Ashkenazi Jewish DNA: the ancient origins behind the label

"Ashkenazi Jewish: 99.8%." It is one of the most confidently reported results in consumer genetics, and one of the least informative about origins. This post explains what that label measures, what the ancient and medieval genomes have established about where Ashkenazi ancestry comes from, and what a formal qpAdm model does with an Ashkenazi genome. It is part of a series introduced in Jewish ancestry and ancient DNA.

As with every post in the series: ancestry is not identity. Nothing here is a statement about who is Jewish.

Why the label is so confident#

Ashkenazi Jews descend from a population that lived along the Rhine in the early Middle Ages, spread east into Poland and the Russian Empire, and grew from a very small number of founders to several million people. The genetic signature of that history is a bottleneck: a period when the effective population was a few hundred people, dated to roughly 600 to 800 years ago (Carmi et al. 2014). After a bottleneck, every member of the population shares long stretches of genome inherited from the same few ancestors. The result is that Ashkenazi genomes resemble one another far more than the genomes of, say, Germans resemble other Germans.

Consumer panels exploit this. Because Ashkenazi genomes form such a tight cluster, the panels treat "Ashkenazi Jewish" as a reference category, and a person with four Ashkenazi grandparents lands inside it almost perfectly. The result is accurate as a statement of resemblance and empty as a statement of origin: the reference group is itself a mixture, and the label cannot see inside it.

What the ancestry actually is#

The genome-wide studies of living populations settled the broad picture some years ago. Ashkenazi Jews cluster with the other Jewish communities of the Mediterranean and Middle East and with Levantine populations, not with Germans or Poles (Behar et al. 2010; Atzmon et al. 2010). At the same time they carry a large European component, which modelling places mostly in southern Europe, Italy in particular, with a smaller share from eastern Europe. Xue and colleagues estimated the European fraction at roughly half, with the admixture dated to the medieval period, consistent with a Levantine-derived community forming in Italy, mixing there, and then moving north of the Alps (Xue et al. 2017).

In 2022 the first substantial set of medieval Ashkenazi genomes was published: 33 individuals from the fourteenth-century Jewish cemetery of Erfurt in Germany, sequenced with the consent of the local Jewish community (Waldman et al. 2022). They are the single most important data point for this question, for three reasons.

1. The modern ancestry profile was already in place. The Erfurt individuals are genetically very close to present-day Ashkenazi Jews. The Levantine and southern European mixture had happened before the fourteenth century, not after.
2. The founder event predates them. The Erfurt genomes already show the elevated runs of homozygosity and the specific disease alleles that mark the modern bottleneck. Whatever the founding population was, it was already small by 1350.
3. There were two groups. Part of the Erfurt community carried more eastern European ancestry than the rest, and the two groups were only partly mixed. Modern Ashkenazi Jews look like a blend of the two, which suggests the eastern component entered through a distinct medieval population rather than through slow, continuous admixture.

Those 33 individuals, labelled Germany_Medieval_Jewish in the Allen Ancient DNA Resource, are in the reference panel we merge your file with. An analyst can use them; you can use them yourself in the Model Lab.

The Levantine half#

The Middle Eastern component of Ashkenazi ancestry is, as far as ancient DNA can tell, the same Bronze Age Canaanite ancestry that underlies every Levantine population. Genomes from Megiddo, Hazor, Sidon and their neighbours describe a population made of the local Levantine Neolithic lineage and a substantial Iranian- or Caucasus-related layer (Haber et al. 2017; Agranat-Tamir et al. 2020), and when Ashkenazi Jews are modelled against those genomes a Canaanite-related population takes a large share, with the rest supplied by a European source. That is why our catalog now carries a dedicated Canaanite source, described in its own post.

One consequence is worth spelling out. A consumer result that reads "99% Ashkenazi, 0% Middle Eastern" is not a finding that there is no Middle Eastern ancestry; it is a finding that the panel put the Middle Eastern ancestry inside the Ashkenazi category before it started counting.

What a qpAdm model does with an Ashkenazi genome#

qpAdm asks a question with a yes-or-no answer built in: can this genome be written as a mixture of these ancient sources, to within statistical noise, tested against these outgroups? It returns a weight, a standard error and a Z-score for every source, and a p-value for the whole model that can reject it. The method is set out in Understanding qpAdm; how to read the numbers is in the reading guide.

For an Ashkenazi genome the two eras of the report usually resolve like this.

Hunter-Gatherer and Neolithic Farmer era. The deep streams of western Eurasia. An Ashkenazi genome draws most of its weight from the Anatolian and Iranian Neolithic farmer sources and the Natufian source (the three ingredients of the Levant), with a measurable Western Steppe Herder and Western Hunter-Gatherer share that marks the European admixture. The steppe and hunter-gatherer weights are the deep-time signature of the Italian and eastern European half; in a Mizrahi genome they are usually indistinguishable from zero.

Classical Antiquity era. The sources are closer to the events. The model the analyst starts from is Canaanite beside Imperial Italy, which is the ancient population of Rome and central Italy in the first centuries AD. The questions the search then settles are whether a Germanic or Early Slavic source earns a place (the eastern component the Erfurt genomes point to), whether the Aegean source fits better than the Italian one, and whether the Levantine share prefers Canaanite to the later Phoenician or Eastern Mediterranean references. Every one of those comparisons is a separate model with its own p-value, and the report keeps the one that survives.

The report also shows something the consumer label cannot: the standard error on the Levantine weight, which tells you how cleanly the model could separate Canaanite from Italian ancestry in your particular file. That number is set by your file's coverage, which you can check for free in the file check before you order.

The bottleneck, from the model's side#

The founder event does one awkward thing to a qpAdm model. Genetic drift after a bottleneck pushes a population a little away from every ancient source at once, in a direction no source can supply. In a well-composed model the effect is absorbed by the noise and the p-value stays comfortable. In a poorly composed one it shows up as a source with a weight that is too large for its history, or as a model that rejects for no visible reason.

This is one of the reasons we do not rotate models automatically. Running every combination of sources until one passes would, for an Ashkenazi genome, reliably find a passing model that is wrong. Every published model is composed by one person who knows what the Erfurt genomes say and what a drifted population does to an f-statistic, and who tries the alternatives before settling. The deeper tiers buy more of that search; the bar is the same at every tier. The buyer's guide for Jewish genomes goes through the tiers.

What the model does not say#

• It does not say any individual from Megiddo, Rome or Erfurt was your ancestor. A source is a reference population the model tests against, never a family.
• It does not identify a tribe, a lineage, a Cohen or Levite status, or a historical person. Those are Y-chromosome questions at best, and the Y-DNA haplogroup tools are the right instrument for them.
• It does not measure Jewishness. A convert's genome will model as their ancestry, and that is not a comment on anything.

Related reading#

• Jewish ancestry and ancient DNA: the series introduction
• The Canaanite source population
• Sephardic Jewish DNA and Mizrahi Jewish DNA, for the communities the Ashkenazi profile is most often compared with
• What a qpAdm ancestry test is
• The full source population directory

References#

• Agranat-Tamir, L. et al. (2020). The genomic history of the Bronze Age Southern Levant. Cell, 181(5), 1146–1157.
• Atzmon, G. et al. (2010). Abraham's children in the genome era. American Journal of Human Genetics, 86(6), 850–859.
• Behar, D. M. et al. (2010). The genome-wide structure of the Jewish people. Nature, 466, 238–242.
• Carmi, S. et al. (2014). Sequencing an Ashkenazi reference panel supports population-targeted personal genomics and illuminates Jewish and European origins. Nature Communications, 5, 4835.
• Haber, M. et al. (2017). Continuity and admixture in the last five millennia of Levantine history from ancient Canaanite and present-day Lebanese genome sequences. American Journal of Human Genetics, 101(2), 274–282.
• Waldman, S. et al. (2022). Genome-wide data from medieval German Jews show that the Ashkenazi founder event pre-dated the 14th century. Cell, 185(25), 4703–4716.
• Xue, J. et al. (2017). The time and place of European admixture in Ashkenazi Jewish history. PLoS Genetics, 13(4), e1006644.