Compare your 23andMe, AncestryDNA or MyHeritage raw data with ancient DNA
Use your 23andMe, AncestryDNA or MyHeritage raw data with ancient genomes: the merge, what qpAdm, f4 and PCA can say about one genome, and limits.
Updated 2 min read
A raw data file from 23andMe, AncestryDNA, MyHeritage, FTDNA or LivingDNA holds your genotypes at several hundred thousand SNPs. Merged with ancient genomes, it can be a target in qpAdm, f-statistics and PCA, just like an ancient population. The results are research-grade models of your ancestry, not a company's ethnicity estimate.
What the raw data file contains
Consumer tests use genotyping chips. The raw data download is a text file, often zipped, with one line per SNP: its identifier (usually an rsID), chromosome, position and your two alleles. A typical file has 600,000 to 700,000 SNPs.
To get it, open your testing company's account settings and request the raw data download. The company prepares the file and sends a link.
How it is merged with ancient genomes
Ancient DNA tools need the target and the references on the same SNPs, in the same format. A merge:
- converts your file to EIGENSTRAT or PLINK format;
- matches your SNPs to the reference SNPs by position, and aligns alleles to the same strand;
- keeps the SNPs that both share.
Consumer chips were not designed for the AADR's 1240K and Human Origins SNP sets, so only part of your SNPs overlap. The overlap depends on the chip version. Fewer SNPs means larger standard errors, but a few hundred thousand shared SNPs are enough for most f-statistics and qpAdm models.
What you can learn from one genome
- qpAdm: whether your genome fits as a mixture of chosen ancient sources, for example western hunter-gatherers, Anatolian farmers and Yamnaya-related steppe herders, and in what proportions.
- f4-statistics: whether you share more drift with one ancient population than another, for example
f4(Mbuti, You; Population A, Population B). - PCA: where your genome projects among ancient and modern genomes.
- DATES: when two ancestries mixed in your family line, for an admixture recent enough to date.
Limits to keep in mind
- One genome is noisy. Expect standard errors of a few percent on qpAdm weights.
- Models, not labels. A qpAdm result says your genome is consistent with a model. Another model with other sources may also fit. Test competing models.
- Recent ancestry is better seen in modern references. For the last 1,000 years, present-day populations are often better sources than ancient ones.
- Chip and imputed files differ. Some companies give imputed or reprocessed data. Use the original raw genotype file.
Doing this in Nexju
Nexju accepts raw files from 23andMe, AncestryDNA, MyHeritage, FTDNA and LivingDNA as .txt, .csv, .zip or .gz. The file is converted privately on Nexju's servers, and it is never shared. You can then use your genome as the target in qpAdm models and rotations, f3, f4 and D statistics and qpWave on every plan. PCA projection and DATES are on the Max plan, as are whole-genome BAM or CRAM uploads by link.
If you use an AI assistant through the Nexju MCP server, the assistant only names your upload; it never receives your genotypes.
References
- Patterson N. et al. 2012. Ancient admixture in human history. Genetics 192. doi:10.1534/genetics.112.145037
- Mallick S. et al. 2024. The Allen Ancient DNA Resource (AADR) a curated compendium of ancient human genomes. Scientific Data 11. doi:10.1038/s41597-024-03031-7