How to back up 900GB of genome data — the storage guide

Your raw genome data — the FASTQ, BAM, and VCF files that come with a whole genome sequencing order — totals roughly 900 gigabytes. That's larger than most people's Google Drive quotas, larger than the storage on the average laptop, and considerably larger than anything you're likely to have ever downloaded in one go before.

It's also irreplaceable in the sense that matters most: your provider may be around forever, or they may not. Nebula Genomics shut down in 2024. Customers who had downloaded their raw files kept their data. Customers who relied on Nebula's portal lost access to their reports overnight.

Your files are the asset. The portal is a convenience. Here's how to back them up properly.

What you're actually storing

The FASTQ files are the bulk. They're also the most valuable because everything downstream (alignment, variant calling, interpretation) can be regenerated from FASTQ using any bioinformatics pipeline — today's tools or 2035's. The BAM files are useful for immediate analysis without re-processing. The VCF is small and easy to keep everywhere.

The 3-2-1 rule applied to genomics

Standard IT backup guidance: keep at least three copies of important data, on at least two different types of media, with at least one copy off-site. Adapted for a 900GB genome:

Copy 1: Local primary (external SSD)

A 2TB portable SSD is the right primary. USB 3.2 or Thunderbolt-connected. Models from Samsung (T7, T9), Crucial (X9), SanDisk (Extreme), and WD (My Passport SSD) all work well in the $100–$150 range. Buy a 2TB even though you only need 900GB — the price difference to 2TB is small, and you'll want room for downstream analysis outputs.

SSD over spinning disk for this use case. You'll rarely write to the drive after the initial dump, but random-access reads for downstream analysis are much faster on SSD.

Copy 2: Local secondary (NAS or second SSD)

If you already run a home NAS (Synology, QNAP, TrueNAS), copy the genome data to a RAID-protected volume. Redundancy at the volume level protects against single-drive failure but doesn't help against ransomware, so keep the NAS copy behind a snapshot-capable filesystem (ZFS, Btrfs, or Synology's snapshot feature).

If you don't run a NAS, buy a second cheap external SSD. A cheap 2TB drive kept in a fireproof safe or at a family member's house counts as your second copy and provides an off-site option in one.

Copy 3: Cold cloud storage

For genuine off-site backup, cold-tier cloud storage is cheap enough to be a no-brainer for genomics.

AWS Glacier Deep Archive is the cheapest at roughly $1 per TB per month, but has retrieval latency of 12+ hours and per-GB retrieval fees. For genome data you almost never need to restore in a hurry, Deep Archive is fine. Backblaze B2 is the pragmatic all-round pick with reasonable price and low retrieval friction. Cloudflare R2 is worth considering if you value zero egress fees over the cheapest per-TB rate.

Encrypt cloud copies

Genome data is the most personal category of information a person can hold. Encrypt before upload. rclone with its built-in crypt remote type is the standard tool — free, open-source, well-audited, and works with all the cloud providers above. Set up once, run automatically on new backups.

Naming and organisation

Use a consistent folder structure so future-you knows what's what:

genome/
  dante-labs-2026/
    raw/
      sample-name_R1.fastq.gz
      sample-name_R2.fastq.gz
    aligned/
      sample.bam
      sample.bam.bai
    variants/
      sample.vcf.gz
    reports/
      pdf/

Include a plain text README documenting the reference genome used (GRCh38 for Dante's current pipeline), the sequencing run date, and the account under which the sequencing was ordered. Future-you will thank present-you when you come back to this in five years.

Verification

Every backup copy should have a checksum. Dante provides md5 checksums for the downloaded files — verify them after download and after each copy. On Linux and macOS:

md5sum sample_R1.fastq.gz against the provider's published hash. If they don't match, re-download. A partial or corrupt FASTQ is worse than no FASTQ — it can silently produce wrong variant calls downstream.

Realistic total cost

One-time hardware for local copies: $100–$300 (one SSD, optionally two). Monthly cloud storage for encrypted off-site: $1–$6 depending on provider and tier. Over a decade that's under $500 all-in to durably protect your genome data.

What not to do

• Don't rely on a single hard drive. All hard drives fail eventually. Single-drive storage is not backup, it's just deferred loss.
• Don't rely on the provider's portal. Companies shut down. Portals get rebuilt. Your local copy is the version that survives corporate changes.
• Don't upload unencrypted to shared cloud storage. Google Drive, Dropbox, iCloud — convenient but not appropriate for genome data at rest.
• Don't upload to genealogy databases casually. GEDmatch, FamilyTreeDNA and similar allow uploads of raw data. These databases have been used in investigative genetic genealogy by law enforcement. Legitimate use case, but understand the implications before uploading.

Bottom line

Buy a 2TB SSD, sign up for Backblaze B2 or AWS Deep Archive, run rclone crypt to sync encrypted backups on a schedule, verify checksums after every copy. Your genome data outlives you by design — back it up like it matters, because it does.