How to Download Your Raw DNA Data From Any Provider

Every consumer DNA service lets you download the underlying data, but each buries it somewhere different and hands you a different kind of file. This page collects the download path for the providers people actually use, and — more usefully — tells you what you end up holding afterwards.

First: which of the two files do you have?

This matters more than the download steps, because it decides what your file can and cannot tell you.

Genotyping arrays are what ancestry kits run. A chip reads a fixed list of pre-selected positions — a few hundred thousand — and reports what it finds at each. Everything off that list is never read. 23andMe, AncestryDNA, MyHeritage and FamilyTreeDNA all work this way.

Whole genome sequencing reads virtually the entire genome, all ~3 billion base pairs. Nebula, Dante Labs and Sequencing.com sell this.

The gap is large: an array file covers a fraction of one percent of your genome, while a 30x sequencing file can cover far more. Neither format proves that a variant is absent, though: arrays never read positions outside the chip, and sequencing can still miss difficult regions or variant types.

Genotyping array kits

23andMe

Open your account's raw data page and request the download. You get a zipped, tab-separated text file whose name starts with genome. Past the header comments, each line is one marker across four columns: rsid, chromosome, position, genotype — your two alleles combined into a value like AG. The chip in use since 2017 is the Illumina Global Screening Array, reading roughly 640,000 positions on GRCh37. The full 23andMe guide goes deeper.

AncestryDNA

From your DNA results, go to Settings, then Download Raw DNA Data. Enter your password, tick the consent box, and confirm. Ancestry then emails you — you must click Confirm data download in that email before the file unlocks. The result is a zipped AncestryDNA.txt on build 37.

One structural difference worth remembering: Ancestry splits your two alleles into separate allele1 and allele2 columns, so its files have five columns where 23andMe has four. Same information, different layout. The AncestryDNA guide goes through it column by column.

MyHeritage

Under the DNA tab, open Manage DNA kits, use the three-dots menu and choose Download. After you confirm and accept the terms, MyHeritage emails a link that expires in 24 hours; follow it, re-enter your password, and the download starts.

Two constraints catch people out. Only the manager of a kit can download it, not someone it was merely assigned to. And it works on a computer or Android device but not on an iPhone or iPad. MyHeritage publishes one format only: a tab-delimited text file with a descriptive header and one line per marker.

FamilyTreeDNA

Go to Results & Tools → Autosomal DNA → Download Raw Data. Two-factor authentication must be enabled before the download is offered at all, so turn that on first if you haven't.

Pick Build 37 Concatenated Raw Data, which covers all autosomal SNPs plus those on the X chromosome. Family Finder exports are comma-separated files compressed into .gz, carrying RSID, chromosome and position columns. Genetic males are also offered a separate Y-SNP file.

Whole genome sequencing services

Nebula Genomics and DNA Complete

Nebula sells 30x whole genome sequencing, and DNA Complete runs through the same portal. Your VCF is the file to reach for — read-level files hold raw output rather than a variant list, and most tools won't take them.

One caveat: Nebula publishes no public help centre, and access to read-level CRAM and FASTQ data has historically gone through customer support rather than a direct download. If you need more than the VCF, check your account and ask Nebula rather than trusting steps that may be a year out of date.

Dante Labs

Downloads live in the Genome Manager portal. Dante provides FASTQ, BAM and several VCFs — separate files for SNPs, indels, structural variants and copy number variants.

There is a deadline attached: FASTQ and BAM are free to download for 30 days after your results are ready, after which regaining access requires a subscription. VCF files stay available. If your results have just landed and you want the read-level data, get it now.

Sequencing.com

Select Files in the header, choose the genome, and use the download arrow beside the file you want. Smaller files such as .vcf and .txt download immediately. Large files like .fastq need preparation — usually one to three days — with an email when they're ready. Downloads work whether or not you hold a subscription.

Once you have the file

The file on your disk is inert. Nothing in it says which of your variants have been clinically assessed, because that lives in a separate public archive: ClinVar, the NIH's record of how variants have been classified — benign, uncertain, likely pathogenic, pathogenic — and how much review stands behind each call.

Matching your positions by hand isn't realistic at several hundred thousand markers, let alone several million. BioDecode does that comparison locally: point it at your file and it matches every position against a local copy of ClinVar on your own machine, nothing uploaded. It reads VCF, TXT, CSV, TSV, GZ and ZIP, covering every export above. If that's the job you're after, download BioDecode.

One thing to check before comparing coordinates against any database: every array export above uses GRCh37 / build 37, while sequencing VCFs vary by provider. The build is an address system, so the same position number means different things under different builds — read the file header rather than assuming.

The trap on the other side of the download

Two limits apply no matter which provider you used.

A missing marker is not a clean result. This is the most common misreading of an array file. The chip only tests the positions it was built to test; if a variant isn't in your data, the array didn't look there. Sequencing narrows that gap but doesn't close it, since some regions stay hard to read.

A flagged marker is a lead, not a conclusion. Consumer raw data isn't clinically validated. A 2018 study in Genetics in Medicine found roughly 40% of concerning variants in direct-to-consumer raw data were false positives against clinical-grade confirmation. Anything that looks significant belongs in an accredited (CLIA) laboratory and a conversation with a genetics professional first.

Held to those limits, your own file is a reasonable place to start. How to find pathogenic variants covers what the classification labels mean.

Frequently asked questions

Do I have to pay again to download my raw data?

No. Every provider here includes the download with the test you already bought. Dante Labs is the one to watch: FASTQ and BAM are free for 30 days after results are ready, after which a subscription is needed to reach them again. VCF files remain available.

Which file should I download if I'm offered several?

For checking variants against a database, take the VCF. FASTQ, BAM and CRAM hold raw sequencing reads rather than a variant list — far larger, and most analysis tools won't accept them.

Can I download my raw data on a phone?

Sometimes, but don't count on it. MyHeritage supports computers and Android devices while excluding iPhone and iPad, and sequencing files run to tens of gigabytes regardless.

Is it safe to upload my raw data somewhere to analyze it?

That depends on the service and what it does with the file afterwards. Is it safe to upload your DNA? covers what changes once a genome file leaves your computer.

This article is educational and is not medical advice.