Results: Assemblies & Bins

After a pipeline completes, SeqDesk parses the outputs and links them to samples in the database. This page covers the result types and their data structure.

Assemblies

An assembly represents a set of assembled contigs for a sample.

Preferred Assembly

Each sample can have a preferred assembly — the assembly used for downstream analysis and ENA submission. When multiple assemblies exist (e.g., from MEGAHIT and SPAdes), the facility admin selects which one to use.

Output Location

MAG pipeline assemblies are found at:

{runDir}/Assembly/MEGAHIT/{sampleName}.contigs.fa.gz

Genome Bins (MAGs)

Genome bins are individual genomes extracted from metagenomic assemblies through binning algorithms.

Quality Metrics

Bin quality is assessed by CheckM, which estimates:

• Completeness — what percentage of a complete genome is present
• Contamination — what percentage of the bin is from other organisms

Common quality thresholds:

Bin Sources

The MAG pipeline can produce bins from multiple binning tools:

1. MetaBAT2 — default binning
2. MaxBin2 — alternative binning
3. CONCOCT — optional (disabled by default)
4. DAS Tool — bin refinement, combines results from other tools

DAS Tool refined bins (at GenomeBinning/DASTool/bins/) are preferred.

Pipeline Artifacts

Beyond assemblies and bins, pipeline runs produce additional artifacts:

Each artifact tracks:

• File path, size, and checksum
• Which pipeline step produced it
• Associated study and sample
• Tool-specific metadata (JSON)

Read Outputs

Order pipelines can also produce read-level results. SeqDesk distinguishes metadata updates from new read files:

For read-cleaning and host-removal workflows, prefer sample_read_candidate with admin_review. This lets admins inspect MultiQC or other HTML reports before switching the order to cleaned reads. Raw and unknown source reads remain preserved; promoted cleaned reads become the active reads used for downstream order pipelines and delivery.

Pending Writebacks (Candidate Review)

When a pipeline stages sample_read_candidate outputs with the admin_review policy (as the Read Cleaning pipeline does), the cleaned FASTQ files are stored as run artifacts and held as pending writebacks rather than being applied automatically. A facility admin reviews the run’s reports, then chooses which per-sample candidates to promote into active reads.

Promotion is non-destructive: existing raw and unknown source reads are preserved, and an existing active cleaned read is superseded (not deleted) by the promoted candidate. The newly promoted reads become the active reads used for delivery and downstream order pipelines.

API

The review flow is backed by per-run endpoints (FACILITY_ADMIN only):

Both require the FACILITY_ADMIN role; promotion is disabled in the public demo. The GET summary marks each candidate as candidate or promoted and returns the per-sample target read data class so the UI can show what will change before an admin confirms.

Reports and Preview Files

Pipeline packages should mark HTML, PDF, TSV, and text summaries as run_artifact, study_report, or order_report outputs. SeqDesk ranks these files on the run output page, shows the primary report first when one is available, and keeps the full output browser available for secondary files.

Assemblies Viewer

The Assemblies page (/assemblies) provides a centralized view of all genome assemblies across studies. The table shows:

Assembly Selection Modes

Viewing Results

Results are accessible from multiple places:

• Assemblies page — centralized view of all assemblies across studies
• Study page → Pipelines tab — overview of all runs and their results
• Sample detail page — assemblies and bins for a specific sample
• Pipeline run detail — all outputs from a single run