Here's what
Martiniano et al. had to say recently in regards to the genetic shifts in what is now Portugal, western Iberia, during the Bronze Age that they saw in their ancient DNA data:
A recurring feature of ADMIXTURE analyses of ancient northern Europeans is the appearance and subsequent dissemination within the Bronze Age of a component (teal) that is earliest identified in our dataset in HGs from the Caucasus (CHG). Unlike contemporaries elsewhere (but similarly to earlier Hungarian BA), Portuguese BA individuals show no signal of this component, although a slight but discernible increase in European HG ancestry (red component) is apparent. D-Statistic tests would suggest this increase is associated not with Western HG ancestry, but instead reveal significant introgression from several steppe populations into the Portuguese BA relative to the preceding LNCA (S4 Text, S6 Table).
...
In the present analysis, fineSTRUCTURE has identified the 3 Portuguese Bronze Age individuals as a genetically distinct population (S23 Fig). When compared to Central or Northern European populations such as Ireland [11], the degree of discontinuity between the Neolithic and Bronze Age in Portugal is not pronounced. However, despite the small sample size we have evidence suggesting complete discontinuity at the level of Y-chromosome lineages with all 3 male Bronze Age samples presenting derived alleles at marker M269.
Although in ADMIXTURE analysis we were not able to observe the presence of the CHG-related cluster in the ancestry proportions of the Portuguese Bronze Age samples, with D(Mbuti, X; Portuguese MN/LNCA, Portuguese BA) we find support for CHG/Yamnaya related introgression and also an increase in EHG [Eastern European Hunter-Gatherer] ancestry.
Despite the authors' conclusion that steppe-related admixture was present in their Portuguese BA samples, the ambiguity created by their ADMIXTURE analysis encouraged some heated debates in the comments at this blog and elsewhere about whether their findings were legitimate, and also suggestions that the Portuguese BA R1b-M269 Y-chromosomes were not derived from the steppe.
To try and put this debate to bed, at least on this blog, let's run the same samples with the qpAdm mixture modeling algorithm. I don't want to get into the details here about the difference between ADMIXTURE and qpAdm, because I don't feel it's something that I can explain accurately. But, suffice to say that qpAdm is a more direct way of estimating ancestry proportions, so, in my experience, it's less likely to lose minor but significant admixture signals in a well thought out and put together analysis.
First up, I need to test whether these Portuguese BA (Portugal_BA) individuals can be modeled as a two-way mixture between EHG and Portuguese Late Neolithic farmers (Portugal_LN).
Portugal_BA
EHG 0.093±0.036
Portugal_LN 0.907±0.036
taildiff: 0.0102798873
chisq: 20.015
Full output
Nope, they can't. But what happens if I add CHG to the model?
Portugal_BA
CHG 0.106±0.048
EHG 0.042±0.042
Portugal_LN 0.852±0.042
taildiff: 0.0367007784
chisq: 14.946
Full output
The statistical fit improves, but it's still lousy, which perhaps suggests that I need a temporally more proximate CHG-related reference sample. How about Yamnaya?
Portugal_BA
Portugal_LN 0.849±0.045
Yamnaya_Samara 0.151±0.045
taildiff: 0.0725988319
chisq: 14.371
Full output
That's not too bad. But let's try a more proximate Yamnaya-related population: Bell Beakers from Germany. Note that some of these Beakers belonged to Y-haplogroup R1b-M269(P312+), which is the most common Y-chromosome lineage among present-day Iberians.
Portugal_BA
Bell_Beaker_Germany 0.328±0.089
Portugal_LN 0.672±0.089
taildiff: 0.109643502
chisq: 13.065
Full output
Somewhat better, and we could probably keep going like this, improving the fits each time, with more relevant reference samples if they were available, like, say, late Beakers from what is now France. I suspect also that using more westerly Hunter-Gatherers than EHG, perhaps from what is now Ukraine, might significantly improve the second model. In any case, my qpAdm analysis provides strong evidence that, unlike Portugal_LN, Portugal_BA harbored CHG-related ancestry that was probably mediated via Yamnaya- and Beaker-related groups.
Citation...
Martiniano R, Cassidy LM, Ó'Maoldúin R, McLaughlin R, Silva NM, Manco L, et al. (2017)
The population genomics of archaeological transition in west Iberia: Investigation of ancient substructure using imputation and haplotype-based methods. PLoS Genet 13(7): e1006852. https://doi.org/10.1371/journal.pgen.1006852
See also...
Late PIE ground zero now obvious; location of PIE homeland still uncertain, but...
Who's your (proto) daddy Western Europeans?
Steppe admixture in Mycenaeans, lots of Caucasus admixture already in Minoans (Lazaridis et al. 2017)
43 comments:
@ Davidski
Could it be possible that the Maykop Culture displaced some CHG people who lived North of the Caucasus(maybe violently) and therefore some women fled and found refuge among the Steppe people ? Some of the earliest swords were found in Maykop Kurgans.
Could you at least explain what 'P-value' and 'Chisq' represent? Maybe a lot of folks here know what they represent but I don't. I take it these tell you whether or not the model is workable or reflects reality or not?
We see violence between farmers and Hunter Gatherers at Talheim and many other places. So the same could have happened North of the Caucasus ?
Thanks for this David.
" I suspect also that using more westerly Hunter-Gatherers than EHG, perhaps from what is now Ukraine, might significantly improve the second model with CHG."
How about from the Carpathians (KO-1) or the Danube? Are any of those samples out, and if so would you be willing to take a look at how that changes the fit?
Still a pretty crap fit. Maybe try the Minoan samples for funzies as one component? Just as another way of adding some extra CHG without directly referencing Yamnaya or CHG.
If you could get a model with a really solid fit here I think that would clarify things. I don't doubt that Bell Beakers brought steppe ancestry with them from Central Europe (though I disagree on how they acquired it), but excess CHG is interesting.
More interesting to look at well be those Hungarian Bell Beakers who seamed to have excess EHG but lacked CHG. If there was already a baseline excess of CHG though that will really muddle things.
@Arch Hades
GIYF
https://en.m.wikipedia.org/wiki/Pearson%27s_chi-squared_test
https://en.m.wikipedia.org/wiki/P-value
David,
When will you be able to get the bell-beaker samples (from the study of May 2017)
@ Davidski
What's the fit if you use Portugal_LN and Hungary_BA:I1502?
Just out of curiosity
How would the fit of your last grouping be affected if the Bell_Beaker_Germany samples had admixed with a population (say WHG) that was also admixed into Portugal_LN---That 33% would then be a very high over-estimate?
@Arch
P-value should be high and chisq low.
@Ryan
KO1 doesn't work, because he doesn't have any CHG and not enough EHG. Minoans don't work because they have very little EHG.
The admixing population has to be very similar to Yamnaya, with roughly equal amounts of CHG and EHG, but some WHG does help to improve the fit, probably because of the Middle Neolithic admixture picked up by the steppe people on their way to Iberia.
@truth
The Olalde 2017 data will be released when the paper is published in a journal.
@Lee
WHG and indeed Europe_MN admixture in the Beakers is probably making the results fuzzy (note the higher standard errors), but it's unlikely to be overestimating the Beaker ancestry proportions to an impossible level, because the program isn't seeing the WHG in the Beakers, but rather the Beakers as a package that includes a specific ratio of WHG.
In fact, these Beakers have around 50% Yamnaya ancestry, with the rest being WHG and Anatolia_N, so very similar to Portugal_LN, and when we look at the successful models, Portugal_BA show around 15% Yamnaya ancestry and 30% Beaker ancestry, which makes sense.
You could even try British beakers or BA, given the Cist grave connection of the Atlantic facade
@Simon
Hungary_BA:I1502 produces a very good fit, but the standard errors are very high, probably because of the shared Europe_MN ancestry with Portugal_LN.
https://drive.google.com/file/d/0B8XSV9HEoqpFS2FSdHBQYjJ0YUE/view?usp=sharing
@Rob
A model with Ireland_EBA looks very similar to the one with the German Beakers.
https://drive.google.com/file/d/0B8XSV9HEoqpFWmg5TDZhemtBakU/view?usp=sharing
Thanks Vincent, I read the Wiki article and also watched a Youtube video explaining P-value, and I'm pretty sure I grasp the basics now. I didn't even know what the term P-value meant before.
@David
So since the P-value of the Bronze Age Portuguese being modeled as German Bell Beaker + Portuguese Neolithic is only .11, this suggests it's not that good of a model? Looking at the P-value numbers of your Mycenaeans being modeled as Sintashta + Minoan, that came out .96. I guess that suggests this model for the Bronze Age Portuguese is nowhere near very realistic?
Actually, the important thing to understand is that the fits clearly improve as I use CHG and EHG, and then more proximate CHG/EHG-related references, instead of just EHG.
The numbers and various measures of significance don't really matter that much, because they're affected by the number and types of outgroups used, the number of markers, sample quality, and probably other factors.
So the main take away point is that Portugal_BA is better modeled with CHG and Yamnaya-related reference samples than without them, and if anyone wants to prove that there was no Yamnaya-related migration to Iberia, then by all means, but you need to show that a very similar but unrelated population existed somewhere other than the steppe.
I have been looking at Davidski's global 10 with ancients PCA. If you plot PC2 vs PC3, you can see some interesting patterns. Kum4---dating to 3500-2900 BCE is very close to the Sintasha mean position. That said, Kum4 is of very low coverage. Aside from that, the Myceneans cluster with Anatolian Chalcolithic from Barcin and the Minoans with BA Anatolians and Tepecik. Minoan, I9130, whose Y chromosome is a G2a and hails from south Herakleion prefecture is closer to the Anatolian Neolithic samples from Barcin and Tepecik, than the two Minoan samples from Lassithi who are J2a-M319*. These two samples are closer to the BA Anatolians. These patterns fit with a substrate Neolithic in both Greece and Crete with secondary movements from NW Anatolia to the Peloponnese and SW Anatolia to Minoan Crete. Of course Kum4 might track the Anatolian I-E languages into Anatolia (it is later than the Anatolian Chalcolithic sample) if one can trust its DNA.
@Davidiski
Okay. I am confused. I get that stats are very much influenced by the number of samples in the groups. Chi square tail probability significance is strongly influenced by degrees of freedom. But if the stats are mostly meaningless how can you judge if the model fitness improves or worsens?
@Lee
The stats aren't meaningless. The numbers are in large part specific to the analysis and not easy to compare directly to other runs, especially where the outgroups are different, but the trends and patterns you see in the data when using different reference samples certainly will tell you which results are more realistic.
In this case, Yamnaya and closely related LNBA Europeans clearly provide the best fits, at the very least relatively speaking.
So who, at least generally speaking, do these results implicate as the donors of the non-Portugal_LN admixture into Portugal_BA? Obviously some sort of BA steppe group like Yamnaya and their offshoots into Europe.
Anyone who disagrees with this conclusion should be able to provide a plausible alternative. So what's the alternative that can produce as good or better stats?
And I won't even get into the uniparental links, both Y-DNA and mtDNA, which similarly implicate BA steppe groups in this process.
@Roy King
Would be interesting to compare this with a 3D PCA of the sort Open Genomes used to make. But yes, it's quite outlandish to assume that the pre-Greek substratum in the Peloponnese unlike the Minoans didn't have CHG-rich Anatolian admixture. From what I've read, the reason why Crete developped the sophisticated Minoan civilisation in the MBA while the rest of Greece remained backwards was that it was spared from the Northwest Anatolian incursion around 2200 BC.
@ Simon W
"@ Simon w
"From what I've read, the reason why Crete developped the sophisticated Minoan civilisation in the MBA while the rest of Greece remained backwards was that it was spared from the Northwest Anatolian incursion around 2200 BC"
Are you sure you understood whatever you read correctly ?
"So, while the first half of the 3rd millennium BC in Thrace is characterised by a (comparatively) moderate level of social and economic complexity and the ideological dominance of pastoral tribes of a north-Pontic origin, there is a real explosion in complexity in the period between 2400 and 2000 BC and the region becomes increasingly included within a much wider network that is now dominated by frequent and highly visible exchange and trade, and new forms of prestige and status expression, as to be detailed below...
There can be no doubt that this new situation emerging after c. 2400 BC, within the Early Bronze Age (EBA) III or Sveti Kirilovo phase in the Bulgarian chronological system, is due to inuence from across the Bosphorus and Dardanelles, where exactly the same developments had happened several centuries earlier in a more gradual manner. In Thrace the change appears to have been more abrupt and levels of societal and economic complexity appear to rise rapidly"
- V Heyd
^^ but I don't suggest they were PIEs
@Davidiski
To be clear, I am NOT trying to assert that some level of eastern influence entered the Iberian Population during the Bronze age. Not much and not nearly the same levels as other parts of central and eastern europe.
But the ancestry does not appear to be the full Steppe package--lacking that teal component. So more HG less Steppe like. So it is possible that the HG ancentral population high in the Yamnaya people and low in the CHG are the actual source not Yamnaya directly. Perhaps they were pushed westward prior to the Yamnya expansion or because of it.
This population could be the exact same population that caused an increase of Iran like ancestry in the Minoans (https://reich.hms.harvard.edu/sites/reich.hms.harvard.edu/files/inline-files/nature23310_0.pdf). Uniparental marker X2c suggest that an ancient Near eastern population entered the Iberia as far back a 9000 years ago (https://www.ncbi.nlm.nih.gov/pubmed/28177087)
So this "eastern" population could be much older and isolated than the Yamnaya...a possibility.
But I am still confused by your rejection of:
Portugal_BA
EHG 0.093±0.036
Portugal_LN 0.907±0.036
P-value 0.0102798873
chisq 20.015
what specifically about its stats say it is a bad fit. StDev look reasonable. the p value is well below 0.05. SO what am I missing on your assessment?
@Lee
So it is possible that the HG ancentral population high in the Yamnaya people and low in the CHG are the actual source not Yamnaya directly.
Eastern European Hunter-Gatherers didn't carry any CHG, and, also, for instance, there's no plausible cultural mechanism by which a Hunter-Gatherer group could have migrated in such large numbers across Europe.
So it had to be a group of pastoralists from the steppe, and certainly of mixed origin. Maybe they had a lower ratio of CHG than Yamnaya Kalmykia and Samara, but they definitely had it, because in the second model above for Portugal BA the CHG ratio is actually higher than the EHG ratio. It's a poor fitting model, but the ancestry proportions do make sense.
the p value is well below 0.05.
Well that's one of the problems. It should be well above 0.05.
But an even bigger problem is the difference in the statistical fits between this model and the Yamnaya model. There's a clear shift in favor of Yamnaya, which is obviously at odds with claims that Portugal BA didn't have any CHG or steppe BA ancestry.
I dunno about this Bronze Age Portuguese sample, but modern Iberians have of course some small Copper/Bronze Age Anatolian admix, otherwise they would not be so different from the Basques, who are actually a perfect mix of Anatolian farmers, Yamnaya and WHG. Modern mainland Spaniards are also about 15% Berber/African as for Lazaridis et al -the Portuese and Iberian islanders have more than that- which raised their score in Natufian/Levantine/SSA compared to SE Europeans, as for Boatigue et al.
I think Italicroots members should be banned from commenting on eurogenes for obvious reasons
@Davidinski
So the qpADM test is for the the null hypothesis?
the probability that X and Y population do NOT contribute to population A
So a significant finding p value less than 0.5 in the test rejects that as a probability?
That seems backwards way of going at it but ok.
italicroots is reknown worldwide for its scientific accuracy, and a lack of bias or any sort of trust issues. Everyone cares very much about what they ahve to say regarding most topics, particularly about Iberia.
I don't see what's wrong about Italicroots last post. Most Spanish and Portugese do have a layer of East Mediterranean stuff related to Minoans and AnatoliaBA and a layer of Berber-type stuff. Basque are an exception to this rule, they only have AnatoliaN, WHG, and Yamnaya.
You might want to check the site first
Maybe Italicroots is a nutjob. I didn't claim he wasn't. We're all a little nuts. And If he is a nutjob this saying is more appropriate than ever "Even a broken clock is right twice a day."
@David - "KO1 doesn't work, because he doesn't have any CHG and not enough EHG. Minoans don't work because they have very little EHG."
I'm talking about using KO1 (or ideally one of the Danube R1b HGs) compared to the Hungarian Bell Beaker that has EHG but no CHG. One of the Olade samples that isn't out yet. I'm not talking about using it for anyone else. There's definitely Yamnaya in Portugal_BA - I'm just curious to know if Yamnaya is the sole source of EHG ancestry.
"The admixing population has to be very similar to Yamnaya, with roughly equal amounts of CHG and EHG, but some WHG does help to improve the fit, probably because of the Middle Neolithic admixture picked up by the steppe people on their way to Iberia."
Am I reading this wrong, or does your model not suggest that Portugal_BA has CHG ancestry above and beyond that of Yamnaya?
Portugal_BA
CHG 0.106±0.048
EHG 0.042±0.042
Portugal_LN 0.852±0.042
P-value 0.0367007784
chisq 14.946
So I'm curious to see if you can nail down whether or not there is in fact an excess of CHG compared to Portugal BA, and if Minoan is a reasonable proxy. (I'd be curious if Portugal LN had an excess of CHG though that's really here no there).
If there is it would lend weight to Dene-Caucasian as a macrofamily, and it would at least be interesting. It would make the Bell Beakers complex to understand but life is complex.
Davidski
"Eastern European Hunter-Gatherers didn't carry any CHG, and, also, for instance, there's no plausible cultural mechanism by which a Hunter-Gatherer group could have migrated in such large numbers across Europe."
i don't know how plausible it is but one possible way to explain "such large numbers" is maybe they weren't large numbers at the time i.e. if it happened at a time when small numbers could have had the same proportional effect e.g. interior HGs (EHG) moving towards coastal refuge regions (WHG) because they couldn't survive in the interior any more for some reason.
so small numbers of ex mammoth steppe people moving into maritime refuges like north Italy, Franco-Cantabria or Black Sea coast *before* farming
@Grey - "i don't know how plausible it is but one possible way to explain "such large numbers" is maybe they weren't large numbers at the time i.e. if it happened at a time when small numbers could have had the same proportional effect e.g. interior HGs (EHG) moving towards coastal refuge regions (WHG) because they couldn't survive in the interior any more for some reason."
They could also have been there for a while. We don't really know more than the broad strokes of the distribution of WHG vs EHG people as the ice sheets retreated. It's seems pretty plausible that hunter gatherers around the Danube could be fairly closely related to the hunter gatherers right next door around the Black Sea. The Danube really does drain right next door to the EHG homeland.
That being said I think it's dead obvious that central European Bell Beakers picked up ancestry from CWC culture and then spread it across Europe. More interesting IMHO is figuring out if there were other bits of ancestry they picked up too.
Andre,
I am well aware of the agenda on ItalicRoots: to exoticize the Iberians and downplay copious post-Neolithic influence in southern Italy from the Near East. They are also in denial about CHG being basal-rich and equate it with WHG, for example...So tired of repeating my points to them regarding this because they all fall on deaf ears.
Nevertheless, we should not employ any underhanded Marxist "banning" tactics and evaluate their individual statements for accuracy(aside from the obvious trollish points which Davidski excises anyway). And to be fair, ItalicRoots is not an obnoxious spammer like xyman.
Anyway, (and this is also for @SamuelAndrews), they are not quite correct. If Iberians have 10% Berber admixure AND copper-age stuff from West Asia, they'd be more downshifted on the PCA than they are(relative to Basques and Copper Age Iberians). The West Asian admixture is there, but not so huge and mainly in the Guadalquivir delta area: south Portugal, Extremadura, and parts of West Andalusia.
@Ryan,
The R1b1a rich Danubian HGs are basically pure WHG with little EHG. They won't fit well as an EHG ancestor of Portugal BA.
@Ryan
Am I reading this wrong, or does your model not suggest that Portugal_BA has CHG ancestry above and beyond that of Yamnaya?
Not really, if you consider the standard errors.
Another thing to consider is that the steppe ancestors of Portugal_BA may have come from west of Samara, like what is now Ukraine, in which case their forager ancestry would be closer to WHG along the WHG > EHG cline, causing a lower ratio of EHG and higher ratio of Portugal_LN in that model.
I wonder how the Samara R1b Samples compare with the earlier Mesolithic R1b samples at Derievka ?
@Sam - "The R1b1a rich Danubian HGs are basically pure WHG with little EHG. They won't fit well as an EHG ancestor of Portugal BA."
Gotcha. IIRC KO-1's HG ancestry is 90% WHG and 10% EHG too.
@David - "Another thing to consider is that the steppe ancestors of Portugal_BA may have come from west of Samara, like what is now Ukraine, in which case their forager ancestry would be closer to WHG along the WHG > EHG cline, causing a lower ratio of EHG and higher ratio of Portugal_LN in that model."
Good point re: this and the standard errors, but it wouldn't also explain that Hungarian Beaker with excess EHG but 0 CHG, no? I think central Europe may have been a pretty mixed and complex place at this point in time. And if it's technically feasible, I think it's always worth looking at if there is are multiple sources of ancestry in a group just to better understand them, even if we know pretty well what the big pieces are.
@Ryan
What Hungarian beaker has excess EHG but zero CHG? I might have missed that one.
@Ryan said...
"They could also have been there for a while."
yeah i think Davidski's numbers point is valid so to me the logical possibilities are:
1) small numbers of EHG when WHG were small numbers also (pre-farming)
2) some peripheral yamnaya with less CHG
or
3) the lower CHG thing isn't actually a thing for some reason
#
"That being said I think it's dead obvious that central European Bell Beakers picked up ancestry from CWC culture and then spread it across Europe."
i think there was a big wave from the steppe (or multiple waves in multiple forms) but also maybe some oddities here and there - not necessarily important in the big picture but interesting from a story point of view.
@epoch2013 - I may have been getting myself confused with Bronze Age samples.
From Martiniano 2017:
A recurring feature of ADMIXTURE analyses of ancient northern Europeans is the appearance and subsequent dissemination within the Bronze Age of a component (teal) that is earliest identified in our dataset in HGs from the Caucasus (CHG). Unlike contemporaries elsewhere (but similarly to earlier Hungarian BA), Portuguese BA individuals show no signal of this component, although a slight but discernible increase in European HG ancestry (red component) is apparent. D-Statistic tests would suggest this increase is associated not with Western HG ancestry, but instead reveal significant introgression from several steppe populations into the Portuguese BA relative to the preceding LNCA (S4 Text, S6 Table).
It's ADMIXTURE so take it with a grain of salt of course, but Simon here did a qAdm run and got similar results:
https://docs.google.com/spreadsheets/d/1g-cEJmS9cZ6r-_9p0qj1f2qszxiJbactk89DOd0TlHg/edit#gid=1320636554
That bronze age samples is from the Mako culture, which immediately preceded the Bell Beakers around Budapest. Simon has them as 55% Anatolian Neolithic, 30% WHG and 15% EHG.
Olade doesn't test for EHG separately. Their ADMIXTURE suggests there could be some samples maybe with extra EHG - Bud1 or Szi maybe. Szigetszentmiklós is an interesting site because there are samples there ranging from 0% to 74% steppe ancestry that are contemporaries, so you are actually seeing the process of locals being assimilated into Bell Beaker culture.
Which is what I find weird about the suggestion that R1b in Bell Beakers came from Yamnaya. So we're meant to believe that Beaker women completely and utterly culturally dominated Yamnaya men, and then the new culture somehow flipped back and became patrilocal?
@Grey - the numbers thing could also just be an extended period of time though - long term contacts up and down the Danube or across the Carpathians. WHG ancestry was also on the rise across Europe at this time so the HG communities may have been more dynamic and organized than we would assume.
@Anthrosurvey
"They are also in denial about CHG being basal-rich and equate it with WHG, for example..."
Well that was one big funny lie from you. There was a big Copper/Bronze migration of CHG rich population who affected South Eastern Europethe the most but also South Western one. Iberians have also some of it and that's why they are different from the Basques. But we need to wait for more papers to be sure. The Berber/African score is 15% in Spaniards as for Lazaridis 2015 and 4%-20% as for Boatigue 2013. Just numbers I got from papers.
@Ryan
Thanks. You refer to the Hungarian EBA, strange indeed.
ItalicRoots,
They are different in large part due to North African influence(which Basques lack).
I didn't deny the possibility of CHG-rich copper age influence in Iberia. I merely implied it was not so significant as in S.Italy and the Balkan sphere. It also isn't so ubiquitous as North African DNA, and mainly restricted to Guadalquivir basin area. Perhaps points along the Catalan coast as well.
15-20% range is typical for Canary islanders. 10% is more typical for mainland Iberians, sometimes hitting highs of 15% and lows of 5%.
Post a Comment