The NCBI Sequence Read Archive (SRA) stores raw, unaligned next-generation sequencing reads and their quality scores. It is not, on its own, a place you submit a single file to — a complete SRA submission is actually three linked records: a BioProject (the study), one or more BioSamples (the biological material sequenced), and the SRA experiment/run data itself (the reads). Funder and journal data-availability mandates that say “deposit your reads in SRA” are really asking for all three, correctly cross-referenced, not just an uploaded FASTQ file.
Not what you were looking for? If you mean the professional association, see SRA International (the Society of Research Administrators International) — a completely different organization from the sequence archive covered on this page.
The three linked records, and how their accession numbers relate
Each level of an SRA submission gets its own NCBI accession number, and the numbering scheme itself tells you which record type you are looking at. Source: NCBI’s BioProject Help documentation and SRA Submission Quick Start guide, last verified 2026-08-16.
| Record | What it represents | Accession prefix | Example |
|---|---|---|---|
| BioProject | The overall research study or initiative — registered once per project, even if it produces many samples and runs | PRJNA (NCBI-submitted; PRJEB/PRJDB for ENA/DDBJ-submitted projects mirrored to NCBI) | PRJNA43021 |
| BioSample | One biological source material — an organism, isolate, cell line, or environmental sample — with its descriptive attributes | SAMN | SAMN00000000 |
| SRA Experiment | One library preparation and sequencing strategy applied to a BioSample | SRX | SRX000000 |
| SRA Run | One actual sequencing run (the read data/FASTQ-equivalent) produced by an experiment | SRR | SRR000000 |
A single BioProject can contain many BioSamples, each BioSample can be sequenced under several experiments (e.g., different library preps), and each experiment can have multiple runs (e.g., re-sequencing, multiple lanes). This one-to-many structure is why a BioProject accession, not an SRA run accession, is usually what a paper’s data-availability statement should cite: it is the stable umbrella identifier that keeps resolving to all associated samples and reads, including any added after publication.
Where does your data belong: SRA, dbGaP, or GEO?
NCBI’s own SRA Submission Quick Start is explicit that not all sequencing-adjacent data goes into the public SRA. Route your data by what it is and who should be able to see it:
| Your data | Correct destination | Access model |
|---|---|---|
| Raw reads from non-human samples, or human reads with no individual-level privacy concern and full consent for open release | SRA (public) | Open, no login required once released |
| Human genotype/phenotype data, or human sequence data requiring controlled access under the consent obtained from participants | dbGaP (Database of Genotypes and Phenotypes) | Controlled access via data access committee request |
| Functional genomics results built on top of sequencing (expression matrices, processed values, experimental design metadata) | GEO (Gene Expression Omnibus) | Open; GEO submissions typically still reference underlying SRA runs |
| Human metagenomic data that may contain incidental human sequence contamination | Contact SRA staff first for screening before public deposit | Screened, then public if cleared |
If you are unsure which of NCBI’s several genomics repositories is the right target for a given data type more broadly — not just reads — CASRAI’s guide to choosing an open data repository and the separate guide to GenBank cover the annotated-sequence case, which is a different NCBI resource from SRA: GenBank holds assembled, annotated sequences (a finished gene or genome record); SRA holds the raw, unassembled reads a sequencer produced.
Submission workflow
NCBI runs all of this through one interface, the Submission Portal, which validates your metadata and then routes the data into the target database (SRA, BioProject, BioSample, dbGaP, or GEO as appropriate). The core sequence is:
- Register (or select) a BioProject. Provide the project title, description, and data type (e.g., genome sequencing, transcriptome, metagenome, targeted locus). If you already have a BioProject accession from an earlier submission on the same study, reuse it rather than creating a duplicate.
- Register one BioSample per biological source. Each BioSample requires a package-specific set of attributes (organism, collection date, geographic location, isolation source, and other fields depending on the chosen BioSample package) — this is the metadata that makes the sample scientifically interpretable to someone reusing the data later.
- Upload the sequencing files and describe the library. For each BioSample, specify the library strategy, source, selection method, layout, and the sequencing platform/instrument used, then upload the read files (commonly FASTQ, BAM, or platform-native formats — see NCBI’s File Format Guide for accepted formats per platform).
- Set a release date, if needed. Data can be held private until a specified date rather than released immediately on submission — commonly used to align public release with a paper’s publication date while still letting authors cite a real accession number in the manuscript before that date.
- Submit and resolve validation errors. The Submission Portal performs format and metadata validation and returns errors/warnings that must be corrected before the data is accepted; NCBI publishes a dedicated troubleshooting guide for the most common submission errors.
- Record your temporary submission ID. Every submission gets a temporary ID in the form
SUB#######before permanent accessions are assigned — quote this ID in any correspondence with SRA staff ([email protected]) if you need help.
Source: NCBI SRA Submission Quick Start (ncbi.nlm.nih.gov/sra/docs/submit) and NCBI BioProject Help (Bookshelf NBK54364), both directly reviewed, last verified 2026-08-16.
Human data and consent
NCBI’s guidance is direct on this point: before depositing human-derived data into the public SRA, the submitter must have consent from the donating individual authorizing release into an unprotected, publicly accessible database. Human data collected under a consent that restricts release to controlled-access use must go to dbGaP instead, never to the public SRA. This is an institutional-compliance point as much as a technical one: an institution’s data management plan and IRB-approved consent language should specify, before data collection begins, which of these two destinations a given dataset is headed for, so the submitting lab is not deciding an access-restriction question after the fact.
Common submission problems
- BioSample package mismatch. Choosing the wrong BioSample package (e.g., a generic package when a pathogen- or host-associated package is required by the target organism/data type) produces validation errors or, worse, a record accepted with missing fields reviewers later ask for. Check NCBI’s current package list against your data type before starting.
- Citing an SRA run accession instead of the BioProject accession in a manuscript. Run accessions (SRR) are numerous and non-obvious to browse from; a data-availability statement that cites only a handful of SRR numbers, rather than the BioProject (PRJNA) accession, makes it harder for a reader to find the complete dataset, especially if runs are added later.
- Submitting before consent/IRB language covers public release. Reversing a public SRA deposit after the fact (because consent did not actually cover open release) is a compliance problem, not just a technical one — confirm the access model before, not after, submission.
- Mismatched or missing metadata between BioSample and the sequencing library description. Organism, collection date, and source fields entered inconsistently across BioSamples in the same BioProject are a common source of post-submission correction requests.
How this fits into repository selection and data management planning
SRA is a single-domain, NCBI-operated repository, not a general-purpose one: it accepts raw sequencing reads and nothing else. For sequencing data specifically, most US federal funders and most journals treat SRA (or an INSDC-partner archive, since NCBI SRA, EMBL-EBI’s ENA, and Japan’s DDBJ synchronize records) as the default expected destination. It has not undergone CoreTrustSeal certification the way some generalist repositories have — see CASRAI’s guide to what CoreTrustSeal certification involves for how that process works for repositories that do pursue it — but its role as the field’s default sequencing-read archive, decades of continuous operation, and INSDC mirroring are generally treated as sufficient trust signals by funders for this specific data type. If your data management plan needs to name a repository for sequencing reads, SRA/BioProject/BioSample is very likely the correct answer to write down. For the manuscript-facing side of this — what to actually write in a data-availability statement once you have your accessions — see CASRAI’s data availability statement worked examples and the data availability statement dictionary entry.
Frequently Asked Questions
Do I need a BioProject before I can submit to SRA?
Yes. Every SRA submission must be linked to a BioProject; the Submission Portal will prompt you to register a new one or select an existing one from your account before you can register BioSamples or upload reads.
What is the difference between BioProject and BioSample?
A BioProject describes the study as a whole (its scope, goals, and data types); a BioSample describes one specific biological specimen used in that study. One BioProject typically links to many BioSamples.
Can I submit sequencing data before my paper is published?
Yes. You can submit and receive accession numbers to cite in a manuscript while setting a release date that holds the actual read data private until publication (or another date you specify), rather than releasing it immediately.
What is the difference between SRA and GEO?
SRA stores raw sequencing reads. GEO stores processed functional-genomics results (expression values, experimental design) and commonly references the underlying SRA run accessions rather than duplicating the raw reads itself.
Can human sequencing data go into the public SRA?
Only if the donating individual’s consent explicitly authorizes release into an unprotected, publicly accessible database. Human data under a controlled-access consent must be submitted to dbGaP instead, not the public SRA.
How long does an SRA submission take to process?
This depends on file size, validation issues, and current queue volume, and NCBI does not publish a fixed turnaround guarantee; using the Submission Portal’s validation feedback to resolve errors before final submission is the main lever a submitter has to avoid delay.
