Trim Galore v2.1.0-beta.1 — Oxidized Edition, now in open beta

[FelixKrueger]

Trim Galore v2.1.0-beta.1 — Oxidized Edition, now in open beta

This is the first public beta of the full Rust rewrite of Trim Galore. It is a drop-in replacement for the Perl v0.6.x tool: same CLI, same output files, same defaults. Testing feedback from real pipelines is the whole point of the beta window — please give it a run and open issues for anything surprising.

What's new in v2.1.0 (vs. Perl v0.6.x)

• Rust implementation. Single static binary, no Perl/Cutadapt plumbing for the wrapper itself — Cutadapt is still invoked for the actual adapter-removal step.
• Multi-adapter support. Repeat -a/-a2 flags or use -a file:adapters.fa / -a2 file:adapters.fa. Multiple adapters are applied in sequence (rounds controlled by -n / --times).
• --discard-untrimmed to keep only reads where an adapter was found.
• Multi-file single-end input — pass several FASTQs in one command.
• JSON trimming report alongside the existing text report, consumable natively by MultiQC.
• MultiQC-compatible Cutadapt section in the text trimming report, matching the output MultiQC has parsed from Cutadapt for years.

Install

Cargo (crates.io)

cargo install trim-galore --version 2.1.0-beta.1

The --version flag is required during the beta window. Plain cargo install trim-galore still resolves to 2.0.0 — cargo excludes prereleases by default, so it won't pick up 2.1.0-beta.1 unless you ask for it explicitly.

Docker / OCI (multi-arch: linux/amd64 + linux/arm64)

docker pull ghcr.io/felixkrueger/trimgalore:v2.1.0-beta.1
# or the moving "latest beta" tag:
docker pull ghcr.io/felixkrueger/trimgalore:beta
# or pin by immutable digest (deterministic across beta iterations):
docker pull ghcr.io/felixkrueger/trimgalore@sha256:3e5dd519ea992ce88eb6f01e8d051a33f356e9ecd2d45e79a9c9bd0ac05c5b77

:latest, :2, and :2.1 are deliberately not moved during the beta — those tags stay pinned to the previous Perl release until v2.1.0 GA ships.

Drop-in compatibility

The CLI is bit-for-bit compatible with Perl v0.6.x — your existing wrapper scripts and pipeline configs should not need changes. Outputs match in file names, format, and content. If you find a real regression, that's a blocker and worth a high-priority issue.

nf-core/rnaseq integration

The integration PR is up and fully green on the complete CI matrix (x86 + arm64 + GPU, both 25.04.3 and latest-everything): nf-core/rnaseq#1789. The PR pins to the image digest above so re-runs are deterministic. No MD5 drift on reference outputs; MultiQC parses both the JSON and the Cutadapt-section text reports cleanly.

If you run other nf-core pipelines that use Trim Galore, pinning your local module copy to the same digest is the recommended way to test this beta without waiting for upstream nf-core/modules to update.

A note on v2.0.0

v2.0.0 was an unintentional early publish to crates.io on 2026-04-01 before the staged rollout was ready. Several user-facing features listed above landed on optimus_prime after that publish, so the release retargets at v2.1.0. v2.0.0 will be yanked from crates.io when v2.1.0 GA ships. Users should install v2.1.0 or later once available; installing v2.0.0 is not recommended.

In the unlikely event v2.0.0 is yanked before GA (only if a correctness bug is discovered in the stable release), cargo install trim-galore with no version will fail — cargo excludes prereleases by default and no non-yanked stable would remain. The beta stays installable with the explicit --version 2.1.0-beta.N flag; watch the releases page or this announcement thread for a hotfix stable if that happens.

How to give feedback

• Bugs or unexpected behavior: open an issue at https://github.com/FelixKrueger/TrimGalore/issues with the beta-feedback label.
• Questions, "does this replace X in my pipeline?", general discussion: reply in this thread or start a thread in the Q&A category.
• Performance comparisons or "I ran the beta on N samples" reports: Show and tell category — very welcome, useful for others deciding whether to adopt.

What's most valuable: running the beta on real data you already have Perl v0.6.x results for, and reporting any divergences — either in output content or in the trimming report fields.

Timeline

• Beta window: minimum 2 weeks from 2026-04-19. Additional -beta.N cuts will iterate freely if material issues surface.
• GA target: on or after 2026-05-03, conditional on no critical beta issues outstanding.
• GA-tracking PR: Trim Galore v2.1.0 (Oxidized Edition) — Rust rewrite #214 (draft, labeled do-not-merge-until-GA).

Platform support

• linux/amd64 ✓
• linux/arm64 ✓
• Windows: not supported in this release; tracked separately.

---

Thank you for testing. Every bug report shortens the path to a confident v2.1.0 GA.

[FelixKrueger]

Update: Beta 1 issues fixed on optimus_prime — please rebuild

Thanks to everyone who kicked the tyres on Beta 1! Three regressions and a handful of improvements have landed on optimus_prime since 2026-04-18. If you've been running Beta 1, please rebuild to pick these up — a formal Beta 2 release will follow later; the branch is the fastest path to the fixes right now.

Bug fixes

• --paired with >2 files was rejected. trim_galore --paired *fastq.gz (common shell-glob usage) now works again — any even number of input files is processed as consecutive R1/R2 pairs, restoring the v0.6.x Perl behaviour. Thanks to the beta tester who reported this on day one.
• Output-path collision pre-flight is now case-insensitive (ASCII) — catches filenames differing only in letter-case on APFS (macOS default) and NTFS (Windows default) as well as ext4 (Multi-pair collision pre-flight misses case-only aliases on macOS/Windows #216).
• Adapter auto-detection runs per pair, not once per invocation. Safer for shell-glob invocations that mix library types or 2-colour/4-colour chemistries across samples.

New

• trim_galore --version now prints a build-provenance line (<git-hash> — <target> — built <timestamp>). Handy for bug reports. -V stays terse — one line, unchanged.
• Reproducible builds: SOURCE_DATE_EPOCH=<unix-ts> cargo build --release produces a bit-identical binary across runs.

How to rebuild

From source (git clone):

git pull origin optimus_prime
cargo build --release

Via cargo install:

cargo install --git https://github.com/FelixKrueger/TrimGalore \
  --branch optimus_prime --force trim-galore

Docker (GHCR): the :beta tag still points at Beta 1. A refreshed tag will follow with Beta 2. In the meantime, build locally:

git clone -b optimus_prime https://github.com/FelixKrueger/TrimGalore
cd TrimGalore && docker build -t trim_galore:optimus_prime .

After rebuilding, trim_galore --version should report version 2.1.0-beta.2 on line 1 and a commit hash dated 2026-04-20 or later on line 2. Please keep the feedback coming in this thread or open issues with the beta-feedback label.

[FelixKrueger]

Update — v2.1.0-beta.2 is out

Tagged 2026-04-20 · Release notes

Headline fix from beta-tester feedback: trim_galore --paired *fastq.gz works again — the beta.1 hard-check that rejected anything other than exactly two input files has been replaced with the v0.6.x semantics (N consecutive R1/R2 pairs). Adapter auto-detection and poly-G scanning now run per pair rather than once per invocation, so mixed libraries or 2-colour / 4-colour chemistries across a glob are handled correctly.

Also in this beta:

• Paired-end output-path collision pre-flight (case-insensitive, catches APFS/NTFS overwrites) — fixes Multi-pair collision pre-flight misses case-only aliases on macOS/Windows #216.
• Byte-equal R1==R2 detection with a precise error message.
• --version now prints build provenance (<git-hash> — <target> — built <ISO-8601 UTC timestamp>) on a second line.
• Reproducible builds via SOURCE_DATE_EPOCH.

Install

cargo install trim-galore --version 2.1.0-beta.2
# or
docker pull ghcr.io/felixkrueger/trimgalore:v2.1.0-beta.2

Pre-built binaries for linux-x86_64, linux-aarch64, and macos-aarch64 are attached to the release.

Keep the feedback coming — earliest GA remains 2026-05-03, so there's time for another bug-fix bump if something important surfaces.

[FelixKrueger]

Update — v2.1.0-beta.3 is out

Tagged 2026-04-24 · Release notes

Third beta, folding in three user-facing PRs since beta.2:

New capabilities

• BGI/DNBSEQ added to adapter auto-detection — no more explicit --bgiseq needed for BGI/MGI/DNBSEQ data. Stranded Illumina stays explicit-only (its sequence differs from Nextera by a single A-tail base and would produce constant ambiguous ties if probed).
• Repeatable -a/-a2 multi-adapter syntax — -a SEQ1 -a SEQ2 now works directly, alongside the v0.6.x embedded-string and file:adapters.fa forms.
• Perl-style A{N} expansion for -a/-a2 (e.g. -a A{10} → -a AAAAAAAAAA).
• Perl-era -r1/-r2/-a2 short-flag forms accepted via transparent pre-parse rewriting — migrating v0.6.x scripts no longer trip on clap's strict single-char short-flag rule.

Bug fixes

• --trim-n is now suppressed under --rrbs, restoring Perl v0.6.x byte-identical output for that flag combo.
• Paired-end report line fix — the parameter summary no longer emits a stray -end suffix (...gets removed-end: 20 bp → ...gets removed: 20 bp).

Framing change

The rewrite's positioning shifted from "byte-identical output to v0.6.x" to "faithful rewrite designed as a drop-in replacement, with useful additions". The old framing no longer held given --poly_g, generic --poly_a, per-pair adapter detection, now BGI auto-detect, and other post-Perl additions. Updated consistently in the README, SUMMARY, User Guide, CHANGELOG, and --help preamble.

Install

cargo install trim-galore --version 2.1.0-beta.3
# or
docker pull ghcr.io/felixkrueger/trimgalore:v2.1.0-beta.3

Multi-arch digest: sha256:3d8e9c90555cb47176ea4f7527230c59995f402af6a32d3ef7b3428ca5d6b337.

Pre-built binaries for linux-x86_64, linux-aarch64, and macos-aarch64 are attached to the release.

Test count went 106 → 147 across the beta.1 → beta.3 arc. Earliest GA remains 2026-05-03. Feedback welcome — beta-feedback label on issues, or reply in this thread.

[FelixKrueger]

Update — v2.1.0-beta.4 is out

Tagged 2026-04-26 · Release notes

Fourth beta. The headline is that the "single static binary, zero external runtime deps" story is now literally true at the artifact level — no Cutadapt, Java, Python, or FastQC tarball needed at runtime.

New capabilities

• Bundled FastQC via the fastqc-rust library — --fastqc works out of the box without Java or an external fastqc binary. Output files (*_fastqc.html, *_fastqc.zip) are FastQC 0.12.1-compatible, so MultiQC parsers and downstream pipelines see identical structure. The Docker image is correspondingly slimmer (~350 MB lighter at the runtime layer; no JRE, no Perl, no FastQC tarball). --fastqc_args continues to translate the common flags (--nogroup, --expgroup, --quiet, --svg, --nano, --nofilter, --casava, -t/--threads, -o/--outdir).

Bug fixes

• --clock and --implicon accept multi-pair input again — the two specialty run-and-exit modes had been narrowed to "exactly 2 input files" by the v2.x rewrite. Now restored to v0.6.x semantics: an even number of files is treated as consecutive R1/R2 pairs, with per-pair headers (=== Clock pair N of M === / === IMPLICON pair N of M ===) and the same case-insensitive output-collision pre-flight that --paired runs. Same shape as the --paired widening that landed in beta.2.

Infrastructure

• .gitattributes linguist patterns (drive-by from @ewels, thanks!) — GitHub's repo language bar now reflects the actual Rust content rather than HTML in docs/.
• Rust toolchain bumped 1.85 → 1.88 (required by fastqc-rust).

Install

cargo install trim-galore --version 2.1.0-beta.4
# or
docker pull ghcr.io/felixkrueger/trimgalore:v2.1.0-beta.4

Multi-arch digest: sha256:9a7f7ddedde9df5400a1819fd22fd0fae20c8c9f41bbcd5ac7ea382178c794f0.

Pre-built binaries for linux-x86_64, linux-aarch64, and macos-aarch64 are attached to the release.

Test count went 147 → 161 across beta.3 → beta.4 (multi-pair specialty validation + fastqc-args translator). GA target is 2026-05-04 ("May the 4th be with you"). Feedback welcome — beta-feedback label on issues, or reply in this thread.

[FelixKrueger]

Update — v2.1.0-beta.5 is out

Tagged 2026-04-26 · Release notes

A point-release with a single change: bumps the bundled fastqc-rust dep from v1.0.0 → v1.0.1 to pick up an upstream precision fix in the >>Overrepresented sequences section.

Background. v1.0.0 was rounding the percentage column to 2 decimals (7.16) instead of emitting Java FastQC 0.12.1's full Double.toString() precision (7.160449112640348). All other sections were already byte-identical to Java FastQC. The divergence surfaced via the nf-core/rnaseq integration matrix on Nextflow 25.04.3 (the older pinned MultiQC there preserves the literal percentage when aggregating, so the rounding cascaded into a downstream MD5 mismatch). Detected sequences, counts, and source classification were always correct — purely a cosmetic formatting deviation.

Fix filed and merged upstream as ewels/FastQC-Rust#2; v1.0.1 published the same day.

No behavioural change to TrimGalore itself. If you've been running beta.4 and don't use --fastqc (or don't have overrepresented sequences in your data), there's nothing user-visible. If you do, the bundled FastQC's fastqc_data.txt is now byte-equivalent to Java FastQC 0.12.1 again.

Install

cargo install trim-galore --version 2.1.0-beta.5
# or
docker pull ghcr.io/felixkrueger/trimgalore:v2.1.0-beta.5

Multi-arch digest: sha256:5a82df1902f33da09c6fef41d6c6b2eba74f36db3bf8c2650c34906a01e33ccd.

GA target remains 2026-05-04 ("May the 4th be with you"). Feedback welcome — beta-feedback label on issues, or reply in this thread.