High-resolution stuff

I just emailed this to the authors of High-resolution genomic ancestry reveals mobility in early medieval Europe, a new preprint at bioRxiv [LINK].

I appreciate that Polish population history is not the main focus of your preprint, and also that you're constrained by the lack of relevant and suitably high quality ancient genomes from East-Central and Eastern Europe. However, I must say that your analysis of the Medieval Polish population and resulting conclusions about Polish population history don't reflect reality.

Your Poland_Middle_Ages genomic cluster is made up of just six samples that don't fully represent the genetic complexity of the core population of Medieval Poland.

As a result, you classified PCA0148 as one of the Poland_Middle_Ages outliers, even though this sample isn't an outlier when analyzed within the context of the full set of published Polish Medieval genomes.

Moreover, PCA0148 is very similar to several Polish Viking Age samples that show Scandinavian-specific genome-wide and Y-chromosome haplotypes, and probably likewise shows some Scandinavian-related ancestry.

This is important to note when attempting to recapitulate Polish population history, because it suggests that Scandinavian-related ancestry played a formative role in the shaping of the core Polish Medieval genetic cluster.

Thus, you might be correct when you claim that the six samples in your Poland_Middle_Ages cluster don't show any "detectable" Scandinavian-related ancestry, but this doesn't necessarily mean that this type of ancestry isn't a key part of the post-Iron Age Polish population history.

Below is a self-explanatory Principal Component Analysis (PCA) plot that illustrates my points. Interestingly, Figure 3c in your preprint shows very similar outcomes in regards to the post-Iron Age Polish population history. But the style and scale of your figure makes it difficult to spot the subtle but likely genuine Northwest European-related genetic shifts shown by PCA0148, the Viking context samples and present-day Poles relative to the Poland_Middle_Ages cluster.

However, I'm also skeptical that your Poland_Middle_Ages cluster doesn't carry any detectable or even significant Scandinavian-related ancestry. That's because I suspect that there might be some technical issues with your analysis that are masking this type of ancestry in the Polish samples.

Your top mixture model for the Poland_Middle_Ages cluster is, in all likelihood, an extreme statistical abstraction of reality, rather than a close reflection of it. That's because, due to a combination of historical, geographical and genetic factors, neither Italy.Imperial(I).SG nor Lithuania.IronRoman.SG are realistic formative source populations for the Medieval Polish gene pool.

One of the reasons why you ended up with such a surprising result is probably the lack of suitable samples from East-Central and Eastern Europe, especially those associated with plausibly the earliest Slavic-speaking populations.

It's also possible that basing your mixture model on formal statistics played a key part.

Formal statistics-based mixture models are known to be biased towards outcomes involving mixture sources from the extremes of mixture clines. If your analysis is affected by this problem, then this would help to explain why you characterized the Poland_Middle_Ages cluster as simply a two-way mixture between a Middle Eastern-related group from Imperial Rome and a Baltic population with a very high cut of European hunter-gatherer ancestry.

I do note that on page 6 of your manuscript you consider the possibility that the Southern European-related signal in the Poland_Middle_Ages cluster might only be very distantly related to Italy.Imperial(I).SG, and that it may even have spread across Poland with early Slavic speakers. This is a great point, and I think it should be emphasized and expanded upon, because I suspect that the problem runs deeper than this.

For instance, if the early Slavic ancestors of Poles carried substantially more Southern European-related ancestry than Lithuania.IronRoman.SG, and this ancestry was, say, more Balkan-related than Italian-related, then this might radically change your modeling of the Poland_Middle_Ages cluster. That's because these early Slavs would be positioned in a very different genetic space than Lithuania.IronRoman.SG, which could potentially require a significant signal of Scandinavian-related ancestry to get a robust mixture model.

Finally, it might be useful to consider Isolation-by-Distance as a partial vector for the Italy.Imperial(I).SG-related signal in Medieval Poland.

The full set of published Polish Medieval genomes includes a number of outliers with obvious ancestry from Western Europe and the Balkans. These people probably don't represent any large-scale migrations into Poland, but rather the movements of individuals and small groups. Over time, such small-scale mobility may have had a fairly significant impact on the genetic character of the Polish population.

Update 26/03/2024: I sent another email to Speidel et al., this time in regards to their analysis of present-day Hungarians.

Your preprint also claims that present-day Hungarians are genetically similar to Scythians, and that this is consistent with the arrival of Magyars, Avars and other eastern groups in this part of Europe.

However, present-day Hungarians are overwhelmingly derived from Slavic and German peasants from near Hungary. This is not a controversial claim on my part; it's backed up by historical sources and a wide range of genetic analyses.

Hungarians still show some minor ancestry from Hungarian Conquerors (early Magyars), but this signal only reliably shows up in large surveys of Y-chromosome samples.

The Scythians that you used to model the ancestry of present-day Hungarians are of local, Pannonian origin, and they don't show any eastern nomad ancestry. So they're either acculturated Scythians, or, more likely, wrongly classified as Scythians by archeologists.

And since these so-called Scythians lack eastern nomad ancestry, the similarity between them and present-day Hungarians is not a sign of the impact from Avars, Hungarian Conquerors and the like, but rather a lack of significant input from such groups in present-day Hungarians.

Citation...

Speidel et al., High-resolution genomic ancestry reveals mobility in early medieval Europe, bioRxiv, Posted March 19, 2024, doi: https://doi.org/10.1101/2024.03.15.585102

See also...

Wielbark Goths were overwhelmingly of Scandinavian origin

Viking world open analysis and discussion thread

Global25 and Celtic vs Germanic coordinates for most of the samples from the recent Margaryan et al. Viking paper are now available HERE and HERE, respectively. Look for the VK2020 prefix.

Feel free to put them through their paces and let me know what you find. Below are a couple of examples of what can be done with these coordinates using Vahaduo Global25 Views.

See also...

Viking invasion at bioRxiv

Commoner or elite?

Who were the people of the Nordic Bronze Age?

Avalon vs Valhalla revisited

Pictured below is a new version of my Celtic vs Germanic genetic map. It's based on the same Principal Component Analysis (PCA) as the original (which can be seen here), but more focused on Northwestern Europe and produced with a different program.

To see the interactive online version, navigate to Vahaduo Custom PCA and copy paste the text from here into the empty space under the PCA DATA tab. Then press the PLOT PCA button under the PCA PLOT tab. For more guidance, refer to the screen caps here and here.

To include a wider range of populations in the key, just edit the data accordingly. For instance, to break up the ancient grouping into more specific populations, delete the Ancient: prefix in all of the relevant rows. This is what you should see:

Conversely, you can leave the ancient sample set intact and instead reorder the present-day linguistic groupings into, say, geographic groupings. To achieve this just delete all of the linguistic prefixes, such as Celtic:, Germanic:, and so on. You should end up with a datasheet like this and plot like this.

Of course, you can design your own plot by using any combination of the ancient and present-day individuals and populations that I've already run in this PCA. Their coordinates are listed here. Indeed, if you're in the possession of your own Celtic vs Germanic PCA coordinates, you can add yourself to the plot. And if you're not, see here.

It's also possible to re-process PCA data via the SOURCE tab. But I don't recommend doing this with the Celtic vs Germanic data, which are derived from a fine scale analysis and don't pack much variation. On the other hand, Global25 data are ideal for such re-processing. I made the plots below from subsets of Global25 coordinates available in a zip file here. To see how, refer to the screen caps here and here.

See also...

Modeling your ancestry has never been easier

Getting the most out of the Global25

Modeling genetic ancestry with Davidski: step by step

Celtic vs Germanic Europe

I have a feeling that ancient DNA from post-Bronze Age Northwestern Europe will be coming thick and fast from now on. To get the most out of such data I've designed a new Principal Component Analysis (PCA) that does a better job of separating the Celtic- and Germanic-speaking populations of Europe than my previous efforts of this sort (see here and here). Below are two different versions of the same PCA. The relevant datasheet is available here.

And here's a Discrimination Analysis (LDA) plot based on the 25 principal components. It further differentiates many of the populations along the east > west cline of genetic diversity.

The difference between the Germanic Anglo-Saxons and the Celtic and Roman Britons of what is now eastern England is obvious. The Anglo-Saxons could pass for Scandinavians, while the Celts and Romans both cluster between the Irish and French. This makes good sense, and is exactly what I was looking for. It's also interesting to see the presumably Celtic-speaking Hallstatt samples from Bylany, Czechia, clustering with the Belgians.

Update 14/12/2019: Pictured below is a new version of my Celtic vs Germanic genetic map. It's based on the same Principal Component Analysis (PCA) as the original, but more focused on Northwestern Europe and produced with a different program.

To see the interactive online version, navigate to Vahaduo Custom PCA and copy paste the text from here into the empty space under the PCA DATA tab. Then press the PLOT PCA button under the PCA PLOT tab. For more guidance, refer to the screen caps here and here.

To include a wider range of populations in the key, just edit the data accordingly. For instance, to break up the ancient grouping into more specific populations, delete the Ancient: prefix in all of the relevant rows. This is what you should see:

Conversely, you can leave the ancient sample set intact and instead reorder the present-day linguistic groupings into, say, geographic groupings. To achieve this just delete all of the linguistic prefixes, such as Celtic:, Germanic:, and so on. You should end up with a datasheet like this and plot like this.

Of course, you can design your own plot by using any combination of the ancient and present-day individuals and populations that I've already run in this PCA. Their coordinates are listed here. Indeed, if you're in the possession of your own Celtic vs Germanic PCA coordinates, you can add yourself to the plot. And if you're not, see here.

It's also possible to re-process PCA data via the SOURCE tab. But I don't recommend doing this with the Celtic vs Germanic data, which are derived from a fine scale analysis and don't pack much variation. On the other hand, Global25 data are ideal for such re-processing. I made the plots below from subsets of Global25 coordinates available in a zip file here. To see how, refer to the screen caps here and here.

See also...

Modeling your ancestry has never been easier

Getting the most out of the Global25

Modeling genetic ancestry with Davidski: step by step