Which DNA Test Has the Best Raw Data for Third-Party Tools?

23andMe, AncestryDNA, MyHeritage, FamilyTreeDNA and Living DNA all export raw data, but the chips differ. What each holds, where they overlap, which to pick.

If you plan to do your own analysis, the test that gives you the best report is not necessarily the one that gives you the best file. What matters for third-party tools is which markers are on the chip, whether the export includes the sex and mitochondrial chromosomes, and how well the file’s marker set overlaps with the references you want to use. Here is how the major services compare on the things that affect what you can do with the download.

The chips underneath

Every major consumer service uses an Illumina genotyping array, but not the same one, and not the same version over time. A few points that stay true across versions:

  • 23andMe moved to a customised Illumina Global Screening Array (GSA) with its v5 chip in 2017. It reads roughly 640,000 markers, with a meaningful set of Y-chromosome and mitochondrial positions and a curated selection of medically relevant variants.
  • AncestryDNA uses its own customised array - v2 since 2016 - with around 670,000 markers. Its design is oriented toward ancestry and relative matching, and its content overlaps less with the GSA family than the other services do with each other.
  • MyHeritage, FamilyTreeDNA and Living DNA all moved to GSA-based chips in the late 2010s, each with custom additions. Their files therefore share most of their markers with 23andMe v5 and with one another.

Marker counts are approximate and change with chip versions; the ISOGG comparison chart tracks the current state better than any static article can.

What “overlap” means in practice

Third-party tools do their work by looking up specific rsIDs. A trait tool has a list of markers it knows how to interpret; an ancestry tool has reference panels built on particular SNP sets. A file is only useful to a tool to the extent that the file’s rsIDs match the tool’s.

This produces a pattern that surprises people:

  • Two GSA-based files - say 23andMe v5 and MyHeritage - typically share the large majority of their markers.
  • An AncestryDNA v2 file and a GSA-based file share far fewer, because the arrays were designed independently.
  • Older files from the same company (23andMe v3 and v4, for instance) differ substantially from the current chip.

So the answer to “which file works best with tool X” is usually “the one whose chip the tool was built around”. Ancestry tools built on academic reference panels tend to favour GSA-style content; genealogy databases that accept uploads have tuned themselves to handle every major chip.

Sex chromosomes and mitochondria

If you care about haplogroups - the deep paternal and maternal lineages - the difference between services is stark. 23andMe includes several thousand Y-chromosome and mitochondrial SNPs, enough to place a haplogroup reasonably well. AncestryDNA includes some Y and mitochondrial positions but does not report haplogroups and covers them more thinly. The GSA-based services fall between, with Living DNA in particular emphasising haplogroup coverage.

For medical-adjacent lookups, 23andMe’s chip deliberately includes many of the variants people search for - APOE, BRCA founder variants, pharmacogenomic markers - because its own reports use them. Other chips include some by coincidence of the GSA design and miss others. If you have a specific marker in mind, the honest answer is to check whether it is in the file before choosing a tool; a missing marker is far more common than a wrong one.

Export formats

Every service lets you download a plain-text raw file, but the layouts differ, which is a small obstacle that trips up a surprising number of tools:

  • 23andMe: tab-separated, header lines commented with #, one genotype column (AG).
  • AncestryDNA: tab-separated, two allele columns (A and G separately), with sex and mitochondrial chromosomes coded as numbers 23 to 26.
  • MyHeritage and FamilyTreeDNA: comma-separated with every field in quotes and the genotype in a RESULT column.
  • Living DNA: tab-separated, similar to 23andMe.

Genespiral reads all of these directly, along with VCF files, and normalises them before any analysis - see our guides to downloading from 23andMe, AncestryDNA and MyHeritage or FamilyTreeDNA.

If you are choosing a test for the data

For most people who want a single file to explore with third-party tools, a current 23andMe file is the most broadly useful: strong marker overlap with academic references, the best haplogroup coverage, and the most medically relevant content. AncestryDNA has the largest matching database, which matters enormously for genealogy but not at all for what you can do with the download. If you already have a file from any service, the practical move is to use it and check its coverage before paying for another; the DNA file report on the Genespiral dashboard shows how many of the trait markers it reads are present in your file.

And if you do end up with two files, comparing them is a useful exercise in its own right - our guide to comparing two DNA files privately walks through it.

This article is educational only and is not medical advice. Chip contents and formats change; verify against your own file.

References

Further reading