Available Pipelines

The installed catalog is built from whatever is found under pipelines/ in the application directory — or wherever SEQDESK_PIPELINES_DIR points, which matters on clusters whose compute nodes do not share the app’s filesystem. Each subdirectory holding a valid manifest.json becomes one catalog entry; directories whose names start with . or _ are skipped, which is why the pipelines/_example/ scaffold never appears.

Eleven packages ship with SeqDesk today:

Two further visibility flags come from each package’s registry.json: visibility.showToUser decides whether researchers see the pipeline’s status at all, and visibility.userCanStart decides whether they may launch it. Every shipped package sets userCanStart: false — launching is always a facility admin action. Independently of both flags, a run’s results stay private until an admin marks it Use as final; see Publishing a run.

Study pipelines

MAG Pipeline

Wraps: nf-core/mag v3.0.0 · Target: study · Requires: paired-end reads on every selected sample

The reference metagenomics workflow: quality control, host removal, assembly, binning, bin refinement, bin QC and taxonomic classification, ending in a MultiQC report. It is also the pipeline that produces the Assembly and Bin records the ENA submission path depends on.

MAG is the only shipped package with a hard compatibility gate. Its samplesheet requires paired short reads, and metadata validation refuses an order whose sequencing technology is long-read or single-end-only:

Sequencing technology "…" incompatible with MAG.
MAG requires short-read technologies that support paired reads.

Steps (from definition.json, in dependency order): Input Validation, Raw Read QC, Read Trimming, Host Removal, Trimmed Read QC, Assembly, Assembly QC, Binning Prep, Genome Binning, Bin Refinement, Bin Quality, Taxonomy, Annotation, MultiQC Report.

Configuration

Two conditional rules apply on top of the flag mapping: enabling skipBinQc or skipBusco also adds --skip_quast --skip_gtdbtk, because those stages depend on bin QC.

Narrowing with runAt: selected-technologies produces its own launch errors:

MAG is restricted to selected sequencing technologies. Found disallowed technology IDs: … Allowed: …
MAG is restricted to selected sequencing technologies in pipeline settings.
Some selected samples are missing order/technology selection metadata.

Outputs

Bin completeness and contamination are parsed from GenomeBinning/QC/checkm_summary.tsv and joined on the bin name. If that file is absent the bins are still created — with empty quality columns.

Submit to ENA

Wraps: submg  v1.0.0 via a custom runner · Target: study · Category: submission (appears under Publishing)

Builds the submg manifests from SeqDesk’s own records, submits reads, assemblies and — optionally — bins to the European Nucleotide Archive, then parses the receipts and writes the accessions back onto the sample, read and assembly records.

Steps: Validate Inputs → Generate Config → Submit to ENA → Parse Accessions.

Prerequisites. This is the pipeline with the longest list of preconditions, and all of them are checked before a run is created:

• the study has an ENA study accession (PRJ…), registered on the same server the submission targets;
• on the ENA test server, that registration must be less than 24 hours old;
• every sample has a taxId and the checklist fields submg requires;
• every sample has paired reads with MD5 checksums — run FASTQ Checksum first if they are missing;
• every sample resolves to an assembly file.

Configuration

Outputs. Two run artifacts — the submission logs (logging*/**/*) and submg-metadata.json. The real result is in the database: sample accessions (ERS/SAMEA), run and experiment accessions (ERR/ERX) and assembly accessions (ERZ), rendered as an ENA submission results table on the pipeline page with links into the ENA browser.

See ENA Submission for the full flow.

Quality Overview

Package: reads-qc v0.1.0 · Target: study · Runtime: seqkit 2.8.0 plus Python via conda; runs locally on Apple Silicon

Per-sample FASTQ statistics rolled up into a study-level HTML overview: read count, total bases, average quality and GC content. It is the lightweight alternative to running FastQC on every sample when all you want is a comparison across a study.

Steps: Compute statistics → Collect statistics → Generate report.

Configuration: none.

Outputs

Note the writeback: although Quality Overview is study-scoped, it updates the same canonical Read fields that the order-side QC pipelines write. Scope governs which target a run belongs to, not whether reads may be touched.

Study MultiQC

Package: multiqc v0.1.1 · Wraps: bioconda::multiqc=1.21 · Target: study

Runs a single MultiQC pass over verified QC artifacts produced by earlier runs in the same study — FastQC archives and NanoPlot NanoStats tables — and produces one consolidated report. It re-analyses nothing; it only aggregates.

The report is written as study-multiqc.html rather than MultiQC’s default multiqc_report.html, so it can never collide with the report MAG emits in the same study.

Configuration

Outputs

Study Demo Report

Package: study-demo-report v0.1.0 · Target: study · Runtime: none — a single awk-based process with no conda profile at all

A deterministic, hermetic pipeline that produces an HTML report, a Markdown summary and a per-sample TSV without doing any bioinformatics. Its purpose is to prove that execution, monitoring, output discovery, preview handling and the results UI all work end to end on a given installation.

This is the right first run on a new install. It needs no reference data, no reads and no network, and it finishes in seconds — so if it fails, the problem is the environment, not the data.

Configuration

Outputs: report/demo-report.html (study report), report/demo-report.md and tables/sample-summary.tsv (run artifacts).

Taxonomic Profiling (Kraken2 + Bracken)

Package: kraken2-bracken v0.1.0 · Target: study and order · Wraps: kraken2 2.1.3, bracken 2.9, krakentools 1.2 + krona 2.8.1

Classifies each sample’s short reads against a Kraken2 database, re-estimates abundances with Bracken at a configurable rank, and renders an interactive Krona chart per sample. The top taxon and its fraction are parsed out of each Bracken table and shown in the sample result column.

This is the only shipped package that declares both targets, so it appears in both the study and the sequencing-order catalogs.

Steps: Classify Reads (Kraken2) → Estimate Abundance (Bracken) → Visualize (Krona) → Collect Summary.

Configuration

Outputs

No Read writeback: everything stays as artifacts.

Sequencing Order pipelines

Simulate Reads

Package: simulate-reads v0.2.0 · Target: order

Generates FASTQ files for the samples of a sequencing order and links them to the canonical Read records. It is the one order pipeline that does not require existing reads — it produces them. Use it for demos, smoke tests, and to give the rest of the order-side chain something to work on.

It can either synthesise reads or replay real template FASTQ pairs that the facility has staged, which produces far more realistic input for downstream QC.

Configuration

Writes back: file1, file2, checksum1, checksum2, readCount1, readCount2. Also produces: summary/simulation-summary.tsv as a run artifact.

FASTQ Checksum

Package: fastq-checksum v0.1.0 · Target: order · Requires: linked reads

Computes MD5 checksums for the FASTQ files already linked to the order’s samples and stores them on the canonical read record. ENA submission validation refuses to proceed without them, so this is worth running as soon as data arrives.

Configuration: none.

Writes back: checksum1, checksum2 (from checksums/*.json). Also produces: summary/checksum-summary.tsv.

The sample result column shows the first 8 characters of each checksum, or Not computed.

FastQC

Package: fastqc v0.1.0 · Wraps: bioconda::fastqc=0.12.1 · Target: order · Requires: linked reads

Runs FastQC per sample, publishes the HTML reports and zip archives, and extracts read counts and mean quality back onto the read record.

Configuration: none.

Outputs

Writes back: fastqcReport1, fastqcReport2, readCount1, readCount2, avgQuality1, avgQuality2. The sample result column shows the report links and the average quality per read direction.

Read Cleaning

Package: read-cleaning v0.1.0 · Wraps: nf-core/detaxizer v1.3.0 · Target: order · Requires: active reads whose dataClass is raw or unknown

Screens reads for host or contaminant sequences — by default Homo sapiens — filters them out, and writes per-sample cleaned FASTQ files. Those files are staged as candidates, not applied: they become run artifacts and wait for an admin to review the evidence and promote them. See Read writeback and candidate review.

Samples whose active reads are already cleaned are reported as Needs raw or unknown reads and skipped, so the pipeline cannot be run twice over its own output by accident.

Configuration

Outputs

NanoPlot (Long-read QC)

Package: nanoplot v0.1.0 · Wraps: bioconda::nanoplot=1.42.0 · Target: order · Requires: a linked read file per sample

Computes long-read quality metrics — read count, total bases, mean/median/N50 read length, mean Phred quality — and an interactive self-contained HTML report per sample.

Configuration: none.

Outputs

Read-length N50 and mean length are kept in the artifact metadata rather than on the Read record, because there are no canonical read columns for them.

Beyond the bundled catalog

The pipeline store

SeqDesk hosts a curated store that facilities can browse in the application or from the server shell. Under Admin → Pipelines, the Per-Sample Pipelines and Per-Study Pipelines tabs each offer three views — Installed, Needs setup and Available — and Available is the store catalog filtered to packages this deployment does not have yet, with an Install action per entry.

The same catalog is available through the operator CLI:

seqdesk pipelines list
seqdesk pipelines list --catalog order
seqdesk pipelines list --catalog study
seqdesk pipelines install <name>

Both surfaces use the same installation, readiness, and activation services. See Installing & Setting Up Pipelines for states, guided setup, and automation.

The store’s catalog is versioned independently of the application, which is what makes a pipeline installable into an existing deployment without shipping a new release. It is a subset of the bundled packages plus private entries — today it carries mag, submg, reads-qc, read-cleaning, simulate-reads, fastq-checksum, fastqc and the private metaxpath, so a package can ship in the repository without yet being offered in the store.

To propose a package for the store, see Contributing to the Official Pipeline Store.

MetaxPath — private add-on

Not part of the bundled catalog. MetaxPath (hzi-bifo/MetaxPath-Nextflow, v0.1.6) is a private, study-scoped package for long-read clinical metagenomics: host filtering, taxonomic profiling, assembly, virulence-factor prediction, AMR detection and reporting. Operators with access install it separately; nothing in the public catalog references it.

Two of its properties are worth knowing because they shaped SeqDesk’s own monitoring logic:

• It ships no step definitions, so trace progress cannot prove completion — which is why SeqDesk finalizes such runs only from a positive exit marker or a terminal scheduler state. See How a run is finalized.
• Its profiling processes request 20 CPUs by default, which a small SLURM allocation rejects pre-flight. See Execution modes.

Your own package

Any directory under pipelines/ with a valid manifest is picked up at startup — no registration step, no code change. See Adding Custom Pipelines.

Related pages

• Study vs Sequencing Order Pipelines — what each target means.
• Running a Pipeline — launching any of the above.
• Results & Outputs — what the outputs above turn into.
• Adding Custom Pipelines — writing your own package.