search this blog

Tuesday, September 29, 2020

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?

313 comments:

«Oldest   ‹Older   201 – 313 of 313
ambron said...

Matt, of course, because this is the Polish median, but Poles are characterized by a wide range of genetic variation.

David, of course, but this is probably a problem for all models of modern populations.

Matt said...

@ambron, yes more in South, less in North. Isolation by distance

We'll find out as more data come through. I suspect that it will be closer to truth that something like your model happened, with geneflow from north and south (Baltic+Carpathian Basin), than something like "Poland ancestry basically a EBA descendent of GAC+Corded Ware plus big drift; then that population was actually ancestral to Baltic BA".

zardos said...

Compare with Unetice and following cultures, there was an influx from both the East and the relative South (Carpathian-Ukrainian region). I think a lot will be related to Sintashta and chariots plus later advanced horsemanship, as well as the Iron Age transition.
Even Unetice was not just local Corded Ware plus GAC to start from.

Simon_W said...

@Rob
"Lastly, some have advocated for a rather late Germanicization of Scandinavia (Osten-Dahl, Udolph), more Common Era than anything Bronze Age, and i;d tend to agree"

Absolutely; there was not Germanicization during the Bronze Age, because what defines Germanic languages is the Germanic sound shift (Grimm's law), and it has been established by linguistic means that this occured late, after 500 BC.

Simon_W said...

@Gabriel
"But what I was saying is that I don’t think West Germanics had Halberstadt_IA ancestry, unlike what some users said here."

When I'm modelling West Germanics (Baiuvari) with Jutland_IA and Halberstadt_LBA, which geographically and archaeologically makes sense, then they score lots of Halberstadt_LBA ancestry. That the yDNA of Halberstadt_LBA doesn't match West Germanics doesn't contradict this, because as a substratum, its yDNA might have got largely replaced. However, as Samuel Andrews pointed out, Jutland_IA is low coverage, so it's probably wrong to use him. And in fact, when I model West Germanics (Baiuvari) with Swedish_IA and Halberstadt_LBA, then they come out as predominantly Swedish_IA, suggesting higher affinity with Northern Europe than with the LBA North German plain.

Simon_W said...

@weure
" in the Celtic vs Germanic PCA, Jutland IA almost is in the Celtic kind of cluster, may be low coverage, but weren't they Cimbri like? (Germanic/ Celtic intermediate?).

@WeightofAudio
"It will be interesting to see if any of this sheds light on the Cimbri-“question”...I don’t think they were straight up La Tene celts, but Perhaps represent some remnants of something that Norse and Ingvaeonics didn’t completely level out."

I don't believe there is anything Celtic about the Cimbri and the Teutones. Their names suggest that they didn't complete the Germanic soundshift. Which makes sense, given the rather late occurence of that shift, see my reply to Rob. In that case, they were simply pre-Germanic, but still descendants of the Nordic Bronze Age, and thus similar to other North Europeans.

Norfern-Ostrobothnian said...

Off topic but do you think that the reason Mari (and Chuvash) are hard to model in a good way is because of genuine drift or how G25 is structured? It would make sense that the Mari would have drifted in isolation, but I am not quite sure if it's correct.

Here's how Udmurts, Chuvash and Mari act on ancient individual samples:

Target: Udmurt
Distance: 1.6045% / 0.01604516


Target: Chuvash
Distance: 4.9587% / 0.04958733

Target: Mari
Distance: 7.8835% / 0.07883508

If Mari is put to the sources


Target: Udmurt
Distance: 0.4415% / 0.00441475

Target: Chuvash
Distance: 0.3831% / 0.00383083

So maybe this is an indication of a bottleneck among the Mari which further radiated to the Udmurt and Chuvash peoples? It wouldn't be that far fetched considering Udmurts and Chuvash have lived in the Mari periphery since medieval times, and there's no doubting that Mari participated in the ethnogenesis of the Chuvash.

Samuel Andrews said...

@ambron,

There's little genetic variation in Poland, Ukraine, Russia. Compared to Germany, France, Italy even the land size of the countries are the roughly the same.

This is because Poland, Ukraine, Russia derive mostly from single ancestral population, proto-Slavs.

It will turn out more complicated with ancient DNA. But, the overall picture is Eastern Europe's population history was hugely changed in Middle Ages. So much that Slovenians on Meditereaen sea are probably 70% Slavic, so very related to Russians in Northern end of Europe.

Matt said...

@ambron, just one further point about a hypothesis of a Third Population that is somewhat intermediate in Steppe-HG-Anatolian proportions between Baltic_BA and Carpathian_CA-EBA, and has a high enough level of the "Balto-Slavic" drift to explain both Baltic_BA and present day Polish folks, with some different regional admixture for each. I can simulate such a population in PAST3 (it's not too hard to "cook up"), and the roughly best fitting ghost coordinates (without too much work) are here: https://pastebin.com/qfA9c3Hf
(Position of this "ghost" on G25 Vahaduo Views: https://imgur.com/a/m3DQTJv)

While it does roughly work in general, and works well for Polish folks being about 50% derived from this population and 50% from generalized predecessor populations that are about 50:50 GAC:earlyCordedWare, the problem with the ghost is that to then explain the Baltic_BA Latvian and Estonian populations then, in vahaduo models, it seems to need to mix with essentially unadmixed Narva+EHG hunter-gatherers. That basically seems pretty less than likely to me, unless we think of the Corded Ware colonization of Baltic as having essentially completely failed and HG to have re-emerged.

If we try to compensate to move around to the ghost to allow for some continuation of preceding Early Corded Ware ancestry in the Baltic, then we would find that the ghost needs to become quite a bit more heavily Anatolian in balance to work. (If you went the opposite way and made the ghost have an steppe:HG:Anatolian ratio like Sintashta, you'd get the opposite problem and it would suggest that mixture of the ghost with an Anatolian+HG population would be required to explain Baltic_BA, and the ghost would need to mix with populations with quite high Anatolian:Steppe ratio to explain the present Polish folks too).

So that's the basis I'd find it less likely, but who knows - ancient dna can be very surprising.

@Sam, on the topic using the G25 data, here are the Maximum, Median and average Euclidean distances between G25 modern samples within population (scaled data because unscaled data are not meaningful on this one):

All Polish: Max: 0.069, Median: 0.04, Average: 0.04;
German Excluding East German: Max: 0.085, Median: 0.05, Average: 0.05;
French, Excluding Corsica, French South, Breton: 0.079, Median: 0.044, Average: 0.044;
All Irish: Max: 0.065, Median: 0.038, Average: 0.038;
All English: Max: 0.048, Median: 0.031, Average: 0.030;
All Dutch: Max: 0.079, Median: 0.049, Average: 0.049;
All Czech: Max: 0.52, Median: 0.037, Average: 0.036
All Spain, excluding Pais Vasco and Canary Islands: Max: 0.063, Median: 0.035, Average: 0.036.

There is some effect where larger nations have higher internal distances and smaller nations lower (e.g. England, Ireland, Czech Republic), though note Netherlands is kind of an exception to this tendency... (Switzerland would be another). Also nations with few internal barriers, for'ex Spain for instance in particular seems internally more "well mixed" than Italy (despite some rugged terrain).

Davidski said...

@Draft Dozen

Old school physical anthropology/typology is against the rules at this blog.

You have to base your arguments on genetic data.

Gabriel said...

@Simon_W

Yeah, I just doubt that there was a back-migration of Germanic speakers with Halberstadt_IA admixture to Denmark, rather than higher GAC ancestry, given Danes have a West Germanic influenced profile.

@zardos

I doubt Sintashta had any impact, genetically speaking, on proto-Balto-Slavs or even Tumulus, given you’d see something like R1a-Z2124 there or even on later samples, and you don’t see that.

Samuel Andrews said...

@ambron,

South Slavs have significant recent Northeast European ancestry. Slovenians, Croats, Bosnians are 70% Polish-like and 30% ancient Balkans. This comes from the Slavic invasions.

There's definitely a lot of Slavic ancestry in Southeast Europe. It is reasonable to say the same is true for East and West Slavs.

Davidski said...

@Matt

There's no need to assume that there was any continuation of early Corded Ware ancestry into the BA in the East Baltic, nor that there was any continuation of local Corded Ware/GAC ancestry in Poland into the present.

What could have happened, and probably did happen, are massive, almost total population replacements starting during the late Corded Ware period.

See that's probably why Y-chromosome clades associated with modern Balto-Slavs only start appearing in and near the Baltic from the very late Corded Ware period, and that's also when Balto-Slavic specific drift makes its appearance.

Also, keep in mind that in turn Baltic BA was probably largely replaced by another migration wave that gave rise to modern Balts, again probably from somewhere in the southeast.

ambron said...

Matt, this is a fascinating game of genetics. Let's keep having fun while waiting for new samples. I'm looking forward to the end result!

Davidski said...

That's the problem with Matt's take on things here. It's just a game of genetics.

He doesn't even seem to know that East Baltic languages are recent (post BA) arrivals in the East Baltic.

Matt said...

I don't think that really solves much; if both 100% replacement from same source, then doesn't explain why they differ (why BalticBA both much more intense HG, without admix, and most intense position on B-S cline?), then if invoke local substrate then back at situation from above post.

Not sure how much languages do or don't tell about story; people argue for linguistic turnovers without much match to genetics all the time...

Davidski said...

@Matt

What I'm saying is that Baltic BA didn't form in the East Baltic, but came from somewhere else where there were populations with high levels of HG ancestry and Balto-Slavic drift.

And I'm saying that this is also where Balts came from, but somewhat later.

Since Slavs are closely related to Balts, but have much lower levels of HG ancestry, then they probably formed just south of this zone.

I think you'll find that this actually fits archeology and linguistics rather well.

ambron said...

Sam, in Poland there is an autosomal border on the Vistula. Eastern Poles (especially towards the north) are identical to Belarusians, northern Ukrainians and western Russians. This, of course, is the result of a shared ancestral population. However, they were not Slavs. They were the Balts. The hydronyms of this area are Baltic. Contemporary Balts were to come from this area to the Baltic Sea, as Dawid wrote a moment ago. Slavisation of this area began only in the Middle Ages, and has not ended until today (Podlasie).

The proto-Slavic genome is an averaged Slovak genome, which also includes some Polish, Czech, Hungarian, Croatian, Sorbian and Slovenian genomes. He lies exactly in the center of the Slavic genetic variation. Such a genome, by adding an admixture to the Baltic genome, gives the East Slavic genome, contributing to German genome - gives Czech and West Polish genome, and contributing to Greek genome - gives South Slavic genome.

According to the Nestor, Poles (Lechites) are Croats, Serbs and Slovenes (Carantani) who came from the Danube. On the other hand, the Russians are Poles who moved to the east - Wiatycze, Radymicze. Therefore, the genetic link of the Russians with the Slovenes is obvious.

ambron said...

David, linguists argue that the Slavs could not have arisen near the Balts and still existed in their vicinity for millennia. If this were the case, the differences in the Baltic and Slavic languages would be, for example, the same as in the group of Romance languages. The difference between the Baltic and the Slavic groups is so diametrical that the languages are completely incomprehensible to each other.

Davidski said...

@ambron

I'm operating here in the context of mainstream linguistic theories, which posit that Balts and Slavs are very close linguistic relatives.

This obviously fits well with genetics too.

ambron said...

David, consent, but they are linguistic relatives dating back 3,500, not 1,500 years.

Gabriel said...

@ambron

I doubt Ukrainians have much Baltic admixture, considering they are among the peoples geographically closest to the proto-Slavic homeland, and so are Eastern Poles, not necessarily the Northeastern ones.

Also Hungarians have German and Balkan admixture. And it’s wrong to say South Slavs are admixed with a Greek-like population, considering modern Greeks are partly Slavic.

Rob said...

Slavic is no more distant than west and east Baltic are to each other

ambron said...

Gabriel, I simplified. I meant the cyproid genotype.

Gabriel said...

@ambron

There is nothing Cypriot-like about the non-Slavic ancestry in South Slavs. They were like Balkan Iron Age people, maybe with some recent Near Eastern. But not like Cypriots.

Samuel Andrews said...

@Gaberial,

The non-Slavic stuff in East Balkans including Serbia is intermediate between Iron age Bulgaria and Bronze age Central-West Anatolia. It has Near East admix but it doesn't resemble Cyprus who is basically North Levant.

The non-Slavc ancestry in Slovenia, Croatia, Bosnia is different. It is Bronze age West Balkans-like.

Samuel Andrews said...

@Gaberial,

I guess I repeated what you said. Sometimes I assume people dis agree.

Samuel Andrews said...

@Simon_W,

Yeah, figuring out Germanic origins is confusing with current ancient DNA, considering non-German substrate maybe basically identical to Proto-Germanic population.

Davidski said...

@Samuel

Disagree is one word. You keep writing it as two.

ambron said...

Gabriel, Samuel, Hellenthal calls the non-Slavic admixture in the South Slavs and the Greeks - Cypriot-like. And this genotype I meant, regardless of its name.

Davidski said...

The Hellenthal paper was a pile of crap.

weure said...

@Chevalier de Balibari, I can't see no genetic connection between Jastorf/Elb-Germanics and the Franks that were rooted just above the Rhine.

Rob said...

There’s probably no unitary “pre-Slavic” people in the Balkans on the event of the Slavic migration. It depends which region or even individual site one looks at

ambron said...

Gabriel, northern Ukrainians look the same genetically as Belarusians and Northeast Poles, i.e. the Balts culturally assimilated by the Slavs. South-west Ukrainians are genetically similar to Slovaks and southern Poles. And somewhere close you have to look for the Proto-Slavic homeland.

Gabriel said...

@weure

Jastorf wasn’t necessarily just Elbe Germanics. Nobody knows yet which culture Franks came from but just because Franks were mostly around the Rhine doesn’t mean they couldn’t have originated in the western areas of the Jastorf culture.

Ioannis Gavras said...

@ Weure

I think Franks are more related with Northwestblock related culture. A ghost BB like population contribute in their ethnogenesis. They definitely didn't arrived in modern France as 100% scandinavian/Nordic like.

So, Slavs are Balts without the strong WHG shift? It was the Baltic homeland a hideway for WHG folks?

Rob said...

Franks are a later confederation of Chattuari, Brucetii, Chamavii, etc
Theyre Rhine-Weser Germans, see my above

Ioannis Gavras said...

@ Samuel


South Slavs have none of such admixture as you described above. They are mostly IA Balkans+Slav and maybe some limited germanoceltic.If you see some post-Roman admixture it depends individual because many of them have also Vlach roots or Vlach related ancestry and Vlachs are known to be mixing with Roman soldiers and generals.We are also dont avoiding the fact that some post Neolithic west Asian admixture to arrived during BA or even IA from Anatolia to Balkans but not crazy things to describe South slavs as partly Cypriot like. And ofc modern Cypriots are not exactly a North Levant group. They are between South Italians/Greeks islanders, the have a strong BA Anatolian admixture(Hattic, Hittite like) witch btw IMO covers most of their ancestry and the rest is Lebanese-Druze like but that depends as I mention individual and region. Eastern Cypriots are more Phoenician like,but again overall Cypriots are neither mycaenean or Phoenician but mostly a BA Anatolian subgroup. And this has to do because the folks there, before the arrival of Greeks and Phoenicians were related to people with BA Anatolia origins/roots.

Ioannis Gavras said...

@ Rob

We know these things.What we don't know know is west germanics genetic make up.I really doubt Rhine-weser Germans were That different from Jastford(Elbe Germans).Without samples/DNA from Northwest block and Jastford culture we cannot be sure for anything.The specific area of northeast Netherlands and Northwest Germany was probably a hotspot of BB related folks and these folks probably assilmated into those germanic tribes you mention while others might even assilimated to the southern Celtogaulish tribes.The whole point here is that western germanics were not exactly nordic/Scandinavian like. They were very close to them but not exactly the same and I think even by ydna kind of view R1a/I1 etc would have been more rare lineages compared to what we seen in modern Nordic people. But I understand that is also hard and difficult to separate BB autosomal from Nordic. It is obvious I think that some Germanic people/tribes were less WHG admixed compared to other tribes.So, without samples/DNA from these 2 cultures we just theorizing.

weure said...

The clue for the Bell Beaker steephead. They had a thinner vault because of brain improvement. The thinner vault caused brahycephaly/ flat occiput.....

It's a real possibility that within the Single Grave Culture/ Protruding Foot Beaker this adaption in the form of 'Dinarization' took place. The Steppe herders/pastoralist, as robust dolicho's (=thicker vaulted) settled down in the Rhine Delta/ North Sea area adapted to the environment and took the Bell Beaker package and so a different, more social complex, life style. This meant a need of a new 'improvement efficiency of the brain'. Their close knit endogamous character meant a push in the spread of this steephead phenotype.

Discus on:
https://anthrogenica.com/showthread.php?19321-Dinarization-and-sexual-selection&p=707327&viewfull=1#post707327

Ric Hern said...

Who were the people of the Elp Culture ?

weure said...

@Ric the Elp ares is this area:
https://en.wikipedia.org/wiki/Elp_culture#/media/File:BronzAgeElp.png
In fact the whole Southern North Sea area.
Common gene pool: is TRB (West), Single Grave Culture and Bell Beakers.
The Elp-Culture is the Bronze Age culture that started with the Sögel-Wohlde period with very outspoken warrior graves. The Sögel-Wohlde culture gave a push to the development of the Nordic Bronze Age. And was expansive. During the Bronze Age the Valsømagle culture from the Danish Isles and Scania was more prevailing. The Elp culture is marked by the specific long houses for example.

Draft Dozen said...

@ambron

" On the other hand, the Russians are Poles who moved to the east - Wiatycze, Radymicze"

Under the "Lyakhs", the chronicler could mean the lands that in his time (Nestor) were under the rule of the "Lyakhs" , it could be the lands of Great Moravia, for example. Archeology does not mark traces of migration of a large tribe or group from Poland to Russia. According to modern views, Vyatichi (as well as Radimichi) were carriers of one of the branches of the romenskaya archaeological culture. The origins of the romenskaya culture interpreted differently by different specialists. For example, as a result of A.V. Grigoriev's review can be noted, that all the small differences between the early romenskaya culture and the "common slavic" culture to some extent find their prototypes in the materials of monuments of the VII - 1st half of the VIII century, located in the Lower Danube region. By the way, in this area, according to Theophanes report, in the context of the same period, there was a large slavic union of the Severs
https://pl.wikipedia.org/wiki/Siewiercy
Bulgarian-Byzantine wars, internal troubles (for example, the Telets revolt) could cause an outflow of the Severs to the northeast. And what is interesting, is that, not only Vyatichi and Radimichi, but also Severians belonged to the romenskaya culture.

"northern Ukrainians look the same genetically as Belarusians"
South Belarusians

"South-west Ukrainians are genetically similar to Slovaks and southern Poles."

I read, that the population of Chernivtsi region is close to Moldovans. And according to data of Ukrainian genetic Utevskaya Olga, Ukrainians quite close the Balkans. Her words "there is a common genetic space - the area covering Poles, Belarusians and Russians - and Ukraine for some reason, falls out of it; probably, this steppe space (Ukraine) is geographically something else, some layers are different there"
Genographic maps
Poles
https://ic.pics.livejournal.com/1981dn/50916219/128666/128666_800.jpg
Russians
https://ic.pics.livejournal.com/1981dn/50916219/129011/129011_800.jpg
Belarusians
https://ic.pics.livejournal.com/1981dn/50916219/129076/129076_800.jpg
Ukrainians
https://ic.pics.livejournal.com/1981dn/50916219/129297/129297_800.jpg
Map of genetic distances from Hungarians and Moldovans
https://ic.pics.livejournal.com/1981dn/50916219/129697/129697_800.jpg
https://ic.pics.livejournal.com/1981dn/50916219/129911/129911_800.jpg

Davidski said...

@Draft Dozen

Ukrainians aren't especially similar to Balkan populations. You can have a look here.

https://vahaduo.github.io/g25views/#Europe1

The source you are using is wrong.

Ioannis Gavras said...

Does anyone know anything about House urns culture?

Rob said...

@ Weure

LOL Im sure BB were smart, but I don't think their flat occiput was cause by a bulging brain. Its just an epigenetic feature.

@ Chevalier

''The whole point here is that western germanics were not exactly nordic/Scandinavian like.''

Sure, agree. One wonders how much continuity there was with preceding LBA in the Rhine region. Their formation does seem condition by encroaching Romans upon the Rhine border. But their material culture is fairly unassiming, - simple shifting farmsteads gradually enlarging toward propper villages, cremation burials are the rule, unfortnately for aDNA.

Ioannis Gavras said...

@ Rob


It is getting annoying how slow European genetics are going.I have a goal to learn pretty much everything about European-West Eurasian genetics before i die.I think the issue's of Germanic and Slavic DNA are going to be hard and delayed.Greek and Anatolian genetics are a similar case.These 4 are my primary goals and i will die happy.

1=Anatolian
2=Greek
3=Slavic-Germanic
4=BB and where they lasted

Gabriel said...

@weure

It is unlikely that the Nordic Bronze Age was triggered by a significant migration from outside Scandinavia. There is no genetic evidence for such event, and R1b-U106 already appears in samples with local isotopes and dates too early for a foreign arrival.

Romulus said...

@Gabriel

Really? Name some U106 from older than the Nordic Bronze age in Scandinavia

Davidski said...

Well, duh...

Sweden_LN RISE98 2275-2032 calBCE

And there's a nice little story that goes along wit this one.

https://eurogenes.blogspot.com/2019/09/commoner-or-elite.html

Samuel Andrews said...

And there's also R1b U106 in Netherlands 1700bc.

R1b U106 is more common in Netherlands, Germany, England than in Scandinavia. Which suggests there is some non-Germanic R1b U106.

Davidski said...

Yeah, there's also U106 in upcoming Dutch Bell Beaker samples.

Gabriel said...

@Samuel Andrews

Sure, but most R1b-U106 is of Germanic origin, and there is no equal distribution of R1b-U106 in Scandinavia either. West Germanics were probably higher in R1b-U106 than other Germanic tribes, and that contributes to the modern distribution of R1b-U106.

Samuel Andrews said...

@Gaberial,

Probably but a deep analysis of R1b U106 is needed to be sure. The problem, is non-Germanic substrates might have been close cousins of proto-Germanics.

R1a Z284 is probably a non-germanic substrate in Norway. 25% Norwegians carry R1a Z284. It is not due to a recent founder effect. A much lower percentage of Swedes belong to R1a Z284 and R1a Z284 is more rare in continental Europe. This supports the idea of non-Germanic substrate in Norway.

In Viking samples, R1a Z284 is only found in samples from Norway.

weure said...

@Rob more efficient brains meant thinner vault. Thinner vault opened the way for a flattened occiput (can go together). It was an adaption of the SGC pastoralist/herders (very robust doliocephaly) to settlement and a more complex culture (BB package with archery etc). So an evolutionary adaptation that leant them certain advantages, signed in the phenotype by 'Dinarization'. The endogamous behavior meant that within a few generations it was their 'marker'.

We see here the adaption by Gerhardt (1976):
https://www.mupload.nl/img/5vsbs1as.11.03.png

Gabriel said...

@Samuel Andrews

Yeah, but that doesn’t change the fact that if West Germanics had higher U106, whether or not they were closer to proto-Germanics, then U106 could still be mostly late Germanic in Northwest Europe.

weure said...

@Rob all based on:
'My research is based on the theory that skulls with a more spherical shape are stronger, and thus are not required to compensate with a thick cranial vault (Demes 1987, Nawrocki 1991, Lieberman 1996).'
https://ufdcimages.uflib.ufl.edu/AA/00/06/00/40/00001/bwalter-Undergrad_Honors_Thesis.pdf

ambron said...

Dozen

Nestor wrote that "Lyakhs" called themselves these Danubian Slavs who settled on the Vistula. Vyatichi and Radimichi were to come from these Lyakhs. Nestor does not mention the Severians here, but goes on to say that the three tribes shared customs.

Rob said...

@ weure
Okay point taken; thanks

Matt said...

OT: @Vahaduo, if you're watching this thread, quick question at you:

New preprint by Florian Prive (https://www.biorxiv.org/content/10.1101/2020.10.06.328203v1) discusses a result where squared euclidean distance on PCA approximates to a linear relationship with Fst distance. This is true in for distances frm the G25 PCA: https://imgur.com/a/hi3OQSc

(Some caveat that this is not true for populations whose distinct genetic variation not well encompassed by G25, like Onge, plus founder effects not covered, etc, but true in the main for the modern comparison set I tried. Correlation obviously would be closer if no populations are projected and all PC dimensions are used, as in Prive's paper, rather than G25 where we project some and just keep top 25 PCs. Relationship with Fst is probably also disturbed in ancient dna with high levels of damage / artefacts which projecting onto PCA defined by others tends to remove, but we would expect that.)

The simple euclidean distance is less linear with Fst: https://imgur.com/a/XE3piL4 (though the correlation is comparable).

(Here are some other figures to illustrate the relationship between Fst and PCA distance using G25: https://imgur.com/a/6ksV7Ag).

This all being the case, how simple would it be to offer a mode in Vahaduo Admixture to minimize squared distance rather than the simple euclidean distance?

Assuming we're doing something *like* trying to minimize Fst by linear methods, then using a transform that fits our proxy distance to the closest linear relationship seems like it potentially makes sense.

Vladimir said...

New article "On the genetic composition of the Sarmatian population of the Lower Volga region (data from paleogenetics)" by scientists of the Siberian branch of the Russian Academy of Sciences. The subject of the study was the single-marker MtDNA and Y. In total, 36 out of 42 structural variants and 54 out of 62 mtDNA samples studied (87.1% of the series) belong to the West Eurasian cluster. It is necessary to note the high phylogenetic diversity of haplogroup H. At this stage of the study, we identified variants of the H6b2, H8, and H20a subgroups among Sarmatian samples. Thus, the variants of haplogroup H in the Sarmatian gene pool are phylogenetically diverse. The second most frequent Western Eurasian cluster in the Sarmatian gene pool (17.7 %) is haplogroup T (11 samples, 4 structural variants). Haplogroups U4, J, and U5a are represented with frequencies of about 10% and 3-4 structural Variants each. The minor East Eurasian component of the mtDNA gene pool of the Sarmatian population of the Lower Volga region is represented by variants of haplogroups A, C, F1b, G2a, the total frequency of which is only 12.8% of the total series studied. As for Y, of the 11 individuals studied, 8 had haplogroup R1a1a, 2 individuals had haplogroup R1b, and one individual had haplogroup N2 (Y6503). No East Asian haplogroups were found. It is interesting, of course, to find N2 (N-Y6503) among the Sarmatians. This may explain the current presence of this haplogroup in the Balkans.

Unfortunately, the article is only in Russian.
https://hfrir.jvolsu.com/index.php/en/component/attachments/download/2297

Draft Dozen said...

@ambron

There are many examples where Nestor's testimonies differs from the facts, so, he wrote that the Severians descended from the Krivichs ("It is there that the Krivichs dwell, and from them are the Severians sprung"), but is not true, and that the Drevlians from the Polan/Polinians, is also not true, etc. As for him: "There were in fact among the Lyakhs two brothers, one named Radim and the other Vyatko. Radim settled on the Sozh', where the people are known as Radimichi, and Vyatko with his family settled on the Oka. The people there were named Vyatichi after him." As you can see, according to legend, two brothers were from the "Lyakhs", and that the locals were called by their names.
In fact, I wrote, that the Radimichi and Vyatichi descended from the Severians, all three tribes were carriers of the romenskaya culture, try to understand, the bearers of the romenskaya culture or its variant/branch in the Oka region were newcomers in this area, they came from the Dnieper region, and the Severians were, probably, immigrants from the Lower Danube, descendants of the Severs that I mentioned above.

Matt said...

@vahaduo just as a followup on my previous comment, here is the same plots and correlations without the outlying populations which have highest major Fst increase that seem unmodelled by G25 (Onge, Ket, Kalash, Saami, Igorot): https://imgur.com/a/Q3esVhg

The correlation is really tight between G25 distance^2 and Fst, without those outliers... (improves for G25 raw distance too).

(last plots in above show population labels - how far a population in comparisons is on X / Y generally is just a measure of how much it is close to the centre of populations used in this set of comparisons and not anything deeper than that, but the offsets have some information in them).

Pastebin of the datasheet used for above plots: https://pastebin.com/36x03kTV

ambron said...

Dozen.

These are the details. Overall, I meant something else ... Jordanes and Nestor locate the Slavic homeland between the Danube and the Vistula, and genes seem to confirm it. The Slavic homeland in the middle Dnieper is just fairy tales told by some archaeologists.

vahaduo said...

@Matt

Thanks for the link!

This all being the case, how simple would it be to offer a mode in Vahaduo Admixture to minimize squared distance rather than the simple euclidean distance?

This algorithm already does that.

It only checks which distance is lower, and it doesn't matter how big the difference is. If A > B is true then A^2 > B^2 is also true.

To calculate the Euclidean distance you need to calculate the square root at the end, and omitting this step gives you the squared Euclidean distance. Since in case of this algorithm there is no difference between them, the squared Euclidean distance is used. Thanks to this the algorithm doesn't have to calculate the square root 500 times per cycle, which is one of the many optimizations that make it so fast.

Matt said...

Aha, thanks. That makes sense actually. Just to check though, final distance in Vahaduo Admixture is the squared euclidean distance and the sources are minimum squared euclidean distance combination to the target?

Samuel Andrews said...

@Gaberial,

True.

Davidski said...

@All

So do Scandinavians and Brits really have recent Viking Age-era Italian ancestry, as claimed by Margaryan et al?

I've been doing other things since I made this post, so I haven't even started looking at this question. But I will next week.

vahaduo said...

@Matt

All reported distances are Euclidean. Squared Euclidean distances are used only internally to speed up calculation as switching to them doesn't change the results and requires fewer operations to be performed.

Samuel Andrews said...

@Davidski,

I have not looked into that. It sounds unlikely, considering Y DNA J, E, R1b U152 don't really exist in Scandinavia.

There is up to 5% J2a, E1b V13 in France, Germany, Belgium, and I think England. But it makes more sense this is from the Roman empire than from the Middle Ages.

Rob said...

@Ambron

''Jordanes and Nestor locate the Slavic homeland between the Danube and the Vistula, and genes seem to confirm it''


The area in question was within the Chernyakov cultural horizon before 250-450 CE; and like many other regions it experienced population decline due to the ‘Germanic‘ out-migrations. The people which remained, e,g the Cotisa-Botosana groups (450-500 CE) are intuitively Dacian-oid, and probably represent the I2a1b-rich element in the proto-Slavic Genesis

vahaduo said...

For exploring the dataset and shifts between different populations I recommend toying with the distance comparison modes.

Distance difference: ( AC - BC ) ↑
A: VK2020_NOR_Mid_MA
B: VK2020_NOR_Mid_VA
C: ↴
-0.02631190 ITA_Tivoli_Renaissance
-0.02626794 Italian_Lazio
-0.02623164 Levant_Ashkelon_IA1
-0.02617476 Italian_Aosta_Valley
-0.02615009 Italian_Umbria

Just note that the top population is not always (or even usually) the best match for a 2-way model.

Target: VK2020_NOR_Mid_MA
Distance: 1.9008% / 0.01900837
80.6 VK2020_NOR_Mid_VA
19.4 ITA_Tivoli_Renaissance

Target: VK2020_NOR_Mid_MA
Distance: 1.7690% / 0.01769014
67.8 VK2020_NOR_Mid_VA
32.2 Italian_Aosta_Valley

In case of the Vikings also comparisons between females and males from a given site can be interesting and informative.

Distance difference: ( AC - BC ) ↑
A: DNK_Jutland_VA_Male
B: DNK_Jutland_VA_Female
C: ↴
-0.00993412 VK2020_Faroes_EM
-0.00909297 VK2020_Scotland_Orkney_VA
-0.00904138 Scottish
-0.00893596 English
-0.00856395 Irish
-0.00823353 English_Cornwall
-0.00820836 Orcadian
-0.00768026 VK2020_GreenlandE_VA
-0.00676130 DEU_MA
-0.00655765 VK2020_DNK_Sealand_IA

Distance difference: ( AC - BC ) ↓
A: DNK_Jutland_VA_Male
B: DNK_Jutland_VA_Female
C: ↴
0.00929426 Bell_Beaker_HUN_EBA
0.00881992 Scythian_HUN
0.00800465 UKR_Chernyakhiv_Shyshaky
0.00799861 VK2020_UKR_Lutsk_MA
0.00796095 Montenegrin
0.00790644 Croatian
0.00790029 SWE_BA
0.00788789 Slovakian
0.00776725 Bell_Beaker_Bavaria
0.00774476 Moldovan_o

Distance D3: ( AC - BC ) / ( AC + BC ) ↑
A: DNK_Jutland_VA_Male
B: DNK_Jutland_VA_Female
C: ↴
-0.25248427 English
-0.24042850 Scottish
-0.21791252 VK2020_Faroes_EM
-0.21463489 Orcadian
-0.21215551 English_Cornwall
-0.19393208 Irish
-0.18588322 Dutch
-0.18178720 Welsh
-0.17902221 VK2020_Scotland_Orkney_VA
-0.15435717 DEU_MA

Distance D3: ( AC - BC ) / ( AC + BC ) ↓
A: DNK_Jutland_VA_Male
B: DNK_Jutland_VA_Female
C: ↴
0.19838422 Bell_Beaker_Bavaria
0.17701722 Bell_Beaker_HUN_EBA
0.15703888 Bell_Beaker_CZE
0.13331368 Austrian
0.12133276 DEU_Lech_EBA
0.09949451 Hungarian
0.09810819 Scythian_HUN
0.09571046 DEU_Singen_EBA
0.09378567 German
0.09032119 Croatian

You can find the spreadsheets here:

individual samples
https://gist.github.com/vahaduo/d500087bd049e98393916b593cbb7e77

averages
https://gist.github.com/vahaduo/21dcdbc2804f39cf10b86bd5851eede7

Gabriel said...

The idea Paleo-Balkan people have influenced Slavs has long been popular, but unless early proto-Slavs were even more steppe-rich than Balts, that theory is very unlikely.

@Samuel

They may also be of farmer origin or even Iron Age Celtic in places like Belgium, England and parts of Germany considering Iron Age Celts were rather farmer-rich compared to local Beaker populations.

Samuel Andrews said...

@Davidski,

There is slight "southern" ancestry in England, because they are slightly south of what would be a Saxon-Briton mix.

If their "southern" ancestry was from Roman empire then Welsh would be more "southern" but they're not.

My guess is it is from France (Normans) not Italy.

Samuel Andrews said...

@Gabriel,

Yeah, there could have been some variation in Iron age British Isles. It might be wrong to assume Irish for example are a proxy for Celtic England. As Ireland, seems to have more local Beaker ancestry, less Celtic ancestry than few Iron age England samples. Maybe some Celtic pops in England were extra Celtic/French-like.

ambron said...

Rob

On the basis of the palynological data, we can already make the first and cautious conclusions regarding demographic movements.

In the early Middle Ages, agricultural activity declines in Lesser Poland (upper Vistula) and increases in Greater Poland and Brandenburg. Logic dictates that in the Slavic homeland the size of the population should decrease and increase in the areas of Slavic migration.

https://www.sciencedirect.com/science/article/abs/pii/S0014498315000443

ambron said...

Gabriel, the oldest uniparental Slavic genomes, autosomal look like this (Arza's models):

Target: DEU_Singen_EIA:MX265
Distance: 2.2907% / 0.02290667
38.4 HUN_Baden_LCA
34.6 Corded_Ware_DEU
23.0 Baltic_EST_BA
4.0 POL_Globular_Amphora

Target: Scythian_HUN:DA197
Distance: 1.3804% / 0.01380425
38.8 HUN_Baden_LCA
33.2 Baltic_EST_BA
25.6 Corded_Ware_DEU
2.0 POL_Globular_Amphora
0.4 SWE_Motala_HG

Garvan said...

vahaduo said...
For exploring the dataset and shifts between different populations I recommend toying with the distance comparison modes.


A) In you notation, what does "Distance difference: ( AC - BC )" mean?

Is AC the sum of the product of P values of A and C for each P in G25?
Is it the sum of the absolute difference between A and C for each P in G25?
Is it the sum of the difference between A and C for each P in unscaled G25?
Something else?

B) Should I use scaled or unscaled coordinates with these distance measurements?

C) Is D3 expected to correlate with D statistics?

E) I find the advanced tools on your site difficult to work with because of lack of examples. If you or somebody else could make a short tutorial that would be much appreciated.

Thanks.

Gabriel said...

@Samuel Andrews

Maybe, but Wales wasn’t the core Roman area, that’s why they speak Welsh and not British Romance.

I do agree the Roman ancestry would be minor though.

Rob said...

@ Ambron

''In the early Middle Ages, agricultural activity declines in Lesser Poland (upper Vistula) and increases in Greater Poland. Logic dictates that in the Slavic homeland the size of the population should decrease and increase in the areas of Slavic migration''


The paper doesn't imply what you state

''The mountainous hinterland of Lesser Poland offers a
good comparison for the pattern of growth visible in Greater Poland, as it presents a totally different scenario.
No visible growth takes place before the tenth and eleventh
centuries, and this is not surprising given both the
peripheral location of the area within the early medieval
political–economic networks,''

SO Lesser Poland simply grew slower than other areas. It does not state that it was a source for other regions

Ambron, the archaeological picture was made clear in the 1970s by Godlowski.

Bernadette said...

Any thoughts on the supplementary paper claiming that Danish Vikings had a higher Swedish component and that is how they said that the English had 6% input from the Danish Vikings and the rest of their Danish component was Anglo-Saxon.

UK populations all have high ‘Denmark’ ancestry. Although Anglo-Saxon and Danish Viking
ancestry are hard to distinguish, Viking-era Danes have too much ‘Sweden’ ancestry to have
contributed more than around 6% ancestry into England, whereas they could plausibly have
contributed all (up to 16%) of the Scottish and Irish signal. Anglo-Saxon samples are needed to explore this further.

pg 137 of the supplementary paper.


AWood said...

@Gabriel

British Romance? The Britons (at least the common people) never spoke a Latin language.

AWood said...

@Gabriel/Samuel

I'm rather hedging my bet that the higher southern European among modern south English like myself is due to Iron Age migrants from northern Gaul/Belgium we commonly refer to as Belgae. They were all throughout southern Britain at the time of Roman expansion, but certainly came much later than those first BB L21+ guys dated to 2500 BC. They also had very minimal impact on Ireland for example. I guess they would have been related to the Halstatt 'Celts' and certainly made a larger genetic impact than Rome or the Normans.

Matt said...

@Vahaduo, ah, fair enough; to explain, thought the correlation of ranked distance and ranked distance^2 might not actually be perfect (e.g. best fit of distance might not be best fit of distance^2) for some reason of brain malfunction, but of course it is.

@Garvan, you obviously want the answer from vahaduo, so this is not in lieu of that, but as I'd understand it:

A) AC = The distance between pop A and some member of set C, 2) BC: BC: The distance between pop B and that same member of set C. (AC-BC) then the difference between those two.

This would be useful where you're looking at a two populations/samples which are highly correlated in their overall distances (from all the other populations in set C), and you are specifically interested in highlighting what affinities to other populations are different between them (e.g. useful for the finescale VK sample affinities, as Vahaduo highlights).

B) I think you could technically use either set of data, but obviously only scaled will actually give you results that relate to the underlying genetic differentiation; distances on unscaled data obviously don't recreate genetic distance meaningfully.

(E.g. to explain why this is the case, say you have a PCA with 5 dimensions where the bulk of the variance in genetic differentiation, say, 90% is in the first dimension, then 10% is split between the next 4, with scaling this is recreated in distances. Whereas, unscaled, each of the 5 dimensions has equal weight so populations who are close together in dimension 1 and close together in reality would seem to be the most differentiated, in complete contradiction to their actual genetic distance.

But this is the same point we have repeated over and over again in regards to scaling, and which seems by now only misunderstood by people who don't know what scaling is, i.e. that it is weighting dimensions to restore underlying distances in the input data, on unweighted dimensions which are all of equal size, not some kind of strange "distortion" of data which already recreates the input distances, which some people wierdly still seem to believe.)

C) As I understand it it, D3 would correlate with D-statistics in the sense that distances generally do, but it's not particularly close to the same thing! As I understand it it basically weights the D1 so that values converge on 0 for those populations that are very distant from the target set C. Let's say that two populations A and B are highly different in their affinities to some target in the set C, C1, but C1 also is really distant to both A and B. The D3 distance scales that distance D1 down (converging towards zero) in proportion to the the degree that C1 is distant to both A and B. So this is basically telling you "Which members of Set C show differences in relationship to A and B *and* are also close to both A and B" (again useful for exploring finescale differentiation).

That said, Vahaduo probably could built pseudo-f and pseudo-D statistics calculations though, on the basis that the euclidean squared distance is pretty correlated with Fst, and Fst relates to f2 statistics, and then the higher f3 and f4 statistics can be expressed as combinations of the f2 statistic.

ambron said...

Rob

Godłowski is a fairy tale writer, as is his mentor - Kossinna. The palynological data do not show a settlement void in Poland, and the genetic and linguistic data do not show the Slavic homeland in Pripyat.

Matt said...

Incidentally, talking about different distances we can calaculate from PCA, if we did want to recreate outgroup a similar property to f3 outgroup statistic with Global 25 (albeit as a distance score, *not* a similarity), what seems to work is the following:

1:

A: Outgroup, B: Target, C: Pop2
XY: Squared Euclidean Distance between X and Y
PseudoF3 = -1*((AC-BC)-AB)

2:
A: Outgroup, B: Target, C: Pop2
XY: Simple Euclidean Distance between X and Y
PseudoF3 = -1*((AC-BC)-AB)

...

Version 1 looks to correlate well with outgroup f3 statistics (down to the pattern where Lithuanian / Latvian tends to "max out" statistics in Northern Europe and Sardinian as we move to Southern Europe).

While Version 2 looks to have some interesting properties where local drift is relatively preserved, but drift in populations that form a clade to a pair of populations is heavily reduced.

Few example plots: https://imgur.com/a/AnlivKh

Example Datasheet: https://pastebin.com/JjDbvF6r (download and convert to .csv).

Rob said...
This comment has been removed by the author.
Gabriel said...

@AWood

But that’s what I meant with Iron Age Celts, they include the Belgae and other tribes, not just Hallstatt Celts.

Ioannis Gavras said...

@ Awood

I totally agree with this.Some people say south England influenced by Romans or Normans but i believe that during IA a Gaulish-Belgian tribe immigrated to the south of the island hence some Celtic lineages and their more southern autosomal there.

J.S. said...

@AWood @Chevalier @Gabriel

In temperate Europe, and more particularly in France, the Iron Age is divided into two periods: the First (800-400 BC) and the Second Iron Age (400-25 BC). The latter is often associated with Celtic cultures, the unity of which has been shown through the study of Celtic art. However, this apparent unity is now being rediscussed through recent work in archeology. While cultural diversity is well known from an archaeological point of view, it is still little discussed from a biological point of view. This work therefore offers an unprecedented palaeogenetic and palaeogenomic analysis of individuals from three necropolises in the north of France, distributed along the Seine valley, a major axis of exchanges betweenMancheand Burgundy. A total of 106 haplogroups, 87 mitochondrial haplotypes as well as 15 paternal lines were characterized. In addition, 12 low coverage genomes were obtained.

At the local level, the data obtained were systematically compared with the available biological and archaeological data, making it possible to highlight distinct functions. The necropolis of Urville-Nacqueville (Normandy) seems to welcome a cosmopolitan population, while that of Gurgy ‘Les Noisats’ (Yonne) is probably used by a local community. The cases of Barbuise ‘Les Grèves de Frécul’ (Yonne) and Urville-Nacqueville(Manche, Normandy)also reveal the complexity of the social organization of these Iron Age groups through the structuring of the funeral space. Although these necropolises host diverse communities, they share significant mitochondrial diversity, a lack of clustering based on maternal ties, and low diversity of paternal lineages. These results form a cluster of clues supporting a patrilocal-type matrimonial system and a patrilineal-type filiation, consistent with the data in the literature. At the regional level, the results show that the sites located in the lower valley of the Seine share more affinities with the groups in the south of England, while those in the upper valley of the Seine are closer to the populations of the east of France and occupy an intermediate position between the north and the south of France,thus highlighting a genetic structure of these groups according to their location along this river axis.

Finally, on a continental scale, the results show that the Iron Age communities of Western Europe form a coherent genetic cluster and exhibit genetic continuity with the Bronze Age groups. The data acquired agree with archaeological hypotheses favoring an economic, political and / or climatic transition to explain the transition between the Bronze Age and the Iron Age, in accordance with the local evolution of groups perceived at the genetic level. .
http://www.theses.fr/2019BORD0332
Free access 01/012021

Garvan said...

@Matt

Thanks for the clear explanation of the notation, I mistakenly thought Vahaduo was using a non-Euclidean distance.

Regarding scaling, G25 scales used the square root of the eigenvalue. Do you know why it uses the square root? I reread the old posts related to this but could not find the reason for the choice. I think the eigenvalue is proportional to the percentage variance.

I normally use unscaled G25, and create a new PCA using only the samples I am interested in. Should I be working with scaled data in this case?

Interesting ideas for PseudoF3. The link to the example data did not work for me.



Groo Salugg said...

@weure

Flat ociput is the norm in adults.
Round ociput in adults is a juvenile neotenic trait.
Any thinning, shrinkage etc are growth defects, that are typical for farming populations (either due to lack of "prunning", or being actually selected for).
The general rise of alleles that are correlated with higher IQ score in modern populations started 4000BP IIRC, that is 500 years after Beakers.
Any correlation between thin skull, flat ociput, and high IQ in Beakers, if there is one, is just a coincidence.
The thin skull and flat ociput is not really a requirement a better functioning brain.

Ioannis Gavras said...

@ Groo Salugg

Skull shape has many peculiarities. There are many factors for being brachy or dolicho or even meso.I Think the whole theory of BB's were Brachy with flat occiput is a extreme in some way. First of all we are talking for a Corded ofshoot and we know that Corded was definitely not Brachy.So,we are probably have to do with a special subgroup that for some reason its individuals have turned into high levels of CI.Hmmm okay!But can you guys explain why the modern ethnic groups with the majority of their ancestry being BB why their CI is very low? Take as an example Irish and British or even the south Basques/Iberians as well. These ethnicities have one of the lowest CI in Europe taking serious anthropological studies and researches. As I mention above there are many peculiarities to have a high or low CI even flat occiput nowdays witch is a crazy issue for young generations btw. I think the whole theory about BB' being Brachy with flat occiput is extreme and overrated. But let's focus on genetics and leave these things.

Davidski said...

About three quarters of Dutch Bell Beakers had skulls like this guy.

https://eurogenes.blogspot.com/2016/06/the-man-with-flat-occiput.html

And this type of head shape was pretty common among most Beaker groups.

So it's something that really sticks out about the Bell Beaker phenomenon. But yeah, it wasn't passed on to Western Euros with R1b-P312, or even with Beaker-like autosomal ancestry.

Matt said...

@garvan, no trouble.

1) Here are some demonstrations of how I understand the D1 and D3 distances from Vahaduo to work (albeit I've calculated these in a spreadsheet on the basis vahaduo has given, and the distance I'm working with are using a conversion of G25's distance to estimated Fst, but this shouldn't effect the basic principles): https://imgur.com/a/gpiqKh4

You can see the D1s are giving you the populations (C) who are most different in affinity to population A and B (here 1 and 2), while the D3s are giving you the populations that are most different in affinity to A and B *weighted* for their overall distance to both (such tha A and B each are 1 and -1 on that scale).

So (within the set of samples I've chosen to look at!): D1 for Yamnaya Samara:Early Corded Ware peaks Sarazm_En and CHG at the YS end WHG+GAC at ECW end vs , while D3 (which weights by summed distance from A+B) peaks Progress_En at the YS end vs at Sintasha+Beaker NLD for the ECW end.

Similarly, D1 for Baltic_EST_BA:Beaker_NLD peaks at WHG/Iron Gates on the Baltic_EST_BA end vs Minoan and Ganj_Dareh on the Beaker_NLD end (reflecting greater CHG-like ancestry), while running D3 for the same pair gives more present day Latvians and Lithuanians on the Baltic_EST_BA end, against Irish, Norwegian, Icelandic on the Beaker_NLD end.

D1 is useful for the question of "Which populations are most related to the stream(s) of ancestry which differentiate A and B?" (I've basically used the same thing sometimes in comments here) while D3 is a ingenious new idea by vahaduo (that I hadn't thought of or seen before at all) to find close populations close to A and B which are most differentiated in respect of them. (On both stats, D1 and D3, populations that essentially form a clade to A and B fall around 0).

The D3 function in Vahaduo seems cool and people should probably be using it if they care about discussing which populations close to a pair of populations AB show different affinity to them. (Albeit at some point you're eventually going to run into limits of G25's ability to distinguish between populations and its going to turn into noise; but it doesn't for at least 2/3 of those comparisons in my plots linked above).

2) No, unfortunately I don't actually know the technical reason for why that's the right way of scaling, though I can remember in previous discussions at some point I looked it up (I think there may have been a section about this in PAST3's help file, since PAST has an eigenvalue scaling option in PCA / PCoA output, or there may be something on the web).

Certainly I can say that when you ask PAST3 to output a PCA / PCoA that is scaled / unscaled, on the exactly same data, and then compare them directly, it *is* scaling by the square root of the eigenvalue, and euclidean distance values on the scaled data are exactly the same as on raw inputs. So that is the way that it is doing it, and not our arbitrary choice. But I can't really say, technically speaking, why it is right. If Vahaduo is still following the thread, he might be able to offer a comment.

3) I may have not explained properly about that pseudo-F3 datasheet; when downloaded as a .txt file and extension changed to .csv and loaded as a spreadsheet, you should see in the same spreadsheet 3 vertically stacked PAST datasheets. One with the squared G25 distance, then the two pseudo-f3 measures. If you're not getting this when downloading I'm not too sure what's going on! :/

Archi said...

Gabriel said...
"But that’s what I meant with Iron Age Celts, they include the Belgae and other tribes, not just Hallstatt Celts."

The Belgae and all other Celtic tribes are the Hallstatt Celts. All Celts come from Hallstatt culture.

@Chevalier de Balibari

"First of all we are talking for a Corded ofshoot and we know that Corded was definitely not Brachy."

Bell Beakers were not Corded Ware offshore.

-------

I don't understand at all what kind of scaling by eigenvalues. There is no such thing. The eigenvalues have nothing to do with the scale of the PCA, they are simply the variance of the principal values, and coords cannot be multiplied or divided by them.

Vladimir said...

I think this is an interesting article:
aYChr-DB: a database of ancient human Y haplogroups

Abstract
Ancient Y-Chromosomal DNA is an invaluable tool for dating and discerning the origins of migration routes and demographic processes that occurred thousands of years ago. Driven by the adoption of high-throughput sequencing and capture enrichment methods in paleogenomics, the number of published ancient genomes has nearly quadrupled within the last three years (2018–2020). Whereas ancient mtDNA haplogroup repositories are available, no similar resource exists for ancient Y-Chromosomal haplogroups. Here, we present aYChr-DB—a comprehensive collection of 1797 ancient Eurasian human Y-Chromosome haplogroups ranging from 44 930 BC to 1945 AD. We include descriptors of age, location, genomic coverage and associated archaeological cultures. We also produced a visualization of ancient Y haplogroup distribution over time. The aYChr-DB database is a valuable resource for population genomic and paleogenomic studies.

https://academic.oup.com/nargab/article/2/4/lqaa081/5920394


Samuel Andrews said...

@Rob, Archeaology question.

Were Megalith graves the main type of burial for all of Western Europe, 4000-2800 BC? Or at least were they very common everywhere. I have read it was common in Northwest but was it also common in Spain?

Was it common in Globular Amphora as well? DNA shows Globular Amphora is derived from Funnel Beaker. Funnel Beaker did lots of Megalith burials so it would make sense Globular Amphora did too. (Which fits with Globular Amphora having lots of Western Euro farmer ancestry).

I'm asking in part because it seems Corded Ware, Bell beaker both went into mainly Megalith burial territory. There is a chronology switch from Megalith burials to Single Burials across Europe in 3rd millenium BC.

Ioannis Gavras said...

@ Archi

Do you have any source or a proof that BB derive their origins from a Corded related group?Can you explain what different they got and its making them foreigners?

Archi said...

@Chevalier de Balibari
"@ Archi Do you have any source or a proof that BB derive their origins from a Corded related group?"
It is necessary to ask Davidsky, he puts forward this hypothesis. I don't have any.

"Can you explain what different they got and its making them foreigners?"

They differ in everything from anthropology and languages to archeology and genetics. What they have in common is that in the process of moving to the territory of CWC, the BBC adopted elements of the CWC.

https://i.ibb.co/fXcg0GH/CWC-SGBR.png
https://i.ibb.co/ysMC0zL/Centum-IEgroups.png

Davidski said...

I'm not sure why the idea that Bell Beakers were an offshoot of the Corded Ware culture is still such a big deal for some people, considering that R1b-L51 has been found in Corded Ware samples and the earliest Bell Beakers are basically identical to nearby Corded Ware groups in terms of genome-wide ancestry.

I've blogged about this quite a bit in recent years, and I'll be blogging about it again soon when more samples are published from western Corded Ware/Single Grave sites.

Vladimir said...

@Davidski “I'm not sure why the idea that Bell Beakers were an offshoot of the Corded Ware culture is still such a big deal for some people, considering that R1b-L51 has been found in Corded Ware samples and the earliest Bell Beakers are basically identical to nearby Corded Ware groups in terms of genome-wide ancestry.“

I agree. I think it is much more intriguing where R1b-L51/L52/Z2118 and R1a-M417/CTS4385/Z645 were located from 4000 to 3000 BCE, I.e. in the period immediately preceding the CWC, since this is probably where IE finally formed. Slightly smaller, but also interesting where were R1b-M269/L23/PF7562 and R1a-M198/M417 from 7000 to 4000BCE as proto-Anatolian-Tocharian-Indo-European language began to form during this period

Rob said...

@ Sam

Yeah Megaliths are just a form of monumentalism to commemorate the dead. Various types of monumental tombs spreang around Europe c. 5000 - 4500 BC, in the border zones where LBK/ SBK & their descendants interacted with HG-dense communities. So in the north European lowlands you see long mounds, Megaliths and their various forms in Atlantic Europe, and of course, kurgans in the north Pontic. Megaliths probably first sprang around western France (which had a lot of large stones in the landscape), then spread into Iberia and northern Europe. Many had gone out of use even before early CWC. I.e. ~ 32/3000 BC, esp. in Iberia & France and communial burial was already shifted toward more individualised form of burial before the true 'single grave' burials of CWC/ EKG arrived; whilst in Germany passage graves were still in Use.

Im not 100% sure of the origin of GAC. Intuitively, it would make sense for GAC to be a specific faction within TRB which expanded. However, the GAC lineage (I2a2a1b2) has not yet been documented in TRB. Moreover, from a GW-perspective, it seems TRB (at least in Poland) are an LBK / WHG mix, with only piecemail western farmer admixture. On the other hand, GAC are definitively 'western farmers' from their Neolithic ancestry, but their HG ancestry has some EHG and their lineages points to a possible E. Baltic origin. They burials were variable, inclduing megaliths, flat graves, single burials , communal burials, etc


weure said...

@Groo Salugg, there are in history lots of swings between dolichocephaly and brachycephaly. In the case of the Northern Bell Beakers, they were most probably in the Protruding Foot Beakers a subset of Single Grave/Corded Ware. The Steppe Pastoralist literally settled down, became agriculturalist, their lifestyle and diet changed. And their culture/ society became more complex. So better brains needed ;) The change in diet already meant some less stress on the bones, a more spherical form is more protective, the vault could be less thinner, which is all more efficient. That do we probably see along the Bell Beakers the 'robust dolicho's' became more brachycephalic with a flat occiput!

vahaduo said...

@ Garvan
Hi.

A) In you notation, what does "Distance difference: ( AC - BC )" mean?

Euclidean distance between A (first selected Target) and C (Source) minus Euclidean distance between B (second Target) and C.

B) Should I use scaled or unscaled coordinates with these distance measurements?

Scaled.

C) Is D3 expected to correlate with D statistics?

Theoretically it should be correlated, see this publication:
https://academic.oup.com/mbe/article/36/12/2878/5543093

E) I find the advanced tools on your site difficult to work with because of lack of examples. If you or somebody else could make a short tutorial that would be much appreciated.

Sooner or later I'll probably have to start a blog.

@ Matt

I'm working on another update right now, so I'll respond later.

BTW if by "D1" you mean the distance difference then in this case switching to squared distances changes results.

Norfern-Ostrobothnian said...

@Vahaduo
Could it be possible to create "partitioned" calculations like GEDmatch oracles? Say I want to see what 50/50 mix works for the target population, the calculator would output the combination with the shortest distance to the target. The size of a partition could be inputted somewhere with 0.5 as the top and at least bigger than 0. Could you see if it's doable?

Samuel Andrews said...

Were Megalith burials also used in Italy?

The I2a2a1b2 in GAC as you know is the same I2a2a1b2 in Mesolithic Serbia. So, GAC must have gotten it from Central European hunter gatherers probably in Poland. This doesn't refute the idea they are descendants of TRB. As, it seems it was common for Western European farmers, and their descendants, to adopt local hunter gatherer Y DNA.

Btw, closest relative of Poland/Ukraine GAC is Sweden TRB. Closest relative of Sweden TRB is Poland GAC.

Vladimir said...

@Davidski
Do I understand correctly that here scientists have caught the moment and place of “mongolization” of part of haplogroup Q, originally former ANE?

https://www.biorxiv.org/content/10.1101/2020.10.12.336628v1

Rob said...

@ Sam

No, Italy isn;t really a Megalithic province (aside from Sardinia)
Fig 5 here shows distribution and chronology of spread https://www.pnas.org/content/116/9/3460

Yes TRB & GAC seem to part of a similar gene pool, although some early TRB seems linked to Danubian -LBK, whilst later the western flow increases. The other thing i had not mentioend is one important cultural facet which characterises GAC is cattle burials, which were borrowed from the Baden culture. So the specific I2a2 in GAC could indeed be from a north Balkan lineage.

Garvan said...

@vahaduo

Thank you for the explanation. A blog with examples of how to use your tools would be nice. Looking forward to it.

old europe said...


@ Rob

The relation between Italy and megalithism goes the other way round.
The western european farmers that created megalithism were predominantly cardial people that pushed into central western France from mediterranean France and north western Italy. There they entered in contact with local WHG and developed megalithism that later on extend eastwards with the trail
Michelsberg-TRB-GAC and further east to Ukraine GAC.
You can follow them with the expansion of the alpine jadite axe all the way from the alpine border between France and Italy to Brittany, England Germany Poland, Slovakia and Bulgaria ( Varna).
It would be interesting to investigate if the EEF component in Sredni Stog II and Dereivka comes from the se folks or from the balkan farmers.

Rob said...

@ Old Europe

''The western european farmers that created megalithism were predominantly cardial people that pushed into central western France from mediterranean France and north western Italy. There they entered in contact with local WHG and developed megalithism that later on extend eastwards with the trail''

Well, that's counter to the radiometric dating. In fact, we don't really see Megalithism in Cardial-Impresso groups. It emerges in central & northwestern France, and here the farmers stream was mixed post-LBK & Cardial, which in turn mixed with WHG.
We can observe the process of WHG introgression already by the middle Neolithic, in Michelsberg culture & similar groups, with I2a mixed with E1b, G2a, etc. But Michelsberg isn't a megalithic grave culture, its an enclosure culture, which is a different form of monumentalism. When we look at Megalithic groups, it's essentially WHG lineages.

Varna , Sredni-Stog aren't really part of that phenomenon, just some similar processes.

Matt said...

Following may be of interest Vahaduo if you're following the thread, related to the discussions on Fst:

A while ago I posted in some comments on here, that applying the following procedures in PAST3 to an Fst matrix transforms it into a dimensional space that roughly resembles the same results we see in Global25 etc: Ordination->Principle Coordinates Analysis->Settings:User Supplied Distance + Transformation Exponent C=1, then Eigenvalue Scale and copy out scores.

(Here's an example result: https://pastebin.com/6M007aPN / https://pastebin.com/Wds4LjLc split into 2)

One problem with this is that due to high amounts of error in Fst, although the low dimensions are dominated by the major trends, you get a lot of dimensions emerging which are essentially dominated by errors in Fst data. (Worse sometimes these errors are not very systematic, but only present between random pairs of ancients!). It's also confounded by systematic attraction of moderns to moderns and ancients to ancients.

So one way around this is using Vahaduo custom PCA projection.
Take the above dimensional data output and then copy into Vahaduo custom PCA, placing modern samples in source and ancients in projected (e.g. Source: https://pastebin.com/1E0p09mK, Project: https://pastebin.com/RC9tPcYL).

That then gives you an output that constrains the dimensional transform of the PCoA data such that if dimensions don't exist in modern samples, they are removed, which essentially removes all the dimensions that are just related to inflated Fst from damage in ancient samples (though also probably at least some real drift that's not represented in modern panels), as well as a lower and much more manageable number of dimensions.

An example output: https://pastebin.com/ecCCYLbh (with a little adjustment for estimated projection bias)

To show the reduction in long branching (e.g. likely error), here's a comparison of dendrograms on orignal Fst matrix, set against one from the above technique of constrained Fst, and G25 data: https://imgur.com/a/dyLeZWU

And here's direct correlations with a Fst estimated from only G25 distances: https://imgur.com/a/ERbMN7u

The "constrained Fst" (using PCoA+Vahaduo Custom PCA) method, has R^2 correlation of about 0.95 - 0.97 with the estimate of Fst from G25 distance. Note that they also have lower intercepts; there is possibly some general inflation in the Fst scores that is not present after this control (though some of this may well be very localised founder effects, e.g. in HG populations). (NJ tree structure also seems highly congruent between estimated on "constrained Fst" and the Fst estimated from G25 distance).

So to me this seems like potentially a useful method to get some estimates of population relationships, using only the Fst matrix, removing some of the sources of error (albeit probably removing some real ancient founder effects and drift, which just aren't represented in modern people).

Basically it's a way to turn a big batch of Fst matrix into dimensional data, then remove trends in that dimensional data that are unrelated to trends in modern dimensional data (e.g. are ancient founder effects, are in populations who left no modern descendents, or are simply errors in ancient dna) and then you can also retransform "back" to Fst by euclidean distance ^2.

«Oldest ‹Older   201 – 313 of 313   Newer› Newest»