Review subworkflows: remove thin wrappers, migrate utility functions to plugins

Summary

We've audited all 113 subworkflows and categorized them by complexity. A number of subworkflows are thin wrappers around 1-2 tools and don't meet the bar of what a subworkflow should provide (3+ tools or reasonably complex channel operations/functions). Additionally, several utility subworkflows contain pure functions that should migrate to plugins.

This issue is for community discussion on which to remove, consolidate, or migrate.

---

Complexity breakdown

Category	Count
TRIVIAL (1 tool, no logic)	2
SIMPLE (1-2 tools, minimal logic)	27
MODERATE (3+ tools or complex ops)	52
COMPLEX (3+ tools + complex logic)	30
Empty/placeholder	2

MODERATE and COMPLEX subworkflows are considered justified and are not listed here.

---

TRIVIAL — Strong removal candidates

These are single-tool wrappers with no meaningful added logic:

• mafft_align — wrapper around MAFFT_ALIGN with empty optional inputs
• vcf_annotate_ensemblvep — wrapper around ENSEMBLVEP_VEP with a join for vcf+tbi output

---

Empty/placeholder directories — Remove

• multiple_impute_glimpse2 — README only, no main.nf
• preprocess_rnaseq — README only, no main.nf

---

SIMPLE — Review for removal or consolidation

Single tool + sort/stats subworkflow (repeated pattern)

These all follow the same pattern: one aligner call + BAM_SORT_STATS_SAMTOOLS. Could potentially be consolidated into a single parameterized subworkflow, or users can call the aligner module + sort/stats subworkflow directly.

• fastq_align_bowtie2 — Bowtie2 + BAM_SORT_STATS_SAMTOOLS
• fastq_align_bwa — BWA MEM + BAM_SORT_STATS_SAMTOOLS
• fastq_align_chromap — Chromap + BAM_SORT_STATS_SAMTOOLS
• fastq_align_hisat2 — HISAT2 + BAM_SORT_STATS_SAMTOOLS
• fastq_align_star — STAR + 2x BAM_SORT_STATS_SAMTOOLS (genome + transcriptome)

2 tools, linear pipeline (A → B)

These are simple sequential tool calls with basic channel wiring. Users could call the two modules directly.

• bam_cnv_wisecondorx — WISECONDORX_CONVERT + PREDICT
• bam_docounts_contamination_angsd — ANGSD doCounts + contamination
• bam_methyldackel — MethylDackel mbias + extract
• bam_stringtie_merge — StringTie assembly + merge
• bedgraph_bedclip_bedgraphtobigwig — bedClip + bedGraphToBigWig
• cache_download_ensemblvep_snpeff — download VEP + SnpEff caches
• vcf_annotate_snpeff — SnpEff annotate + tabix
• fastq_fetch_clean_hostile — fetch reference + clean with hostile
• fastq_extract_kraken_krakentools — Kraken extract reads
• fastq_download_prefetch_fasterqdump_sratools — prefetch + fasterq-dump
• h5ad_removebackground_barcodes_cellbender_anndata — CellBender + AnnData
• fastq_ngscheckmate — patterngenerator + vafncm
• fastq_index_filter_deacon — Deacon index + filter
• fasta_clean_fcs — FCSGX_RUNGX + FCSADAPTOR_CLEAN
• fasta_build_add_kraken2 — Kraken2 add + build
• bam_ngscheckmate — BCFtools mpileup + vafncm
• fastq_contam_seqtk_kraken — SeqTK subsample + Kraken2

Minimal logic wrappers

• bed_scatter_bedtools — 1 tool (BEDTOOLS_SPLIT) with map + transpose
• fastq_decontaminate_deacon_hostile — pure routing to 2 other subworkflows, no tools

Sort/dedup + stats combos

• bam_sort_stats_samtools — sort + index + BAM_STATS_SAMTOOLS
• bam_markduplicates_picard — MarkDuplicates + index + stats subworkflow
• bam_dedup_stats_samtools_umicollapse — UMI-collapse dedup + index + stats subworkflow
• quantify_pseudo_alignment — Salmon/Kallisto quant + tximport subworkflow

---

Utility subworkflows — Plugin migration candidates

These contain pure functions/logic with no tool invocations. Their functions should move to Nextflow plugins:

• utils_nextflow_pipeline — version check, params dump
• utils_nfcore_pipeline — 45+ custom utility functions
• utils_annotation_cache — cache path resolution logic
• utils_references — reference file resolution with custom functions

(utils_nfschema_plugin and utils_nfvalidation_plugin are already plugin-based.)

---

Guiding principles

• Subworkflows should have 3+ tools or reasonably complex channel operations/functions
• Pure functions should live in plugins, not subworkflows
• Simple linear A→B pipelines can be called as two modules directly in a workflow
• Repeated patterns (align + sort/stats) could be consolidated

Next steps

Community feedback on each item above. Please comment with 👍/👎 or discussion on specific subworkflows you'd like to keep/remove/consolidate.

Start by reviewing the listed subworkflow directories under subworkflows/nf-core, including each README.md and whether main.nf exists. Compare the wrappers, linear pipelines, and utility functions against the stated principles, then comment with a specific keep, remove, consolidate, or plugin-migration recommendation for each item; done means every candidate has a community-backed disposition.