March 08, 2012

A first look at the genome of the Tyrolean Iceman

Thanks to the publication of the Tyrolean Iceman's (reconstructed on the left) genome sequence, by Keller et al. (2012), I have been able to include him in a joint analysis with 2,671 other individuals, previously assembled for the K7b and K12b calculators of the Dodecad Project.

I have essentially repeated the unsupervised ADMIXTURE analysis from scratch, with two differences:
  • The dataset now includes Oetzi, and
  • The number of markers has been reduced to 44,435 by intersecting the 166,770 SNPs used in K7b/K12b with the SNPs available for Oetzi
I had predicted that Oetzi would turn out "Mediterranean" in his genomic profile, and the published analysis seems to have confirmed this by discovering that his closest modern kin are Sardinians, a Mediteranean population. Sardinians are usually modal for the various "Mediterranean", and "Atlantic_Med" components inferred by the Dodecad Project.

Thankfully, the ADMIXTURE analysis in both the K=7 and the K=12 case, resulted in the same solution as in K7b and K12b, despite the inclusion of Oetzi and the fact that a smaller number of markers were used. There are slight variations in the admixture proportions compared to the K7b/K12b numbers, and an expected increase in "noise levels" in the minor components, but the overall solution appears to be the same.

The results for Oetzi and the reference populations, including Dodecad populations with 5+ members are included in the spreadsheet.


Below, I will briefly comment on them.

K=7


Oetzi turns out to be 51.9% "Southern" and 43.1% "Atlantic_Baltic" in the K=7 analysis, with noise levels of the other components. The salient point is that he seems to be lacking the "West Asian" component, unlike most Europeans, except Basques and Sardinians, who have:

Basques: 27.6% "Southern" and 69.5% "Atlantic_Baltic"
Sardinians: 46.2% "Southern" and 52% "Atlantic_Baltic"

So, Oetzi does appear to be most Sardinian-like in this analysis, and indeed to be a little more "Southern" than extant Sardinians. This is consistent with Keller et al. (2012) which finds him to cluster with Sardinians and to be a bit more "southern" (in PCA space) than Sardinians.

K=12

Oetzi turns out to be 57.7% "Atlantic_Med" and 22.3% "Caucasus" in this analysis. Once again, this makes him quite similar to Sardinians who are 64.9% "Atlantic_Med" and 21.3% "Caucasus". Two components are salient in their absence: the "Gedrosia" component which is present in non-Sardinian modern Italians, and the "North_European" component which is also present in the same.

Discussion


If we compare Oetzi with modern Europeans, he certainly appears to be most Sardinian-like. If we compare him to modern Italians (the samples N_Italian_D, North_Italian, as well as the neighboring Tuscan, TSI30, and C_Italian_D) populations, he appears to possess many of the same components (including Atlantic_Med, Caucasus, Southwest_Asian, and Northwest_African in non-trace quantities).

This of course does not indicate that Oetzi was of "mixed" ancestry, since ADMIXTURE components are but imperfect reflections of ancestral populations. It does, however, indicate that his profile consisted of many of the same elements found in Italy today, but in different proportions. Moreover, any additional or "exotic" ancestry in Oetzi cannot have existed in substantial amounts so as to render the inferred proportions (in terms of modern populations) nonsensical.

The three exceptions to this rule are the "West_Asian" component in K=7 and the "Gedrosia"/"North_European" components in K=12, which are conspicuous in their absence. And, given that all three of these have their present-day center of weight in more eastern longitudes, we can be fairly certain that present-day Italians have been affected by people from the east, by one or more routes.

This is, again, reasonable, and in agreement with the observation of Keller et al. (2012) -pictured on the left- that the Iceman occupies a "western"/"southern" position relative to modern Italians, and is consistent with the idea that modern Italians were formed by an eastward shift (in the case of more southern Italians) and a northeastward shift (in the case of more northern Italians).

We can have near-absolute certainty that a "North European" element was present in northernmost Italy by at least the first half of the first millennium BC, the time of the Hallstatt culture of Central Europe and of historical references about the arrival of Celts in Cisalpine Gaul. Oetzi is dated to the latter half of the 4th millennium BC. Hence, there is an intervening period of about 2.5k years between the two; it will be interesting to sample individuals from that time period to see when -within that time interval- the arrival of this trans-Alpine element occurred. It will also be very interesting to sample individuals from further north, east, and west, to ascertain the full spatial extent of "Oetzi-like" individuals during the early Copper Age.

Similarly, we can be certain that more "eastern" elements were present in Southern Italy by at least the time of the Greek colonization, which occurred at much the same time as the arrival of the Celts in the north. However, since all Italians are eastward shifted relative to Oetzi, even those from furthermost north, it is likely that the eastward shift of modern Italians preceded the historical colonization and that some of the myths preserved by Roman and Greek writers about eastern origins of various pre-archaic Greek and pre-Etruscan Italian peoples may hold an element of truth.

At present, we are disadvantaged by the fact that we are trying to place ancient individuals within a modern genetic landscape. Oetzi's genome suggests the fluidity of this landscape as we move into the past. Thankfully, it is all but certain that we will soon obtain full genome sequences from other prehistoric and historic individuals from different spots in space and time, so our understanding of how West Eurasians came to be is only bound to improve.

25 comments:

Dienekes said...

@ Edwardo Pinto,

You put the numbers in the wrong order in Dodecad Oracle.

The correct results:


> DodecadOracle(c(0,0,5.7,2,57.7,0,1.5,2.4,7.6,0.7,22.3,0))
[,1] [,2]
[1,] "Sardinian" "13.8152"
[2,] "Andalucia_1KG" "22.8797"
[3,] "Murcia_1KG" "24.5412"
[4,] "Canarias_1KG" "25.2208"
[5,] "Baleares_1KG" "26.4044"
[6,] "Castilla_La_Mancha_1KG" "26.643"
[7,] "North_Italian" "27.0747"
[8,] "Aragon_1KG" "27.1599"
[9,] "Castilla_Y_Leon_1KG" "27.2406"
[10,] "Valencia_1KG" "27.2947"



[1,] "80.8% Sardinian + 19.2% Egyptans" "4.3119"
[2,] "72.7% Sardinian + 27.3% Morocco_Jews" "4.7931"
[3,] "81% Sardinian + 19% Jordanians" "5.4624"
[4,] "18.8% Palestinian + 81.2% Sardinian" "5.5228"
[5,] "83.3% Sardinian + 16.7% Yemenese" "5.6492"
[6,] "74.4% Sardinian + 25.6% Sephardic_Jews" "5.7106"
[7,] "80.9% Sardinian + 19.1% Lebanese" "5.7583"
[8,] "81.4% Sardinian + 18.6% Syrians" "6.1189"
[9,] "24.4% Ashkenazi_D + 75.6% Sardinian" "6.3367"
[10,] "20.1% Algerian_D + 79.9% Sardinian" "6.3728"


But, actually the K12b Oracle should not be used in this case, as the admixture proportions for it are derived from the full ~167k markers, whereas Oetzi's have been calculated on only ~44k.

If anyone knows how to use R, they can modify it quite easily by changing the X table to reflect the new frequencies of the new K=12 analysis.

Dienekes said...

Deleting these two comments, as you've realized the mistake and our results now agree.

Maju said...

I downloaded the spreadsheet and, yes, on first sight, Sardinians are the ones who share closest apportions of the Mediterranean ("Southern") and Atlantic ("Atlantic-Baltic") components. But, visually scrolling around, I found that maybe Sicilians had a better Med/Atlantic ratio, one closer to that of Ötzi (1.2).

So I got the spreadsheet to work and effectively, Sicilians have one of the closest Med/Atl ratios to Ötzi (after Ashkenazim and Turkmens). Excluding those with very low levels of Atl and Med components, the closest Atl/Med ratios are (Otzi's +/-20%):

Ötzi: 1.20
Ashkenazi Jews: 1.19
Ashkenazi-D: 1.15
Sicilian-D: 1.10
S.Italian-Sicilian-D: 1.07
Kumyks: 0.97

Further away (in the - 50% zone) score other Italians (including Sardinians (0.89) but excluding North Italians), Greeks and several Caucasus groups. In the +50% zone are Turks, Sephardic Jews and Moroccan Jews.

In brief: Sicilians (but also other peoples) look a lot like Ötzi plus other admixture. This other component is invariably the West Asian one, which is in the 20-30% zone for Italians, Greeks and Jews and in the 40-55% zone for the rest. The only exception is Sardinians, who have almost no third component (remained "pure" since those times probably) but have a Med/Atl ratio not as good as that of Sicilians.

Dienekes said...

That is consistent with my observation in the previous post that Sardinians differ from Otzi in an Iberian direction, while Italians in a Balkan/Anatolian one.

Westgoth said...

I have one (maybe silly) question: Is it possible to determine the Y-STR's from the genome (by simply browsing it)??

Dienekes said...

I believe not, because of the short reads used to assemble the genome, but I'm no expert on Y-STRs to tell for sure.

anthrospain said...

I would like to see other plots with Ozi on it, other than the PCA from Keller et al.

royking said...

Great work Dienekes!
Your results add more weight to the proposition that:
1) Oetzi was an admixture of Mesolithic ( Atlantic-Med) and Neolithic ( SW Asian and Caucasus) components. he does not differ much from Sardinians and Italians in these proportions.
2) Oetzi lacked the European lactase persistence allele (along with LBK samples) and, accordingly lacked the Northern European component which likely expanded into Northern Italy after Oetzi.

Dienekes said...

I see no particular reason to assign the entirety of the Atlantic_Med component to the Mesolithic substratum. Its divergence with the Caucasus component has been estimated by me as 9.9ky

http://dienekes.blogspot.com/2012/02/timeline-of-human-prehistory.html

So, I consider it as a Neolithic component mixed with a Paleolithic substratum, in an unknown proportion, depending on how divergent the Paleolithic substratum was originally; if we are to go by Fst estimates of early Neolithic LBK vs. Mesolithic Europeans, pre-Neolithic Europeans were quite divergence from incoming Neolithic peoples, hence, the 9.9ky divergence estimate corresponds to a mainly Neolithic population that has absorbed a minority dose of Paleolithic Europeans.

aspromavro said...

Dear Dienekes, could you create an Oracle edition based on this with Otzi in it? Even better with a three-population mixedmode option. Would that be feasible?

Maju said...

I would say that those estimates about the age of divergence of the West Eurasian sub-components are most arguable, Dienekes.

If, as it's the case, Fst distance between West Africans and West Eurasians (with unsupervised Admixture) is of the order of 0.148-0.214 (avg. 0.187), between Siberians and West Eurasians is of 0.120-0.160 (avg. 0.147), and between West Eurasian sub-components of 0.060-0.140 (rough avg.0.100)...

... then we can estimate that WEA subdivisions are (very roughly) half as old as the OoA migration (actually a bit less than half). The time line of the OoA migration is still a bit uncertain but there is quite strong archaeological evidence for c. 90-80 Ka. ago (it could be older but hardly more recent). So it's legitimate to estimate the WEA subdivisions are on average from c. 40-50 Ka ago.

I know you use a more complex formula, making assumptions about effective population size but those assumptions are hard to justify. Still you do get most ages as pre-Neolithic and then claim that the estimates are "inflated".

I just cannot concur. For me the only evidence of Neolithic genetic flow in Europe is when we see a K=2 Europe vs. West Asia (what comes naturally when samples are balanced or at K=4 when Siberians and North Africans are included) and we see how the West Asian component penetrates in Europe, notably the Balcans - but even Basques have some of it (Lithuanians almost not but they have some Siberian influences instead). In this linked graph the red (West Asian) component represents the Neolithic input, while the purple (European) component represents the Paleolithic persistance, probably, which is dominant.

Of course, once the Neolithic was established in the Balcans, most of what it carried on to other lands was European "Paleolithic" components, so it's not as straightforward. But this European recycled Paleo-into-Neo blood is not what you mean, do you?

Dienekes said...

@Maju,

Fst does not increase linearly with age, so while the comparisons you make are useful for ballpark figures, they are not as accurate as the ones using the formula.

It is also important to take effective population size into account; small populations drift strongly (high Fst), which would inflate divergence times, but plugging in the effective size corrects for this factor. This is most evident in the case of Papuans in my own calculations.

As for your K=2 Europe vs. West Asia comment, it is fallacious to associate these components with "Neolithic" vs. "Paleolithic". This is most evident in the Reich et al. analysis of Indians, where it was shown that populations of southernmost India that tend to score almost completely "South Asian" in ADMIXTURE analyses actually possess a hefty dose of Ancestral North Indian.

0% of an admixture component X indicates minimum influence from X within the studied dataset, it does not mean no influence from X.

Justin said...

what goofball made his arms?! why do these recreations always suck so bad?

Maju said...

Do you know which are the relevant Fst distances of the Kalash population? The Kalash are an excellent example of your claim: "small populations drift strongly". I feel that this is correct to some extent but not as much as you seem to think.

I'd test that myself today but I'm busy right now and you may know the answer.

"This is most evident in the case of Papuans in my own calculations".

I'm sorry but I'm not aware of those calculations. Do you get figures older than 50Ka? Because that's the minimal age for the colonization of New Guinea.

If so, I wonder what ages would we get without correction. I have not worked with Papuans either but the Siberian component already gives a vibe (by my admittedly simpler method) of 60-70 Ka ago for East-West Eurasian divergence (what is coincident with Liujiang and some other evidence).

"This is most evident in the Reich et al. analysis of Indians, where it was shown that populations of southernmost India that tend to score almost completely "South Asian" in ADMIXTURE analyses actually possess a hefty dose of Ancestral North Indian".

I have not read the paper because it is pay per view but this is the first time I heard any such claim. I have been following Harappa Ancestry Project for quite a while and I have always seen both South and North South Asians having both 'Indian' components. While there are some individuals (invariably Iranians) who show almost no "South Asian" (ASI) component, all show some "Baloch-Caucasus" (ANI) component (example).

This is not consistent with what you say.

Dienekes said...

>> This is not consistent with what you say.

http://dienekes.blogspot.com/2011/03/note-of-caution-on-admixture-estimates.html

The Mala have 1.1% of the "West Eurasian" component and 38.9% of ANI.

The Mala have _minimum_ Caucasoid admixture, but that admixture is not zero, but about 40%.

Similarly, certain Europeans that show ~0% "Near East" ADMIXTURE components in a K=2 analysis are not necessarily descended mainly from Paleolithic Europeans as you presume.

>> I'm sorry but I'm not aware of those calculations. Do you get figures older than 50Ka? Because that's the minimal age for the colonization of New Guinea.

http://dienekes.blogspot.com/2012/02/timeline-of-human-prehistory.html

The effective population sizes can be found in Li et al.; estimates of Fst divergence can be found in world9 calculator

http://dodecad.blogspot.com/2011/12/world9-calculator.html

jeanlohizun said...

Dienekes:
For the sake of the argument I’m going to ignore the fact that unsupervised ADMIXTURE runs are highly dependent in the populations being sampled, as well as the number of participants of each population. In your entry found here: http://dienekes.blogspot.com/2012/02/timeline-of-human-prehistory.html you mentioned that you used the following formula to estimate the time of divergence in generations:

t=23*log(1-Fst)/log(1-1/(2*Ne))

You said you used the following set of populations for the as modal for “six” Caucasus: Adygei; Atlantic_Med: French_Basque; North_European: Orcadian; Southwest_Asian: Bedouin; Northwest_African: Mozabite; Gedrosia: Brahui. You also said you used the data from Li et al(2008) found in supplementary table-1 to find the Ne values.

(http://www.sciencemag.org/content/319/5866/1100/suppl/DC1)

I’m don’t know how you arrived to the conclusion that you could essentially average out the effective population size of two populations in a linear form, but I’ll take your word for it, for the sake of the argument. Why you chose to use 23 years is also puzzling, given that it is “OK” if we are talking about long term human generations, but is it t < 20 kya really considered long time. Nonetheless let’s stick with it for the sake of argument. Using the data from Li et al(2008) we find the following Ne values for the chosen populations. French Basque (6137), Orcadian (5379), Bedouin (9002), Adygei (6699), Mozabite (7014), Brahui (7948).

Now I understand why you chose Adygei to be modal for Caucasus instead of Georgian who exhibit this component at 73.9%, and this is because there is no data for the Ne of Georgians in Li et al(2008). Now, what I don’t understand why you chose the Orcadians as the modal for North European, when in fact the Russians from HGDP exhibit this component at 65.4%, whereas the Orcadians from HGDP exhibit it at 45.6%.

This can all be seen here:

https://docs.google.com/spreadsheet/ccc?key=0ArJDEoCgzRKedEY4Y3lTUVBaaFp0bC1zZlBDcTZEYlE#gid=0

Also unlike the Georgians, the Russian from HGDP do have the Ne listed on the Li et al(2008) study. So I went ahead and calculated the divergence time of Caucasus(Adygei) vs. North European(Russian). Now the Ne of the Russian is 7034. So the average Ne would be 6867. This means that:

T=23*log(1-0.041)/log(1-1/13733)=13.2 kya

So just by changing the modal population from Orcadians to Russians we see that the divergence time increases to 13.2 kya.

Now the Atlantic_Med is almost equally distributed amongst Sardinians(70.5%) and French Basque(73.1%). Now using the Sardinians (Ne=6624) instead of the French Basque changes the Atlantic_Med-vs.-Caucasus divergence from 9.9 kya to 10.3 kya which is not a big change.

Dienekes said...

when in fact the Russians from HGDP exhibit this component at 65.4%

HGDP Russians are ~10% admixed with Mongoloids.

https://docs.google.com/spreadsheet/ccc?key=0ArAJcY18g2GadHZ6SHpiLTNTa3lsUmZJY2pQblVRR2c

I’m don’t know how you arrived to the conclusion that you could essentially average out the effective population size of two populations in a linear form, but I’ll take your word for it, for the sake of the argument.

It's an approximation; the effective size of the ancestral population is unknown; extant populations may have much lower apparent effective size -due to drift-, or much higher apparent effective size -due to admixture.

Dienekes said...

I also strongly suspect that the Adygei Ne is biased upwards. It appears to be (6699) about halfway (6707) between relative pure Caucasoids (Basques/Sardinian = (6137+6624)/2) and (Russians=7034)

and according to K7b, the Adygei are 5% Mongoloid, whereas the Russians are 10%.

jeanlohizun said...

Dienekes said:

HGDP Russians are ~10% admixed with Mongoloids.

I presumed you are implying that because of this admixture HGDP Russians would have a bigger Ne than that of the ancestral North European population. Fair enough!! But aren’t the Adygei also admixed with “Mongoloids”.

Putting that aside, the Atlantic_Med presence in both French Basque and Sardinian got me thinking:

Back in April 2011 you did an ADMIXTURE analysis where at K=9 both French Basque and Sardinian components formed from the South European component, even before the North/Central European component separated into Northwestern and Northeastern European.

http://dienekes.blogspot.com/2011/04/genetic-structure-of-west-eurasians.html

At K=9 we see that both the French Basque and Sardinian components are such that using the Ne for French Basque (Ne=6137) and Sardinians (Ne=6624) is more accurate in determining the time of divergence of their respective components than using either the Ne of French Basque or Sardinian to determine the time of divergence of Atlantic_Med.

Here is the K=9 population painting, and the Fst distances between the components

http://1.bp.blogspot.com/-YJPxeZCqL1M/TbDJ0Vml8FI/AAAAAAAADiA/LqozZjewXrg/s1600/ADMIXTURE_9.png

Using the Adygei(Ne=6699) as the modal population for West Asian, and Russians (Ne=7034) as the modal population for N/C European we find that the following divergence times using a generation time of 23 years:

French Basque-vs.West Asian 9.9 kya

French Basque-vs.N/C European 9.5 kya

French Basque-vs.Sardinian 9.2 kya

Sardinian-vs.West Asian 11.2 kya

Sardinian-vs.N/C European 11.8 kya

Now when we look at K=10 we see that the N/C European component has split into two different components NW European and NE European. The Orcadians(Ne=5379) can be used as modal for the NW European component, while the Russians can be used for the NE European modal.

http://2.bp.blogspot.com/-VO3qZEQYilY/TbDLtgVFvcI/AAAAAAAADiI/kdvqPTQaO-Q/s1600/ADMIXTURE_10.png

These are the divergence times we find using a generation length of 23 years:

French Basque-vs.West Asian 11.4 kya

French Basque-vs.NW European 8.1 kya

French Basque-vs.Sardinian 9.5 kya

French Basque-vs.NE European 10.2 kya

Sardinian-vs.West Asian 11.8 kya

Sardinian-vs.NW European 8.7 kya

Sardinian-vs.NE European 12.5 kya

NW European-vs.West Asian 8.2 kya

NE European-vs.West Asian 11.3 kya

NE European-vs.NW European 6.4 kya

So it seems that either some of what was being characterized as French Basque or West Asian on the K=9 was NW European. In any case, is really hard to notice it by just eyeballing the population paintings as the French Basque painting doesn’t really change from K=9 to K=10, perhaps Dienekes could provide some percentages.

jeanlohizun said...

Continue...
Then when we go to K=11

http://2.bp.blogspot.com/-Wk9_ST1CKC4/TbFOmpaIlGI/AAAAAAAADiQ/jhw64bHM1G4/s1600/ADMIXTURE_11.png

We see that some consistency:

French Basque-vs.West Asian 11.4 kya (Did not change)

French Basque-vs.NW European 7.8 kya (Went down from 8.1 kya)

French Basque-vs.Sardinian 9.5 kya (Did not change)

French Basque-vs.NE European 10.5 kya (Went up from 10.2 kya)

Sardinian-vs.West Asian 12.2 kya (Went up from 11.8 kya)

Sardinian-vs.NW European 8.7 kya (Did not change)

Sardinian-vs.NE European 12.8 kya (Went up from 12.5 kya)

NW European-vs.West Asian 7.9 kya (Went down from 8.2 kya)

NE European-vs.West Asian 11.6 kya (Went up from 11.3 kya)

NE European-vs.NW European 6.9 kya (Went up from 6.4 kya)

In terms of time of divergence what we see is that from the point of view of:

French Basques: West Asian>>NE European>>Sardinian>>NW European

Sardinian: NE European>>West Asian>>French Basque>>NW European

NW European: Sardinian>>West Asian>>French Basque>>NE European

NE European: Sardinian>>West Asian>>French Basque>>NW European

West Asian: Sardinian>>French Basque>>NE European>>NW European (For K=10)

West Asian: Sardinian>>NE European>>French Basque>>NW European (For K=11)

I want to see what other people think about this. In my opinion many theories come to mind, involving Magdalenian expansions, Neolithic expansions, etc.

Dienekes said...

I presumed you are implying that because of this admixture HGDP Russians would have a bigger Ne than that of the ancestral North European population. Fair enough!! But aren’t the Adygei also admixed with “Mongoloids”.

See above post.

And, yes, they are, and this shows that the divergences of the West Asian component with the others may in fact be slightly younger, further reinforcing my main thesis.

In fact, even without accounting for admixture we're only really getting at most early pre-Neolithic dates. And, if we account for admixture (which inflates Ne, Fst, and t) we will get even younger dates for the divergence of populations.

It's pretty clear to me that all West Eurasians are derived from a common source in a Neolithic and later timeframe; the absorption of Paleolithic elements was not sufficient to change that basic fact.

Dienekes said...

* late pre-Neolithic

royking said...

I think that Dienekes and others are losing sight of the evidence that that may have been a major Mesolithic migration from the Caucasus area to Europe/North Africa as evidenced by trapezoidal microlithics that are post-LGM and perhaps even post-Younger Dryas. In West Europe this phase is often termed the Castelnovian horizon (after Chateauneuf les Martigues in Provence) circa 6000-7000 BCE. It predated the Neolithic and the lithics were widespread--including the Mesolithic at Franchthi cave. Some archaeologists joke about the connections between the Caspian in North Iran and the Capsian in North Africa. IMO it is not improbable that Mesolithic people migrated during climatic amelioration. Such a demographic interconnection could explain the low dates based on Fst's from Dienekes analyses without invoking a Neolithic expansion to explain the temporal patterns.

jeanlohizun said...

Dienekes said:
It's pretty clear to me that all West Eurasians are derived from a common source in a Neolithic and later timeframe; the absorption of Paleolithic elements was not sufficient to change that basic fact.

It's not all that clear to me, in fact if we assume a model as you predict where a common Neolithic source gave rise to all West Eurasians, then we will have to assume that Magdalenian expansions did not influence West Asia, and that Paleolithic European populations had been isolated from West Asian populations for a very long time. In fact what the early pre-Neolithic dates suggest to me, is that there was an ongoing gene flow from Iberia/SW France before the Neolithic, which got cut off at the On-set of the Neolithic, this seems as if the Proto-Basques, Proto-Sardinians, Proto-West Asians, Proto-NW European and Proto-NE European were all having a continued gene flow 10-12 kya. Then something happened that cut off the Sardinians from the Proto-NE European, and the Proto-West Asian 12 kya. The Proto-Basques got cut off first from the Proto-West Asian , and then from the Proto NE European 10-12 kya. This looks to me as if the advancement of agriculturist drove the Proto-West Europeans like Europeans back to Western Europe, where as divergence time suggest the Proto-Basque continued interacting with Proto-Sardinians and Proto-NW Europeans, until the Sardinians got cut-off about 9.5 kya, and then the NW Europeans at 8 kya. The fact that the divergence time of Proto-NE European is in the same order as that of Proto-Basque suggest to me that they both remained relatively isolated from incoming Proto-West Asians, but got separated. Given that the Proto-NW European appear to diverge from the Proto-West Asians at a much later date than both Proto-Basques and Proto-Sardinians, this suggest that the former were possibly affected by the incoming Neolithic wave, which also coincides with the separation dates of the Western European components. Finally some gene flow amongst Proto-NE Europeans and Proto-NW Europeans brought these two components together.
My thoughts are that during Magdalenian times there was an ongoing gene flux between all West Eurasian populations, then the agriculturists move to the Balkans 10 kya, and the Proto-Western Europeans and Proto-Eastern European(Baltics) get separated. Proto-Eastern Europeans start accumulating divergence from Proto-Western Europeans. Agriculturist move into Central and Western Europe in the 6-9 kya time frame, and Proto-Basque contract a lot. However the fact that these Agriculturist probably acquired some Balkanic blood means that the interaction between them and Proto-Western Europeans not only would bring the Proto-NW European closer to Proto-West Asian, but also to Proto-NE European. Finally a direct gene flow from Proto-NE European to Proto-NW Europeans bring them closer together, while separating the Proto-NW European component from the Proto-Basque and Proto-Sardinian components. Very likely the actual separation date of Proto-NW Europeans and Proto-Basques is more recent, but given that the Proto-NW Europeans received influence from Proto-West Asians, and Proto-NE Europeans these influences drove the time of divergence upwards.

Dienekes said...

It's not all that clear to me, in fact if we assume a model as you predict where a common Neolithic source gave rise to all West Eurasians, then we will have to assume that Magdalenian expansions did not influence West Asia, and that Paleolithic European populations had been isolated from West Asian populations for a very long time.

That is in fact what we observe in the mtDNA, a relatively pristine U-bearing gene pool and Fst distances with both incoming LBK and extant Europeans comparable to what is found in inter-continental comparisons today.

The same with Y-chromosomes: absence of many of the extant Y-chromosome groups in present-day Europeans, down to the Copper Age, and nothing that suggests persistence of a Paleolithic substratum in the population.

@royking

Perhaps there is such a Mesolithic layer in Europe. The point is, however, that Neolithic and later technologies facilitated the demographic dominance of the incomers over the pre-existing populations, even if some of them had arrived from the same area a few thousand years prior to the onset of the Neolithic. Late pre-Neolithic hunters may have spread into Europe, and perhaps they did so widely, but they didn't have the tech to grow in large numbers.

My comments re: gene pool of Mesolithic Europeans above are also pertinent to the question.

A good analogy is with the westward push across North America. There were small groups of explorers and homesteaders that headed west with horse-technology, but it was the railway that facilitated the full-scale colonization of the West.

Similarly in Siberia, Russian settlers began moving east centuries before the Industrial Revolution, but it was the railway that opened up the region for full-scale colonization.

The same thing may have happened in Europe; different groups of hunters may have drifted there from West Asia, since the 40,000s downto the late pre-Neolithic (albeit no evidence for these groups exists as of yet, due to the largely uniform U-dominated mtDNA gene pool). But it was a combination of new technologies: farming, secondary products, ox- and horse-driven transport, metalworking, etc. that gave the edge to newcomers from the east, just as it was technology that gave the advantage to Europeans moving into thinly populated areas more recently (such as North America, Siberia, or Australia).