Eurogenes K13 is the default calculator of GEDmatch's most-used admixture project, and probably the single most-quoted component breakdown on the internet. It dates from around 2013, it has thirteen components, and it is still worth running — provided you read it as what it is: a fixed set of statistical profiles from before the ancient-DNA revolution, not a measurement of where your ancestors lived.
Background first if you need it: what an admixture calculator is and the map of GEDmatch's projects. This post is about K13 specifically.
The thirteen components#
North Atlantic · Baltic · West Med · East Med · West Asian · South Asian · East Asian · Siberian · Amerindian · Oceanian · Red Sea · Northeast African · Sub-Saharan.
Each is an allele-frequency profile learned by clustering the project's reference samples — academic panels plus project volunteers — with ADMIXTURE-style software. The names describe where each profile peaks, as judged by the calculator's author (Davidski, who later built Global25). They are labels on constructs, and the way to see what a label means is to look at which reference populations score high on it in the calculator's own spreadsheet:
- North Atlantic peaks in the Irish, Danish, Norwegian, Orcadian, West Scottish, English and North Dutch references — and, tellingly, in the French Basque. Read it as northwest European affinity, not as "British DNA".
- East Med is highest in Cypriot, Levantine (Jordanian, Palestinian, Lebanese, Samaritan, Druze) and several Jewish references. No single reference in the panel exceeds ~50% of it — it is a component everyone in a wide region carries a slice of, which is exactly why a modest East Med figure in a southern European profile is unremarkable.
- West Med peaks in Sardinians, Basques and southwest French — the populations richest in early European farmer ancestry, though the calculator predates the language to say so.
- West Asian spans the highlands from Anatolia through the Caucasus into Iran. In modern ancient-DNA terms it blends what we now separate into Caucasus hunter-gatherer and Iranian-related ancestries — K13 cannot tell those apart, because the samples that define them were published after it was built.
- Amerindian is anchored by Karitiana, Maya, Pima and North Amerindian references; trace values in Europeans are almost always noise or very old shared Siberian-related ancestry (the same ancestry that makes Ancient North Eurasians relevant to both continents), not a Native American ancestor.
Who K13 suits#
Community experience and the author's own notes agree: K13 resolves European ancestry — especially northern, eastern and central European — better than the smaller Eurogenes calculators, and its 2013 refresh added reference samples relevant to Central and South Asians, Iranians and Turks. If your ancestry is mostly within West Eurasia, K13 and its sibling K15 (two extra components that split the northwest European signal) are the Eurogenes calculators worth your attention. For fine-grained regional affinity, K36 exists, with sharper caveats.
If your ancestry is largely South Asian, HarappaWorld was built for exactly that and does it better; substantially African ancestry is better served by puntDNAL or Dodecad's Africa calculators.
Reading a K13 result without over-reading it#
Take a typical southern European output: North Atlantic 32, West Med 24, East Med 18, West Asian 14, Baltic 8, plus small change. Four readings, in decreasing order of soundness:
- Affinity, plural. The genome leans toward the northwest European, western Mediterranean and eastern Mediterranean profiles at once — true, and consistent with the deep structure of southern Europe (farmer ancestry plus later layers).
- Comparison. Against a sibling or a neighbour run through the same calculator, differences of a few points in the same components are meaningful as relative statements.
- The Oracle's ranking. Oracle and Oracle-4 turn the profile into nearest-reference-population lists, which is the most interpretable output a GEDmatch run produces.
- Not a census. "18% East Med" does not mean an eastern Mediterranean great-grandparent, a date, or a migration. It means 18% of your alleles are best explained by that fixed profile, given the other twelve on offer.
And the standing rule for every calculator: components under a few percent are unconfirmed noise until a method with error bars has looked. K13 attaches no standard errors — nothing on the screen separates a real 2% from a rounding artefact, and nothing in any calculator can.
What has aged#
K13 predates Caucasus hunter-gatherer genomes, the AADR, and a decade of ancient sampling. Its ancient-adjacent components (West Asian above all) merge ancestries the field now separates cleanly. It also carries the calculator effect all the GEDmatch projects inherit: it describes the populations that trained it better than everyone else.
None of that makes it useless — it makes it a 2013 instrument. For the same exploration against current, dated ancient panels, the free Global25 admixture calculators and the closest-populations tool run era by era in your browser; and when a K13 number is about to carry an argument, the method built for arguments is a formal qpAdm model, which is the one tool in this family tree that can tell you no.
Terms used here are defined in the glossary.
