, bai ] in the test names. This update also adds automatic M-bias trimming for rastair2, using the new rastair2/mbias from nf-core/modules#13678 · rastair2/mbias: new module….

[Lukecele]

Hi @eduard-watchmaker! 👋

Just reviewed the updated test suite following @adamrtalbot's feedback:

• ✅ All channel declarations in tests/main.nf.test are now lowercase channel.of and channel.value.
• ✅ Named closure parameters (meta, vcf -> ...) are properly used throughout.
• ✅ All 35 CI matrix checks across Conda, Docker, and Singularity are completely green.

Both the subworkflow logic and the accompanying test suite look clean, robust, and ready for merge! 🚀

[eduard-watchmaker]

Hi @Lukecele thanks for the review in here too, any chance you can approve the PR so I can merge?

[Lukecele]

LGTM! Verified that all feedback points from @adamrtalbot are addressed (lowercase channel.of, clean assertions, descriptive test names). Tests and snapshots are 100% green. 🚀

[eduard-watchmaker]

@adamrtalbot can you approve too? Thanks!

[Lukecele]

Yes soon give me free minutes free times Il mer 11 nov 2026, 17:30 Eduard Casas ***@***.***> ha scritto:

…

@eduard-watchmaker <https://github.com/eduard-watchmaker> requested your review on: nf-core/modules#13641 <#13641> bam_taps_conversion: optional rastair version 2 caller (use_rastair2). — Reply to this email directly, view it on GitHub <#13641?email_source=notifications&email_token=AU3A7UV6FNYSWO5IWHLFGD35YSH2LA5CNFSNUABQM5UWIORPF5TWS5BNNB2WEL2JONZXKZKFOZSW45CON52GSZTJMNQXI2LPNYXTGNJYGE4TQMJWG42TTJTSMVQXG33OWBZGK5TJMV3V64TFOF2WK43UMVSKKZLWMVXHJLDGN5XXIZLSL5RWY2LDNM#event-35819816759>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AU3A7UQ7CCUXYGWPULXRC735YSH2LAVCNFSNUABFKJSXA33TNF2G64TZHMYTSOBZHAZDAOBTHNEXG43VMU5TMMRZGA3TQNBVG43KC5QC> . Triage notifications, keep track of coding agent tasks and review pull requests on the go with GitHub Mobile for iOS <https://github.com/notifications/mobile/ios/AU3A7USYTLBDA3ITV5HAWD35YSH2LA5CNFSNUABQM5UWIORPF5TWS5BNNB2WEL2JONZXKZKFOZSW45CON52GSZTJMNQXI2LPNYXTGNJYGE4TQMJWG42TTJTSMVQXG33OWBZGK5TJMV3V64TFOF2WK43UMVSKKZLWMVXHJKTGN5XXIZLSL5UW64Y> and Android <https://github.com/notifications/mobile/android/AU3A7USHZ5IZSYTVASO4VMD5YSH2LA5CNFSNUABQM5UWIORPF5TWS5BNNB2WEL2JONZXKZKFOZSW45CON52GSZTJMNQXI2LPNYXTGNJYGE4TQMJWG42TTJTSMVQXG33OWBZGK5TJMV3V64TFOF2WK43UMVSKKZLWMVXHJLTGN5XXIZLSL5QW4ZDSN5UWI>. Download it today! You are receiving this because your review was requested.Message ID: ***@***.***>