Skip to Content
Sequencing FilesOverview

Sequencing Files

Sequencers write files. SeqDesk’s job is to work out which file belongs to which sample, record that link durably, and then control who can see it. This section covers the whole path from “there is a directory full of .fastq.gz” to “the researcher can download their reads”.

Everything here is facility-admin only and lives on a sequencing order’s Sequencing Data page (/orders/[id]/sequencing). There is no standalone Files page: the legacy /orders/[id]/files route redirects — non-demo facility admins to the Sequencing Data page, everyone else to the order page. The order must be SUBMITTED or COMPLETED; a DRAFT order rejects every sequencing action.

The mental model

SeqDesk never moves or renames your sequencing data. It points at it.

site.dataBasePath                 ← one configured root directory
└── 20260415/                     ← whatever the sequencer wrote
    ├── barcode01/
    │   └── FAX12345_pass_barcode01_a1b2c3d4_0.fastq.gz
    └── barcode02/
        └── …

        ▼ scan (read-only)  ▼ match  ▼ assign

Read { sampleId, file1: "20260415/barcode01/FAX…_0.fastq.gz", file2: null,
       dataClass: "cleaned", isActive: true }

Four consequences follow, and they explain nearly every question people have:

  • Paths are stored relative to the data base path. Moving the root directory and re-pointing the setting keeps every link working. Moving a file inside the root breaks its link, and the sequencing overview flags the sample as pointing at a missing file.
  • A file is never copied on assignment. Only uploads and promoted pipeline candidates put new bytes on disk.
  • Assignment creates a Read record on the sample. Reads are the unit that pipelines, delivery and ENA submission consume — not files, not runs.
  • Nothing escapes the root. Every path is resolved through a traversal check; absolute paths and .. segments are rejected outright.

The three ways files get linked

RouteWhen to use itWhat it produces
Auto-Discover on the Associate viewThe normal case — data has landed in the data directorySuggestions you review and apply
Manual associationNames are ambiguous, or discovery found nothingA link you pick file-by-file
UploadThe files are not on the server at allA copied file plus a Read, under a generated path

All three converge on the same Read record, and all three respect the raw-data protection rules described in Assigning Files to Samples.

Why the run plan comes first

Discovery tries a run-plan barcode before anything else. On a Nanopore output tree where every filename is a meaningless flow-cell id, that is the difference between deterministic matching and no matching at all.

Set up the run plan before the data arrives. If your order form collects a per-sample barcode, the run plan builds itself when you create the run.

Linking is not releasing

Assigning a read makes it visible to the facility. The researcher still sees nothing until a facility admin presses Make downloadable to user on the Delivery to user card. Even then, only active reads classified cleaned and customer-visible reports are released.

That two-step design is deliberate: it lets the facility associate raw output, run cleaning pipelines and review QC without ever exposing unfiltered — possibly human-contaminated — reads. See Releasing data to the researcher.

File Discovery & Auto-DetectDirectory scanning, depth and ignore rules, R1/R2 pairing, the in-flight write guard, and scan caching.Assigning Files to SamplesThe barcode-first matching cascade, confidence scores, raw-data protection, and pipeline writeback.Sequencing RunsThe run plan that makes barcode matching possible.Data Storage settingsWhere the data base path and the scan settings are configured.