How to Prepare and Deposit Genomic Data in NCBI SRA: A Workflow

By Dr. Zubair Khalid, DVM, MS, PhD · 2026-11-11

By the end of this guide you will be able to organize your raw sequencing files, generate checksums, register a BioProject and BioSample records, fill in the SRA metadata form, upload your files to NCBI, and confirm that your submission received accession numbers. You will need a computer with a Linux, macOS, or Windows command line, your raw sequencing files in FASTQ format, a text editor or spreadsheet program, and roughly one working session to prepare everything plus patience for the upload itself.

The Sequence Read Archive (SRA) is the National Center for Biotechnology Information (NCBI) public archive for raw sequencing reads. Depositing there is the standard way to fulfill data sharing requirements from journals and funders, and it is how your data becomes reusable by the wider community.

Why Deposit at All: What Happens to Data After Submission

Public archives are not storage bins. They are active research infrastructure. When researchers deposit genomic data in NCBI SRA, other groups can reanalyze the reads in aggregate to answer questions no single study was designed to ask.

A concrete example shows the scale. Thousands of Pseudomonas aeruginosa RNA sequencing profiles are publicly available through the SRA. One team gathered profiles from hundreds of studies produced by more than 75 research groups, uniformly processed 2,333 samples with the Salmon pseudoaligner, and built a gene expression compendium for hypothesis generation, complete with an algorithm to fold in new data as it is deposited [1]. Every one of those samples started as someone's submission. Without the deposits, that cross-study analysis would have been impossible.

This is the case for research data sharing in one paragraph: your raw reads outlive your paper, other people can validate, combine, and extend them, and depositing them is increasingly a condition of publication and grant support.

SRA Compared With Other Repositories

Choosing a repository is part of the workflow, so a short data repository comparison helps. Raw sequencing reads belong in the SRA. Processed, matrix-level data such as normalized expression tables belong in the Gene Expression Omnibus (GEO), which is also run by NCBI. Many genomics papers deposit raw reads in SRA and processed results in GEO, then link the two through the same BioProject.

The submission experience differs between them. GEO requires a multistep submission involving metadata preparation, file transfer, and MD5 checksum validation, and the manual process has been described as error-prone and dependent on bioinformatics experience [2]. SRA submission through the modern Submission Portal follows a similar overall shape, with a metadata form plus a file upload, so the same practical skills transfer between the two repositories.

Other archives in the same ecosystem include the European Nucleotide Archive (ENA) at EMBL-EBI and the DNA Data Bank of Japan (DDBJ). These three archives exchange data under the International Nucleotide Sequence Database Collaboration, so a submission to SRA is mirrored to the others. For most US-based researchers submitting to NCBI directly is the default path.

One practical note on workflow friction: deposition involves repetitive steps, checksums, and template filling, which is why groups have built automation tools. A recent web-based tool for GEO, for example, automated uploads, MD5 calculation, and metadata template population, cutting submission initiation from 2 to 3 hours to under 20 seconds across 30 or more users and more than 50 upload sessions [2]. Similar automation principles apply to SRA deposits, particularly for labs with recurring submission types.

Before You Start: What You Need on Hand

Gather these before touching the portal.

1. Raw sequencing data. Original output from the sequencing facility, ideally in FASTQ format (a plain-text format that stores a sequence read, its quality scores, and the identifiers connecting them). Compress with gzip, which produces .fastq.gz files.
2. Sample metadata. For each sample, you need organism, strain or isolate name, tissue or source, library strategy (for example WGS or RNA-Seq), library source, library layout (single-end or paired-end), sequencing platform, and file names.
3. A project description. A short public summary of the study, its scope, and its funding, which becomes your BioProject record.
4. An NCBI account. Registered at the NCBI site, ideally linked to an ORCID identifier.
5. Checksums. An MD5 checksum is a 32-character fingerprint computed from a file. If a single byte changes during upload, the checksum changes, so NCBI and you can verify the transfer was clean.

Deposit Genomic Data in NCBI SRA: Step-by-Step Workflow

This section walks through the full deposit genomic data in NCBI SRA workflow in order. Each step lists the action, a worked example, and what to check.

Step 1: Register in the Submission Portal

Log in to the NCBI Submission Portal with your NCBI account and start a new submission under "Sequence Read Archive." The portal walks you through a wizard with tabs for submitter information, BioProject, BioSample, metadata, and files. You can save a draft and return to it, which matters for large submissions.

Step 2: Organize and Compress Your Files

Keep one directory per submission, with unambiguous file names that match your metadata. Then compress.

gzip sampleA_R1.fastq sampleA_R2.fastq
ls -lh

Expected output:

-rw-r--r-- 1 user group 4.2G sampleA_R1.fastq.gz
-rw-r--r-- 1 user group 4.3G sampleA_R2.fastq.gz

Gzip typically shrinks FASTQ files to roughly a third of their raw size and does not damage the data. Do not use archiving formats that bundle multiple files, such as tar or zip, because the archive expects individual read files.

Sanity-check the files before uploading. Count the reads in one file:

zcat sampleA_R1.fastq.gz | awk 'NR%4==2{c++} END{print c}'

Expected output:

52341982

For paired-end data, the R1 and R2 counts must match exactly. If they do not, the pairing is broken and the submission will produce unusable data. Also glance at the first records:

zcat sampleA_R1.fastq.gz | head -8

Each read should occupy exactly four lines: an identifier starting with @, the sequence, a + line, and the quality scores.

Step 3: Generate MD5 Checksums

On Linux or macOS:

md5sum sampleA_R1.fastq.gz sampleA_R2.fastq.gz > md5.txt

Expected output (illustrative values):

b1f9c2a4e7d30815c6a2f0b9d4e1a7c3  sampleA_R1.fastq.gz
8e3d5f1b0a9c4627d8e5b3a1f6c09d42  sampleA_R2.fastq.gz

On Windows, in Command Prompt:

certutil -hashfile sampleA_R1.fastq.gz MD5

Save the checksum list with your submission. The portal lets you record these values, and it is the standard way to prove file integrity after transfer. The same MD5 validation step appears in the GEO deposition workflow, which reflects how widely this practice is embedded in repository submissions [2].

Step 4: Create the BioProject

A BioProject is the umbrella record for your study, describing the overall research effort, not the individual samples. In the portal you provide a project title, a public description, the organism scope, relevant grants, and data release timing. Most users choose to release data at publication or on a set future date. NCBI assigns the project an accession in the form PRJNA followed by digits, for example PRJNA1234567.

Step 5: Create BioSample Records

A BioSample record describes one biological specimen: organism, strain, tissue, collection details, and any attributes relevant to your study. You can register samples one at a time through the web form or many at once with a template spreadsheet. Names must be unique within the submission. NCBI assigns accessions in the form SAMN followed by digits.

Step 6: Fill the SRA Metadata

The SRA metadata ties each BioSample to its file and to the technical details of how the library was sequenced. The core fields you will encounter are summarized here.

Fill these values with a spreadsheet template where one is offered, because the portal's validation catches missing or inconsistent values before submission. Template-based completion is one of the mechanisms automation tools use to prevent errors, since it removes free-form typing from repetitive fields [2].

Step 7: Upload the Files

The portal gives you a host, username, password, and a unique upload directory when you reach the file step. Three upload methods are available.

For FTP with the lftp client:

lftp -u <username>,<password> <host>
cd <your_upload_directory>
mkdir sampleA
put sampleA_R1.fastq.gz
put sampleA_R2.fastq.gz
by

For Aspera, the general form is:

ascp -i <key_file> -k 2 sampleA_R1.fastq.gz <username>@<host>:<upload_directory>/

The -k 2 flag makes Aspera resume and verify transfers, which pairs naturally with your MD5 checksums. After the transfer, NCBI checks the uploaded files against the checksums and metadata you supplied, and file integrity validation is a standard gate before a submission is accepted [2].

Step 8: Submit and Wait for Accessions

Once files are uploaded and the metadata tab shows no errors, click the final submit. NCBI staff and automated validation review the submission. You receive email notifications, and the status in the portal changes from submitted to processed. When processing completes, NCBI issues accession numbers: one for the experiment or run for each sample, linked to your BioProject and BioSample accessions.

Processing time varies with submission volume and the complexity of the metadata, so build delay into your publication timeline rather than planning a same-day turnaround.

Step 9: Verify the Public Record

When your data releases, retrieve the BioProject accession from your submission and open the project page. Check three things.

1. Every sample appears with the correct organism and attributes.
2. The run entries point to files that download correctly.
3. Your paper, when written, cites the accessions in the Data Availability statement.

This closes the loop that makes research data sharing work: a reader should be able to go from your paper to the raw reads in a few clicks.

Workflow Tips That Save Time

These are the deposit genomic data in NCBI SRA workflow tips that experienced submitters rely on.

• Start the account and BioProject before the data arrives. Account issues and registration backlogs are easier to absorb early than during a revision deadline.
• Use one naming scheme for files, samples, and metadata rows. The most common validation failures are name mismatches, not scientific errors.
• Compute checksums on the machine where the files originate, not after copying. A checksum of an already-corrupted copy validates corruption.
• Test your upload pipeline with one small file first. Confirm your transfer method, credentials, and directory path before committing to a multi-hundred-gigabyte transfer. Automation work in adjacent repositories shows how much time batch testing saves when uploads and checksum calculations run in parallel rather than one at a time [2].
• Keep the submission email thread. If you need to correct anything later, the submission accessions and correspondence are your reference.
• Write your Data Availability statement now. You can draft the sentence with placeholders and drop in the accessions the day they arrive.

Common Mistakes and Limitations

The most frequent deposit genomic data in NCBI SRA workflow common mistakes, with fixes.

1. Uploading BAM files where raw reads are expected. SRA accepts primary sequencing data. If your study produced raw reads, submit the FASTQ files, and keep aligned BAM files for a different track or a supplemental repository appropriate for processed data. If your data genuinely has no raw-read form, consult the portal guidance for that data type.
2. Broken paired-end naming. Some instrument software names mates R1 and R2, others use _1 and _2. Whatever convention you use, the metadata and the actual files must agree, and the read counts must match. Fix mismatched counts before uploading, not after.
3. Metadata that conflicts with the files. Declaring paired layout while submitting only one file, or declaring an organism that does not match your BioSample, triggers review delays. The fix is to validate your spreadsheet locally before submission and to treat every portal warning as a real problem.
4. Skipping checksums and then wondering about a corrupt file. If NCBI flags a checksum mismatch, regenerate the checksum from your original file, confirm the original is intact, and re-upload the affected file. Checksum workflows exist precisely because silent corruption during large transfers happens, which is why automated deposition tools build MD5 calculation directly into the upload process [2].
5. Assuming immediate public release. Submissions process in a queue, and review can request corrections. If your paper must cite accessions by a deadline, submit with enough lead time and set an explicit release date rather than relying on publication-day timing.
6. Treating the deposit as the end of curation. Data that are hard to reuse still cost downstream users effort. Aggregate reanalysis studies apply their own quality control and filtering to deposited reads, for example excluding samples with aberrant housekeeping gene expression before building compendia [1]. Clean metadata at submission time reduces how often deposited samples get filtered out of later analyses.

A broader limitation: this workflow covers the standard path for raw sequencing reads. Specialized data types, contaminate-prone human data with consent restrictions, and controlled-access requirements involve additional review steps that this general workflow does not detail. When your situation involves restricted human data, follow the consent provisions of your protocol and the portal's controlled-access instructions.

Checking Your Result and Next Steps

After accession numbers arrive, your submission is finished but your data's public life is starting. Do these three things.

1. Link the accessions everywhere. Paper, preprint, grant reports, and your lab website should all cite the BioProject and run accessions.
2. Monitor reuse. Citations to the accession in other papers, or your samples appearing in aggregate reanalyses such as the multi-study compendia model [1], are evidence of impact that funding agencies value.
3. Archive the submission locally. Keep your original files, checksums, metadata spreadsheets, and the submission emails for the retention period your institution or funder requires. Deposits answer "where is it," but your local archive answers "what exactly did we submit."

Frequently Asked Questions

Do I submit raw reads or processed data to SRA?

Raw reads. The SRA stores primary sequencing data, so FASTQ files are the standard input. Processed outputs such as expression matrices belong in GEO or a supplemental file linked through your BioProject.

How long before my SRA data becomes public?

It depends on submission volume and whether validation finds problems. Submissions enter a review queue, and you set the release timing yourself, so plan lead time and choose a release date or release-on-publication in the BioProject.

Can I keep my data private while my paper is under review?

Yes. You can submit now and schedule release at a future date or upon publication. Reviewers and editors can often be given the accession numbers, which resolve once the record is released.

What file format and compression should I use?

FASTQ files compressed with gzip, one file per read direction for paired-end libraries. Keep the original basecaller or demultiplexer output rather than re-encoding it through intermediate tools.

Do I need one BioProject for every submission?

One BioProject per study, not per sample. Each sample gets a BioSample record, and all the BioSample records link back to the single BioProject that describes the overall effort.

What is the difference between SRA and GEO?

SRA archives raw sequencing reads. GEO archives processed, experiment-level data such as gene expression tables. Submissions to both involve metadata and file transfer, and the manual deposition process in GEO is well known for being multistep and error-prone without automation [2].

How do I fix an error in a submission that is already processed?

Contact NCBI through the link provided in your submission confirmation email and describe the correction. Changes to metadata or additional files go through the existing submission record, so keep your confirmation emails accessible.

Can other scientists reuse my deposited data?

Yes, and that is the point of depositing. Publicly released SRA data can be downloaded, reprocessed, and combined across studies, which is how aggregate analyses spanning hundreds of studies and thousands of samples become possible [1].

Sources

1. Computationally Efficient Assembly of Pseudomonas aeruginosa Gene Expression Compendia.
2. GEO uploader: simplifying the data deposition in the GEO repository.