🐲 nf-core/rnaseq 3.25.0 (Plutonium Pangolin) is out! Chunky release this time — highlights: 🧬 New MultiQC "Strandedness checks" section A rebuilt MultiQC section where table rows reflect which strandedness analyses actually ran for each sample, with a stacked composition bargraph and a dataset switcher when both RSeQC and Salmon inferred. Inference methodology is clearer and issues like "sample failed infer_experiment" are now visible rather than silently dropped. 🛠️ StringTie restructured as a proper three-stage de-novo workflow

--stringtie_ignore_gtf

now runs through the nf-core

bam_stringtie_merge

subworkflow as assemble → merge → quantify, publishing

stringtie_merge.gtf

and per-sample

<sample>.denovo.transcripts.gtf

. Much more useful for de-novo transcript discovery than the previous single-shot setup. 🚀 GPU-accelerated rRNA removal New

--use_gpu_ribodetector

parameter enables GPU-accelerated rRNA removal with ribodetector. The GPU CI skip was generalized from

SKIP_PARABRICKS

to

SKIP_GPU

so all GPU tooling shares one gate. 🦠 Prokaryotic profile improvements - Skip StringTie by default (reference-guided assembly isn't informative for bacterial/archaeal annotations) - Derive the gene BED via

gffread --bed

so RSeQC

infer_experiment

gets a usable reference (

ea-utils/gtf2bed

produced an empty BED from CDS-only annotations) - Drop RSeQC

infer_experiment

for

bowtie2_salmon

(transcriptome-aligned BAMs can't be inferred against a genomic BED; Salmon's own inference is surfaced instead) - Bowtie2

-k

default raised 1 → 200 for

bowtie2_salmon

, so Salmon's EM has enough multi-mapping evidence to quantify small transcriptomes correctly 📊 SortMeRNA default DB bumped to SILVA 138 Updated to

smr_v4.3_default_db

— addresses #1354, bringing the shipped rRNA DB up to a much more current reference. ⚠️ Breaking: per-strand bigWigs no longer emitted for unstranded libraries A strand-agnostic

<sample>.bigWig

is always emitted, but for unstranded libraries a

-strand +/-

split carries no biological meaning and has been dropped (#1275). 🔧 Other notable changes: - Fix per-sample MultiQC hanging under

--skip_quantification_merge

— each sample's report now fires as soon as its own contributors arrive -

custom/multiqccustombiotype

fails loudly when featureCounts exceeds

--max_biotypes

(default 100), catching misconfigured

--featurecounts_group_type

values that used to hang MultiQC - Scope MultiQC's

table_sample_merge

to samplesheet paired-end IDs so samples like `foo_1`/`foo_2` are no longer collapsed into a single

foo

row in General Statistics - Centralize pipeline-specific module configs in

conf/modules/

(sarek-style layout) - Skip sentieon tests on the conda profile (license-server-driven output drifts across conda solves) - Sync all nf-core components to latest; retire

ch_versions

plumbing (everything now emits via the

versions

topic) 🙏 Thanks to everyone who contributed — bug reports, feature requests, reviews and code all very appreciated. 📋 Full changelog and release: https://github.com/nf-core/rnaseq/releases/tag/3.25.0 🐛 Bugs and feature requests to https://github.com/nf-core/rnaseq/issues!