search this blog

Thursday, June 30, 2016

The Natufian puzzle


One of the biggest surprises in the new Lazaridis et al. preprint is that the Natufians don't show any Sub-Saharan African admixture when poked and prodded directly with formal statistics. However, TreeMix, which runs on formal statistics, doesn't have much trouble finding Sub-Saharan or related ancestry in both the Natufians and Neolithic farmers from the Levant. So what's going on?

Also worth noting in the fourth graph is the heavy migration edge from near the base of the East Eurasian Dai/Han branch to the early Neolithic farmers from Iran. My bet is that this is a signal of admixture from a Central Asian forager population, perhaps representing a parallel clade to that of the East Asians? Or is it just Ancient North Eurasian (ANE) admixture? I'll try and investigate that soon.







108 comments:

Onur said...

@Davidski

Also worth noting in the fourth graph is the heavy migration edge from near the base of the East Eurasian Dai/Han branch to the early Neolithic farmers from Iran.

It is from the common ancestors of West and East Eurasians. It is a substitute for the divergent position of the Neolithic farmers from Iran in comparison to the other included West Eurasian populations at the other migration edge levels of this TreeMix analysis.

Ariele Iacopo Maggi said...

Off topic once again. I tried to test this Italian/Greek connection with Anatolia Chalcholithic. Here we go! Italian_south seems to prefer Armenia Chalcolithic over Anatolia Chalcolithic...
-Italian_South-
"Remedello_BA" 33.15
"Armenia_Chalcolithic" 32.45
"Iran_Chalcolithic" 13.7
"Levant_Neolithic" 9.45
"Yamnaya_Kalmykia" 8.6
"Israel_Natufian" 1.4
"Ami" 1.1
"Yoruba" 0.1
"Karitiana" 0.05
"Anatolia_Chalcolithic" 0
"Iran_Neolithic" 0
"LaBrana1" 0
"Satsurblia" 0
"Villabruna" 0
"Yamnaya_Samara" 0
"LBK_EN" 0
"Anatolia_Neolithic" 0
"Bell_Beaker_Germany" 0
"Iran_Late_Neolithic" 0
"Hungary_HG" 0
"Iberia_EN" 0
"Iberia_MN" 0
"Somali" 0

Poise n Pen said...

It's only surprising if you believe in recent Out of Africa fairy tale and only if you think that implies that african at that time equated to modern black africans.

capra internetensis said...

@poise N pen

Wtf does it have to do with either of those things?

ryukendo kendow said...

@ Huijbregts

(Pulling the comment here for visibility.) Huij, I'm not sure that removing so many dimensions will be a good idea. As I mentioned before, minor distortions of populations towards the columns should be expected to provide information about the geographic location/identity of the contributing populations.

I went to check out this idea, by analysing how populations behave when given three columns: Anatolia_Neolithic, Iberia_EN, and Bichon. If we assume that only deep ancestry matters, then Iberia_En is just a combination of Anatolia_Neolithic plus Bichon, and the inclusion of Iberia_EN should not be expected to add any new information. But this was not the case.

Here is PC1 vs PC2, which is indeed dominated by WHG vs ENF:

PC1 vs PC2
Zoomed in:
PC1 vs PC2 (zoomed)

Here, Iberia_EN is just intermediate between Anatolia_En and Bichon. But PC3 is very different:

PC1 vs PC3
Zoomed in:
PC1 vs PC3
Zoomed in even more:
PC1 vs PC3

This is not a WHG-vs-ENF distinction, because Ukrainian-West, Srubnaya, and Corded Ware Germany, which are WHG-rich, are shifted towards Anatolia_Neolithic while Sardinian, French South and Italian Bergamo, which are WHG-poor, are shifted towards Iberia_EN.

The most Iberia_En shifted populations are Iberia Chalcolithic, Iberia_EN, Sardinian, Basque, Spanish, French_South, Icelandic etc, while the most shifted to Anatolia_Neolithic are Hungary_CA, Hungary_HG, Mordovian, Belarusian, Cypriot, etc; and the obvious implication is that these populations share a faint but significant drift path with either Iberia_EN or Anatolia_En due to direct recent contribution of these populations to the moderns as opposed to just similarity in the big six or big four, especially as their ratios of the big four will predict opposite affinities instead.

You are of course right that PC3 is very small, but its precisely the fact that otherwise similar populations exhibit small fluctuations in the similarity to the column populations, that make your tool so powerful. These would, of course, be forced into the higher PCs if the sheet is orthogonalized, while the lower PCs can be expected to produce a proxy of the African_Asian_European 'triangle' pattern in the first three PCs, and also proxies of the the Big Four/Six patterns in deep ancestry in slightly higher PCs.

ryukendo kendow said...

Hmmm, I'm actually tempted by this analysis to suggest to David to create a 'Mass-Monte' scenario, where every single population present in the rows is also present in the columns. Populations with missing diagonals, i.e. where its not possible to calculate the relatedness of the population to itself due to single sample or poor sample quality, can always be dropped as columns when they are used as rows, giving us unprecedented flexibility in choosing any populations as contributors.

A datasheet like this should be extremely information-rich; not only will it explore big four or big six, it should be possible to explore even more recent phenomena. E.g. I would expect modern east Europeans to demonstrate a distortion in the Srubnaya column, demonstrating the recent influence of this population, just like the way SW and W Europeans demonstrate distortion to Iberia_En in the previous analysis. Also, Mediterranean Europeans may be expected to show a distortion towards modern Middle Eastern columns. Etc. Etc...

But I expect, David, that it will be far too much work for you haha.

Krefter said...

"One of the biggest surprises in the new Lazaridis et al. preprint is that the Natufian farmers don't show any Sub-Saharan African admixture when poked and prodded directly with formal statistics."

That is surprising. Natufians fit as part African using your D-stats. Also in ADMIXTURE tests they score in African components. So it really looks like they had some African admixture.

Davidski said...

Yeah, everything I've seen, like PCA and Admixture output, suggests that the Natufians have Sub-Saharan ancestry. But straight D-stats and f4-stats don't reflect anything like that.

Chimp Biaka Anatolia_Neolithic Israel_Natufian -0.000422 -1.539 414749
Chimp Biaka Iran_Hotu Israel_Natufian 0.000981 1.199 70803
Chimp Biaka Iran_Neolithic Israel_Natufian -0.000223 -0.566 367632

Chimp Mbuti.DG Anatolia_Neolithic Israel_Natufian -0.000312 -1.113 481333
Chimp Mbuti.DG Iran_Hotu Israel_Natufian 0.000703 0.906 81688
Chimp Mbuti.DG Iran_Neolithic Israel_Natufian -0.000043 -0.104 425175

Chimp Mota Anatolia_Neolithic Israel_Natufian -0.000734 -1.933 481191
Chimp Mota Iran_Hotu Israel_Natufian 0.000686 0.644 81676
Chimp Mota Iran_Neolithic Israel_Natufian -0.000388 -0.768 425056

Chimp Yoruba Anatolia_Neolithic Israel_Natufian -0.000407 -1.407 414749
Chimp Yoruba Iran_Hotu Israel_Natufian 0.000552 0.654 70803
Chimp Yoruba Iran_Neolithic Israel_Natufian 0.000026 0.063 367632


Ariele Iacopo Maggi said...

Israel_Natufian
"Anatolia_Neolithic" 80.1
"Yoruba" 9.5
"GoyetQ116-1" 8.35
"Papuan" 2.05
"AG3-MA1" 0
"LaBrana1" 0
"Ami" 0
"Dai" 0
"Vestonice16" 0
"Han" 0
"Samara_Eneolithic" 0
distance=0.023084

Karl_K said...

Just because Natufians share some ancestral alleles with Sub-Saharan Africans, that does not mean they could only have acquired them through admixture.

They could simply be alleles that were originally inherited by both populations, but a derived allele became fixed only in the branch leading to Crown Eurasians.

ryukendo kendow said...


Here is a plot of shared drift with Eastern_HG compared to shared drift with Ust-Ishim:

Plot 1
Zoomed

Looking at the position of Iranian_Neolithic, it shares as much drift with EHG as Levant neolithic, Korean, Thai, and Palestinians--not very much, even lower than Lebanese or India_South. But Iran_N is quite asymmetrical between WHG and EHG in formal stats on its own, and these two facts can only be reconciled by EHG ancestry in Iran_N coupled with something else that is lowering Iran_Neolithic's similarity to EHG at the same time, other than Basal Eurasian, since Basal Eurasian is present in other West Eurasians, influencing their position relative to U-I and these samples are an outlier. Iran_N's and Israel_Natufians are relatively similar to Ust-Ishim compared to others, and, surprisingly, even more strongly for Satsurblia...

EHG vs Ami comparison:
Plot 2
Zoomed

Iran_N and Satsurblia are shifted towards Ami as well. Israel_Natufian, on the other hand, is shifted away from Ami.

For the Ust-Ishim vs Ami comparison:
Plot 3
Zoomed

Iran_neolithic, Satsurblia and especially Israel Natufian are disproportionately close to Ust-Ishim vs other Eurasians, while Israel Natufian is disproportionately far from Ami.

The same population X that exists in S Asia, breaking off from W and E Eurasians before they break off from each other, can straightforwardly explain Satsurblia and Iran_Neolithic. Not sure there is any population breaking off from any position in the current tree we have than can increase similarity to Ust Ishim while decreasing similarity to Ami though, which appears to be necessary to reach the position of Natufians relative to other West Eurasians. So the topology will probably change when we have populations from North Africa.

There's also something strange going on with Satsurblia, which I haven't noticed before. And interestingly, Villabruna vs Loschbour are quite different in relation to Ami, which supports the low levels of East Asian in West Eurasia discovered by Qiaomei Fu.

ryukendo kendow said...

Another thing, its pretty clear from this that S Asians like Munda either don't have any ANE ancestry, or very little, and most of the shared drift with EHG found in populations like India_South is brought by Steppe migrations rather than Iran_neolithic, which doesn't have that much shared drift with EHG really.

Chad Rohlfsen said...

There's no African in Natufians. The only reason it would appear in Admixture is too few samples for a component to form, so they get Anatolian plus African. Having a lot of Euros in a run will cause this.

I'll post a few stats here tomorrow night, but it's clear when looking at the following...

Mbuti.DG Iran N WHG ANE
Mbuti.DG ANE Natufian Iran N
Mbuti.DG Onge Natufian Iran N

Switching out Natufians for Levant N shows the same, but not quite as strong. I've said it and I'll say it again. Don't be surprised if ancient North Africans are 100% West Eurasian, with Natufians between them and the Levant.

capra internetensis said...

@chad

AFAIK North Africans are expected to be West Eurasian in this time period. And isolated by LGM Sahara, so could have no Sub-Saharan. But that implies that our Sub-Saharan references have no detectable North African ancestry, which is unexpected.

Do you have Dinka?

Shaikorth said...

Treemix finding SSA in Natufians is unsurprising considering Fst-ratios, and also if someone has managed to nMonte-fit them as part African. Dstats might show it too if we had better quality sequences of Natufians and more SNP's. Contemporaneous or earlier North African samples could be more SSA.

That 32% Iran_Neolithic edge branches off before East and West Eurasians and might be required to fit Iran_N into its position next to Iran_Chalcolithic as it's more basal.

Jaydeepsinh Rathod said...

"Also worth noting in the fourth graph is the heavy migration edge from near the base of the East Eurasian Dai/Han branch to the early Neolithic farmers from Iran. My bet is that this is a signal of admixture from a Central Asian forager population, perhaps representing a parallel clade to that of the East Asians?"

Dear David,

Can you add any of the ASI heavy groups like Paniya, Pulliyar, Kharia or Mala into the graph ? Perhaps this may resolve the question of where exactly the migration edge into Iran_N is coming from.

Thanks.

"Another thing, its pretty clear from this that S Asians like Munda either don't have any ANE ancestry, or very little, and most of the shared drift with EHG found in populations like India_South is brought by Steppe migrations rather than Iran_neolithic, which doesn't have that much shared drift with EHG really."

The Munda are an Austro-asiatic group and the Austro-asiatics have only entered into South Asia in the last 4,000 years. So the shared drift with EHG found in populations like India_South being from the steppe cannot be argued on the basis of the Munda.

Jaydeepsinh Rathod said...

"The same population X that exists in S Asia, breaking off from W and E Eurasians before they break off from each other, can straightforwardly explain Satsurblia and Iran_Neolithic."

Dear Ryukendo,

Can you elaborate on this ?

Grey said...

If Bantu expansion who was there before?

Was OoA different from modern SSA and a back migration from north China to north/east Africa combined with Bantu expansion erased them?

You need iodine for brain development, you get iodine from the sea and it declines from the equator (or so a paper i saw once said) so...

I assume from that the first people to have brains enough to OoA would be on one or other of the equatorial coasts

http://www.mapsofworld.com/africa/maps/africa-map.gif

any remote Kenyan tribes with eye folds?

ryukendo kendow said...

@ Immchr

Well, the question is are the Munda or the Indo-Aryans there first? Since I take the former position and you the latter, we're both going off tautologies here, no point using this as a ball to kick.

Quite some time ago, Sein presented his idea about how some 'third Eurasian' group, splitting off before either East or West Eurasians split off from each other, could be present in the mix that makes up the ASI. I was very skeptical at first due to phylogeographic reasons, but I've seen increasing evidence that this is in fact the case. Look at the following plot:

Plot 4

Paniya have only as much drift with ANE as levantines and East Asians, which makes very little sense if ASI is ENA, as Levantines = some WHG-EHG ancestry plus a whole lot of Basal Eurasian, but Paniya = some WHG-EHG ancestry + some ASI in the form of ENA (Which has more drift with ANE than Basal Eurasian) + some Basal Eurasian, so Paniya should outstrip both East Asians and Levantines in ANE affinity, but they do not.

Also, Paniya have much more drift with Ust-Ishim than expected, even more than Finnish, which I found difficult to reconcile with the amount of Basal Eurasian via Iran_N/CHG that they would have.

This problem is solved entirely if we assume the following tree:

Basal (Ust Ishim (X (West Eurasian HGs, East Asians and Onge))))
ASI = X+Onge

Which also makes sense of the mtDNA M being older in SE Asia than in India factoid that you yourself raised, esp if Onge-types overlaid the X-types.

Shaikorth said...

What proportions of those components do Paniya need to account for their relationship to ancients and also to East Asians and South Siberians:

Chimp Paniya Japanese Ket -0.0168 -6.6 134889

Grey said...

Probably not relevant but interesting what you get when you follow randomly generated google paths

Maasai call hunter gatherers dorobo

https://en.wikipedia.org/wiki/Dorobo_peoples

one of the references

"Hobley says (1905:39-40): "The author in 1891 visited Mount Kenia, and while encambed on the lower slopes encountered a few specimens of a bearded race of man who were said to live in the depths of the forests on that mountain.""

ryukendo kendow said...

@ Shaikorth

My guess is that the amount of Iran_N ancestry in Paniya is not as high as usually thought, definitely less than 40%, maybe even <25%. Otherwise they will be more Basal than Finnish or Yamnaya. Maybe about 10-30% of their ancestry comes from the X-types, and a significant amount from the Steppe as well.

Ariele Iacopo Maggi said...

Also Iran neolithic seems to behave like natufians, maybe is really this basal effect that increase the SSA.
--Iran_Neolithic--
"Satsurblia" 52.55
"Anatolia_Neolithic" 19.85
"Munda" 9
"Somali" 8.8
"AG3-MA1" 4.2
"Yoruba" 2.55
"Israel_Natufian" 1.6
"Papuan" 1.45
"Esan_Nigeria" 0
"LaBrana1" 0
"Eastern_HG" 0

distance=0.029425

Jaydeepsinh Rathod said...

Dear Ryukendo,

Thanks for the reply.

"Well, the question is are the Munda or the Indo-Aryans there first? Since I take the former position and you the latter, we're both going off tautologies here, no point using this as a ball to kick."

Perhaps you've misunderstood me. What I meant was, since Munda's presence in South Asia is only 4000 YBP there are not a representative of ancient South Asian. Munda's relatednesss to EHG =/= Ancient South Asian's relatedness to EHG.

---------------

"Paniya have only as much drift with ANE as levantines and East Asians, which makes very little sense if ASI is ENA, as Levantines = some WHG-EHG ancestry plus a whole lot of Basal Eurasian, but Paniya = some WHG-EHG ancestry + some ASI in the form of ENA (Which has more drift with ANE than Basal Eurasian) + some Basal Eurasian, so Paniya should outstrip both East Asians and Levantines in ANE affinity, but they do not.

Also, Paniya have much more drift with Ust-Ishim than expected, even more than Finnish, which I found difficult to reconcile with the amount of Basal Eurasian via Iran_N/CHG that they would have.

This problem is solved entirely if we assume the following tree:

Basal (Ust Ishim (X (West Eurasian HGs, East Asians and Onge))))
ASI = X+Onge

Which also makes sense of the mtDNA M being older in SE Asia than in India factoid that you yourself raised, esp if Onge-types overlaid the X-types."

Thanks for this. I have been thinking of something similar in the last few days. Let me share it with you.

According to me, there was no pure ASI or pure ANI population ever in South Asia. According me, most of the Eurasians descend from an ancestral meta-population that split up somewhere between South Asia & SE Asia.

Did you read this new paper on SEA genetics - http://www.nature.com/ejhg/journal/vaop/ncurrent/full/ejhg201660a.html

It appears, based on the above paper, especially fig. 1a, that Ancestral South Asians split up from the Ancestral SE Asians even before the split of Papuans, since even the Papuan related component is present in modern North & South Indians in trace amounts, more likely due to shared ancient ancestry. It appears that Ancestral South Asians split off from the Ancestral SE Asians before the SE Asians diversified into Papuans, East Asians (such as Chinese) etc.

Basal Eurasian could be that ancestry which is basal to the split of Ancestral South Asians & Ancestral SE Asians.

Perhaps, Ancestral South Asians are the ancestors of most of West Eurasians. Maybe after the split of Ancestral South Asians (ASA) & Ancestral SE Asians, the ASA came to Peninsular India and later on spread into Northern India/SC Asia. Maybe Ust-Ishim & Oase1 are members of that Ancestral South Asian group.

As you can see from this PCA - http://www.open-genomes.org/analysis/PCA/Eurogenes_Ice_Age_Eurasians_PC_plot_1-2-3.html,

Ust Ishim & Oase1 cluster with the ASI groups like Paniya & Pulliyar, while Kostenki, GoyetQ116 & the Vestonice samples cluster with North Indians and SC Asians. Its like they're on a ANI-ASI cline. What if the ANI rich North Indians are just a drifted subgroup of Ancestral South Asians that contributed to the West Eurasian ancestry in Early Europeans and elsewhere ?



Karl_K said...

" Basal (Ust Ishim (X (West Eurasian HGs, East Asians and Onge)))) ASI = X+Onge "

Doesn't ANE fit into this 'X' location? It branched off after Ust-Ishim, but before Kostenki14. Right?

Karl_K said...

@Grey

"Hobley says (1905:39-40): "The author in 1891 visited Mount Kenia, and while encambed on the lower slopes encountered a few specimens of a bearded race of man who were said to live in the depths of the forests on that mountain.""

Clearly pre-modern Hipsters.

German Dziebel said...

@Davidski

"Also worth noting in the fourth graph is the heavy migration edge from near the base of the East Eurasian Dai/Han branch to the early Neolithic farmers from Iran."

I'd like to reiterate my question from a couple of strings ago: where do Amerindians sit compared to East Asians (Dai-Han) when it comes to their relationship to Natufians and other West Asian aDNA samples? Stuttgart, I remember, was closer to Amerindians than it was to East Asians. I'd suspect the same will true of Natufians et al. but I'd like to see the actual data. Lazaridis et al. 2016 didn't include Amerindian samples.

The lack of Sub-Saharan African ancestry in Natufians is consistent with them having Y-DNA E lineages, which are a subset of a Eurasian CT clade that (back) migrated to Africa. Fst stats showing the greater proximity of Natufians to Eurasians than to SSAfricans capture this well. One way to interpret this is that the geographic distribution of Y-DNA E lineages captures the process whereby modern humans entered Africa (around 50-40,000 YBP) and admixed with descendants of African-specific archaics (carriers of A and B Y-DNA lineages) who, as distinct populations, never left Africa.

Karl_K said...

@Dziebel

"modern humans entered Africa (around 50-40,000 YBP) and admixed with descendants of African-specific archaics (carriers of A and B Y-DNA lineages) who, as distinct populations, never left Africa."


Very interesting. But how did the back-migration not bring Neanderthal ancestry into Africa?

Or was the back-migration only of a Basal Eurasian group?

If so, why isn't the back-migration admixture more obvious?

Shaikorth said...

Iran_Neolithic is less Neanderthal and at least as basal as Natufians, but does not show the African affinities Natufians do in TreeMix. The latter are mainly Eurasian with a part of their ancestry related to WHG, but these TreeMix results and fst-ratios in comparison to Iran_N suggest there is something else.

Also, if there was a significant West Eurasian backmigration to Africa the D-stat East_Asian French Mbuti/Dinka Chimp probably should not have Z>2 when run on high coverage genomes.

huijbregts said...

@ Ryu

We both want a clear visibility of the higher dimensions.
I fear that lot of noise is introduced by the higher dimensions.
You fear that usefull information is banned to high dimensions and gets dropped.
That remains to be seen. IMO the PCA suggests otherwise.
In your nice examples you zoom in on PC3 and remark that it is very small.
It is small because it is dwarfed by PC1.
I suggest that you drop a lot of the pops that are distant on PC1:
The Neaderthals: 'Denisovan','Neandertal_Altai'
SAA: 'Esan_Nigeria','Masai_Kinyawa','Yoruba','Somali'
EastAsian/Oceania:'Karitiana','Itelmen','Nganasan','Ulchi','Thai','Bougainville','Papuan','Kinh_Vietnam','Japanese','Korean','Dai','Ami','Han'

In the next table I have collected the percentage of variance by dimension.
In the trimmed model you get a much higher percentage of the variance in PC3

"" "full" "trimmed"
"1" 86.3854898282583 54.8746538824027
"2" 10.0076974051496 27.0442564443205
"3" 1.26814694825649 9.96403733836785
"4" 1.02535733933314 3.97391011934758
"5" 0.500970551920407 1.50453341759423
"6" 0.2752102361601 1.07163895764541
"7" 0.151934430810854 0.582523785598367
"8" 0.103940621153565 0.278516521587389

Ariele Iacopo Maggi said...

Bell_Beaker_Germany
"Remedello_BA" 49.1
"Yamnaya_Samara" 43.2
"Loschbour" 7.2
"Han" 0.5
"Iberia_Chalcolithic" 0
"Yamnaya_Kalmykia" 0
"Armenia_Chalcolithic" 0
"Hungary_CA" 0
"Hungary_EN" 0
"Satsurblia" 0
"Ami" 0
"Iberia_EN" 0
"Armenia_MLBA" 0
"Cypriot" 0

David. I think something is wrong with Remedello (or I did something wrong).
Remedello_BA
"Anatolia_Neolithic" 70.2
"Eastern_HG" 23.65
"LaBrana1" 6.15
"Satsurblia" 0
"Iran_Neolithic" 0
"AG3-MA1" 0
"Esan_Nigeria" 0
"Somali" 0
"Yoruba" 0
"Israel_Natufian" 0
"Munda" 0
distance=0.058306

ryukendo kendow said...

@ Ariele

I don't think you did anything wrong, at least not for the Bell Beaker fit. I've not finished the fitting process yet, but my preliminary fits are also indicating a big role for Remedello in BB, though total Iberia contribution is still ~20%. Possibly because the range of sources are wider than yours.

Ariele Iacopo Maggi said...

@ ryukendo kendow
My Remedello shows Yamna/EHG... I thought that Remedello was supposed to be a standard sardinian-like EEF.
Remedello_BA
"LBK_EN" 74.15
"Motala_HG" 9.3
"Yamnaya_Samara" 9.2
"Hungary_HG" 7.35
"Anatolia_Neolithic" 0
"Satsurblia" 0
"Iran_Neolithic" 0
"AG3-MA1" 0
"Esan_Nigeria" 0
"Somali" 0
"LaBrana1" 0
"Yoruba" 0
"Israel_Natufian" 0
"Eastern_HG" 0
"Munda" 0
"Yamnaya_Kalmykia" 0
"Villabruna" 0
"Hungary_EN" 0
distance=0.040515

Rob said...

But why did Ariele's run get 0% Iberia_Chalc
(Good work btw ) ?

ryukendo kendow said...

Oops Ariele, because I was doing a large number of runs, I misread Iberia Chl 7% contribution as 17% contribution, which when added with a few other small percentages gives me 20%. The actual percentage is approx 10% from Iberia.

Looks like whatever Iberia_Chl was capturing previously is now being captured by Remedello, good suggestion, Rob and Richard Rocca.

[1] "distance%=0.1594 / distance=0.001594"


Bell_Beaker_Germany
"Yamnaya_Kalmykia" 15.6
"Yamnaya_Samara" 13.7
"Remedello_BA" 13.4
"Anatolia_Chalcolithic" 12.15
"Hungary_BA" 11.5
"Iberia_Chalcolithic" 7.4
"Hungary_HG" 5.8
"Iberia_EN" 4.05
"Baalberge_MN" 3.55
"Motala_HG" 3
"Samara_Eneolithic" 3
"Poltavka_outlier" 1.3
"Loschbour" 1.15
"Satsurblia" 0.95
"Unetice_EBA" 0.7
"Iberia_MN" 0.55
"Eastern_HG" 0.5
"Moroccan" 0.45
"Itelmen" 0.25
"Corded_Ware_Germany" 0.2
"Ulchi" 0.2
"BedouinB" 0.15
"Srubnaya" 0.1
"Villabruna" 0.1
"Andronovo" 0.05
"Armenia_EBA" 0.05
"Armenia_MLBA" 0.05
"Hungary_EN" 0.05
"Orcadian" 0.05

As you can see, there are still many populations to drop. I will post the final runs later, but currently I'm working on something else for Huij, so others please feel free to cut in.

Ariele, I will check out Remedello later, but in the meantime, I recommend you drop the EHG-containing populations, and see if the signal reappears in the form of another population. I expect it would reappear, such large percentages usually do, in which case the signal is probably real. Either way, this lends weight to another theory--Jean Manco's theory of travel through Italy (Rob correct me if I'm wrong).

Ariele Iacopo Maggi said...

@ ryukendo kendow
"my preliminary fits are also indicating a big role for Remedello in BB"
Maybe is the contrary. Remedello traditionally is considered the first warrior like culture IE/Kurgan culture in Italy. There is a lot of literature on that particular topic.
BB showed up in Remedello

Remedello_BA
"Iberia_EN" 51.7
"Bell_Beaker_Germany" 48.3
"Anatolia_Neolithic" 0
"Satsurblia" 0
"Iran_Neolithic" 0
"AG3-MA1" 0
"Esan_Nigeria" 0
"Somali" 0
"LaBrana1" 0
"Yoruba" 0
"Israel_Natufian" 0
"Eastern_HG" 0
"Munda" 0
"Yamnaya_Kalmykia" 0
"Yamnaya_Samara" 0
"Villabruna" 0
"Motala_HG" 0
"LBK_EN" 0
"Hungary_EN" 0
"Hungary_HG" 0
"Iberia_Chalcolithic" 0
"Anatolia_Chalcolithic" 0
"Hungary_BA" 0
"Hungary_CA" 0
"Iberia_MN" 0
d=0.025

Rob said...

Yes IIRC Jean Manco suggested steppe movement via Balkans & / or Italy then to Iberia and the BB network, but I'll leave that to her to clarify.

Didn't Kurd find something 'odd' (east shifted) with Oetzi like a year ago ?

Intriguing only, I was just pondering an admittedly thin hypothesis : What if Balkan foragers already had significant EHG -type ancestry ?

Karl_K said...

What is the date for this Remedello genome? Don't the later Remedello graves contain Bell Beakers?

Rob said...

Ariele
Remedello (33--2500 BC) would be mostly anterior to BB (27/500-2200 BC)

ryukendo kendow said...

While waiting for the runs to finish, Immchr, I can run with your theory: you say that Basal Eurasian split from the proto-ANI-ASI population outside India, then the proto-ANI-proto-ASI population traveled to India, and then the ANI population spread out to Europe while the ASI population spread to East Asia. But from the Metspalu paper, which you seem to know well, you know that the ANI population is most like Georgians or CEU (sample of whites from Utah). This means that ANI contains Basal Eurasian, as this is an important component of ancestry in Europeans and West Asians. So this demands that the ancestral ANI population must have traveled outside India, admixed with Basal Eurasians in West Eurasia, which is after all where the Basal Eurasian component originates as the aDNA from the Lazaridis paper tells us, and traveled back to India in later times. In fact, post-Neolithic times, since thats when Basal and non-Basal populations begin mixing in earnest. Are you happy with this theory?

What is undeniable is that West Eurasians as a clade had already existed in Europe and West Siberia by the time of Kostenki, 45000 YA, without Basal Eurasian, while ANI, a mix of Basal Eurasian with post-Mesolithic Hunter-gatherers in Europe and Siberia, had to emerge much later, in the Neolithic and Chalcolithic, which is when populations from south of the Mediterranean and Caucasus began to mix with populations North of it. The current genetic composition of Indians requires an input from populations like Neolithic and Chalcolithic Caucasians and West Siberians, with the East European/West Siberian input biased towards high-caste populations like Tiwari and GujaratiA, and the Middle-Eastern like input biased to low castes populations like Lodhi and Chenchu. This really doesn't leave much room for interpretation, other than Neolithic and post-Neolithic, i.e. recent, population movements that had differential impact on populations of different castes.

K U. said...

@Davidski

Re: SSA in Natufians: What do you think about my theory here: http://www.anthrogenica.com/showthread.php?7489-Lazaridis-et-al-The-genetic-structure-of-the-world-s-first-farmers-(pre-print)&p=166159&viewfull=1#post166159

the tl;dr: Natufian SSA affinity can be explained by unaccounted for Northwest African (Berber-like) admixture in Yoruba/West Africans, and NW Africans share common ancestry w/ Natufians.

Karl_K said...

@Shaikorth

"Iran_Neolithic is less Neanderthal and at least as basal as Natufians, but does not show the African affinities Natufians do in TreeMix. The latter are mainly Eurasian with a part of their ancestry related to WHG, but these TreeMix results and fst-ratios in comparison to Iran_N suggest there is something else.

Also, if there was a significant West Eurasian backmigration to Africa the D-stat East_Asian French Mbuti/Dinka Chimp probably should not have Z>2 when run on high coverage genomes."

Couldn't this all just be a result of which Eurasian population had the most back-migration impact in Africa?

If Natufians, in specific, admixed into Sub-Saharan Africans in the distant past, and admixed the least into the modern Eurasian populations, then the programs would put any alleles that they did not share with other Eurasians (but do share with Africans, due to back migration) into the African clusters.

So. The African component in Natufians could just as likely be the signal for Natufian back migration into Africa.

postneo said...

@jrathod
Just a note: It would be interesting to know admix and linkage disequilibrium of denisovans in SE asians, ASI, ANI, Iran_N etc.
Iran_N is on the same side of the guls as S asians, theres a bigger barrier btw SE asians and S asians.

ryukendo kendow said...

@ Karl K

Karl, in this image:
http://imgur.com/bbsgoEl

If you draw a line from Dai to Abkhasian or Georgian, which was our previous mental model of Indians as ENA+CHG, you will end up with a cline of populations that Paniya does not fall on. Specifically, Paniya is shifted to the left, i.e. has less ANE-EHG ancestry than expected, for her level of Ust-Ishim/ENA similarity and CHG similarity.

Now we know that the actual population is Iran_N, and if you draw a line from East Asians to Iran_N:
http://i.imgur.com/UIINvST.png

South Asians now fall above the line. But then this raises the problem of why Iran_N has so little drift with EHG-ANE. One can go with the author's theory of extreme levels of Basal in Iran_N, which would in fact explain everything, but Iran_N and especially Satsurblia are actually outliers in similarity to Ust-Ishim vs similarity to EHG, indicating that, for all their low similarity to EHG, they are not all that Basal, some of this shift away from ANE-EHG is still unexplained.

Furthermore, since they are not outliers in similarity to Ami in the other plot, but continue being outliers in similarity to Ust, its not ENA that is depressing the similarity; its something Crown Eurasian that is neither ENA or W Eurasian.

Since the tree is:

Africans (Basal (Ust Ishim (X ((ANE-WHG) , (Dai-Onge)) )))

ANE doesn't fit the position.

This is explained if we posit that there was some population in between S Asia and W Eurasia, possibly in SC Asia, that broke off before E and W Eurasians broke off from each other, and occupies position X, which influenced either of or all of Iran_N and ASI and Satsurblia.

Karl_K said...

@RK

"This is explained if we posit that there was some population in between S Asia and W Eurasia, possibly in SC Asia, that broke off before E and W Eurasians broke off from each other, and occupies position X, which influenced either of or all of Iran_N and ASI and Satsurblia."

Sounds reasonable. It is actually surprising that there aren't many more hints of these ghost populations.

German Dziebel said...

@Karl

"Very interesting. But how did the back-migration not bring Neanderthal ancestry into Africa?"

It did. It was found across SSAfrica but the dates that were assigned to its entry were too late. It's an old trace that just got diluted by in-situ archaic African ancestry.

The fact that Basal Eurasian ancestry is adversely correlated with Neandertal ancestry (the lower frequency of Neandertal alleles in Europe is correlated with higher frequencies of Basal Eurasian) disproves the argument made by geneticists that non-Africans picked up their Neandertal lineages in West Asia on their way out of Africa. (Not that this incoherent argument needed to be disproved by precious aDNA data.) The opposite scenario whereby Eurasians entered SSAfricans and admixed with in-situ hominins explains everything very well.

Basal Eurasian (or populations that are high in the BE component) should be expected to be more closely related to Amerindians than it is to East Asians. And by its structure (small, drifting component with an equal relationship to both West Eurasians and East Asians) BE is most reminiscent of Amerindians.

mickeydodds1 said...

Karl K,

That would fit with the widespread incidence of y DNA haplotype E in sub Saharan Africa.
Undoubtedly E was originally a Eurasian Claude.

capra internetensis said...

There are lots of ways Natufians could be related to Africans. The question is why they have this affinity in TreeMix and FST ratios but it's completely absent in D stats.

It doesn't matter if Natufians mixed into Sub-Saharan Africans rather than the other way around, it should still show up in the D stats.

German Dziebel said...

@ capra

"There are lots of ways Natufians could be related to Africans. The question is why they have this affinity in TreeMix and FST ratios but it's completely absent in D stats."

Fst shows Natufians as being closer to Han (0.176) and WHG (0.153) than they are to Yoruba (0.186). Any tree program would build a tree out of any pool of geographic variation. More variation means more upstream for the program regardless whether the nodes are true branching points within an expanding population or admixture moments between a younger superstratum and an older substratum. Natufians are geographically between SSAfricans and Eurasians, so they would look like having an "African" signal. D-stats and Fst must be more discerning.

Shaikorth said...

Natufian migration edges are arrayed in such a way that they don't imply stable Natufian admixture in Africa. For instance in tree3, the edges do not bring Yoruba or the pygmies closer to Eurasia.

Also if both Dinka and Mbuti are shifted towards East Eurasians rather than towards French using high coverage sequences, that is telling. Natufians are closer to West Eurasians than to East Eurasians (admixturegraph suggested their West Eurasian part is close to WHG) so significant admixture from them to SSA isn't suggested by these stats.

Karl_K said...

@capra

It has to be due to the populations chosen as African.

Karl_K said...

@Shaikorth

"Natufians are closer to West Eurasians than to East Eurasians (admixturegraph suggested their West Eurasian part is close to WHG) so significant admixture from them to SSA isn't suggested by these stats."

But if the admixture into Africa was only from the Basal Eurasian ancestral side of Natufians, than the WHG complication is a later unnecessary complication.

Shaikorth said...

German,

"Fst shows Natufians as being closer to Han (0.176) and WHG (0.153) than they are to Yoruba (0.186). "

Fst also shows that ancient and modern Eurasian populations on their supplementary table, including Near Easterners like Palestinians and Jordanians, are closer to Papuans than to Mbuti, with a few exceptions - the Bedouins, Natufians and Levantine_N.

Karl, if it's basal to Africa that is causing the edges how come Iran_N is not showing this affinity? It should have even purer basal as it's less Neanderthal than the Natufians.

German Dziebel said...

@Shaikorth

"Fst also shows that ancient and modern Eurasian populations on their supplementary table, including Near Easterners like Palestinians and Jordanians, are closer to Papuans than to Mbuti,"

It's a matter-of-course if Mbuti is more archaic-admixed than are Papuans.

"if it's basal to Africa that is causing the edges how come Iran_N is not showing this affinity? It should have even purer basal as it's less Neanderthal than the Natufians."

We need stats involving Amerindians to get a full picture.

Ariele Iacopo Maggi said...

@ Rob
"Remedello (33--2500 BC) would be mostly anterior to BB (27/500-2200 BC)"

This Remedello BA could be from a very late period, from around 2000 BC or even younger, if this sample has direct BB admixture both results (BB in Remdello and Remedello in BB) are not surprising at all (if we could have the d-stats with all 3 Remedello samples it would be even more intersting and we could check if BB has early Remedello influence or not).
Btw, as far as I know we completely ignored the fact that this Remdello BA had steppe genetic influence, that is kind of a big deal anyway. If someone has the files of this sample it would be great to put it in a calculator or something.

Grey said...

Karl_K

"Clearly pre-modern Hipsters."

gentrification smh

Karl_K said...

@Shaikorth

"Karl, if it's basal to Africa that is causing the edges how come Iran_N is not showing this affinity? It should have even purer basal as it's less Neanderthal than the Natufians."

We don't have a clear enough picture of the Basal Eurasian population. It must have split well before seperate admixture with the other Eurasian populations.

The branch of Basal leading toward Natufians could have admixed back into SSA. This would explain the affinity.

We just can't say that much without more data. There are several possibilities.


Shaikorth said...

German, what if there is no major archaic admixture in Mbuti though? Also if there is, do you suggest Bedouins and Natufians but not other Near Easterners have the same stuff too?

We can also replace Mbuti and Papuan with Ami and Yoruba, same pattern emerges with Levantine_N, Natufians and populations with SSA being closer to Yoruba and everyone else including Iran_N and Anatolia_N closer to Ami.


These migration edges do not increase the affinity of all SSA's to Natufians which is what we would expect if there was a widespread basal mixture. Instead it varies which makes Natufian affinity to SSA's stable but not vice versa.

epoch2013 said...

Shared ancestral because Basal branched off before Crown?

Karl_K said...

@epoch

yes.

Samuel Andrews said...

@epoch2013,
"Shared ancestral because Basal branched off before Crown?"

Natufians didn't have an exceptionally high amount of Basal. Neolithic Iran had more but aren't showing any signals of African ancestry.

@Chad,

The stat D(Chimp, Natufian)(Mbuti, East Asian) is signifcantlly less positive than the stat D(Chimp, any other Middle Easterner)(Mbuti, East Asian).

Natufian's are as distant from East Asians as heavily African admixed Egyptians. What other explanation is there but African admixture in Natufians?

Alberto said...

Regarding the D-stats with Natufians comparing to Africans, I remember one case when some stats were saying that Lithuanians were as close to Germany_MN as they were to Belarusians and Norwegians. That seemed very weird. Then Matt made a plot with those stats in the Y axis and another ones in the X axis. And it became obvious that while in the Y axis Lithuanians, Norwegians and Belarusians were at the same height as Germnay_MN, on the X axis the former 3 still were equal and therefor clustered together, while Germany_MN was clearly away from them. So I think that these direct D-stats just can't tell us the whole story, only a small part.

Now we have these sheets with many D-stats comparing each population to many others, and with that you can calculate the distance between 2 populations in all the dimensions provided by all the different columns/populations. This should give a much more realistic result.

For example, this method showed that Ust-Ishim is closer to Africans that all other ancient genomes (latest ones not included), which sounds strange at first sight but makes a lot of sense when you think about it. I think it was even closer than all modern Eurasians except maybe those with largest amount of SSA admixture.

In these latest sheet there is no Ust-Ishim, but a few distances to Yoruba for other populations:

Ami = 1.371847
Loschbour = 1.368161
English_Cornwall = 1.301642
Italian_Tuscan = 1.278244
Vestonice16 = 1.273387
Levant_Neolithic = 1.19589
Iran_Neolithic = 1.18255
Palestinian = 1.169303
BedouinB = 1.16451
Israel_Natufian = 1.148884
Moroccan = 0.972796

So not all Eurasians are equidistant to Africans (that would be an amazing coincidence, if not outright impossible). But it doesn't mean that Vestonice has more SSA admixture than Loschbour, or that English have more than Ami. Lower distance should not mean admixture.

But I think that what the direct D-stats suggest is that there is no SSA admixture in Natufians. I guess they would show the same with Ust-Ishim. But maybe not with Palestinian who does have some SSA admixture, even if they are further away from Yoruba than Natufians are? Something like:

D(Chimp, Yoruba; Palestinian, Israel_Natufian)

Might shed some light about admixture vs. distance.

huijbregts said...

@ Ryu
If you trim the sheet as I suggested, the first four Principal Components explain 95.9% of the variance.
They look well on a PCA.
From the fifth component on the components are dominated by one row and column.
PC5: Satsurblia/Kotias
PC6: Kostenki14/Vestonice,GoyetQ
PC7: Mansi
PC8: BedouinB/BedouinB2
PC9: MotalaHG/MotalaHG2
PC10:minusAG3-MA1/Yamnaya_Samara2
I have no idea how nMonte will react on these singular columns; I have little confidence.
If we cannot improve on this D-stat sheet, I think that it is wise to limit ourselves to 4 components.

Alberto said...

@RK

Did you ever look at the possible reason why MA1 has a clearly different behaviour in those direct D-stats (or even f3 stats with outgroup for shared drift) than it does in IBS or in this D-stats based sheets?

If you plotted MA1 f3 shared drift vs. MA IBS there wouldn't be a good correlation, with a line clearly separating Europeans (on the f3 side) and S-C Asians (on the IBS side). And the curious thing is that D-stats sheets for nMonte seem to agree with IBS quite more than with the direct D-stats. So I think that the strange position of Paniya in that plot might be misleading (as in general the relative position of Europeans vs S-C Asians).

Distance from MA1-AG3:

Sardinian = 0.167171
Italian_Tuscan = 0.149482
Spanish = 0.14607
India_South = 0.141406
Bulgarian = 0.139648
English_Cornwall = 0.130954
GujaratiD = 0.122896
Estonian = 0.115291
Srubnaya = 0.112519
GujaratiA = 0.109248
Pathan = 0.109188
Kalash = 0.107852
Andronovo = 0.104631
Tajik_Shugnan = 0.100034
Tajik_Ishkashim = 0.095602

Here is MA1 IBS with modern populations:

https://drive.google.com/file/d/0B2ZfdVZaNXDxcHR2dGs1X2w1M0k/view?usp=sharing

Shaikorth said...

@Alberto

How many SNP's in common with MA-1 was used in the IBS sheet?

Matt said...

@ Alberto, I guess "Distance from MA1-AG3" is from when you run nMonte with a single calc pop (e.g. MA1-AG3) and then pull out distance?

That's kind of a measure of how like the MA1-AG3 row a population row is its relatedness to the columns. In theory because it is descended from that population.

But also a population could score similarly to MA1-AG3 in how it relates to the columns while at the same time having a relatively low relatedness to MA1-AG3 itself. That's particularly true if the columns are not really good at describing the drift history of one or both of the row populations (e.g. Vestonice16 has no particularly relevant column it shares a lot of drift with).

Alberto said...

@Shaikorth

I don't know. Maybe Davidski could tell, since he's the one who ran that test (but it was long ago). Though they are only modern populations, so I guess the overlap should be as good as it can be.

@Matt

Yes, that's the method used. And yes, it's not a specific "shared drift" between the populations, it's more like how related the populations are to each other in a broader way (and I think more accurate way from a population genetics point of view). It overcomes the limitations of using a single D-stat, which we know that cannot tell us all the information about the ancestry of a population.

But also a population could score similarly to MA1-AG3 in how it relates to the columns while at the same time having a relatively low relatedness to MA1-AG3 itself.

In how it relates to all the columns? I don't know if you're thinking about some kind of big coincidence? Like two populations plot in the same place in all the dimensions of a PCA but are otherwise unrelated?

Olympus Mons said...

@Ariele,
I see you are already asking for the same as me.
Instead of running RISE486 (most likely) from 2100 bc, lets us see how the 3 samples individually do because they span a millennia from before bell beaker and very much "german bell beaker" of course.

Thanks.

Matt said...

All I would suggest is the following:

- Take the D-stats sheet, and try taking out columns relevant to European HG (Kostenki, WHG, SHG, EHG)

- Using the modified sheet re-run closeness to Yoruba for Euro HG rows

See how it changes. If the European HGs look like they're now closer to Yoruba (less distance), does it seem like they've actually closer to Africa, or is this just an artefact that you have no relevant columns any more?

Now how does this apply to estimating the closeness of an Ust Ishim row or a Natufian row to Yoruba?

Alberto said...

@Matt

Removing the columns that are important would obviously reduce the accuracy of the result. We have as many relevant columns as possible, and it should be good enough (though of course, I wouldn't complain if we could have more and especially relevant ones). But I think this is beyond the main point.

We've discussed this issue before, and I'd be glad to discuss it again (though probably it's long to discuss it in detail here). But to show with an example what I mean:

There was a datasheet with columns such as Jordanian2, Palestinian2,... And when looking at this columns, for example for Jordanian2 (a randomly chosen group or Jordanians) the highest shared drift was with populations like Anatolia_Neolithic and LBK_EN. Jordanians themselves (the other individuals split from the same group) was in the position 104, just below Karelia_HG and Brahui. Palestinian was in position 99, just below Kalash.

So with that sheet, here are the distances from Jordanian2 for some of those populations using this method:

Karelia_HG = 0.22517
Kalash = 0.118732
LBK_EN = 0.115431
Anatolia_Neolithic = 0.112082
Palestinian = 0.009915

So while the columns themselves are telling us more accurately the shared drift (from a technical point of view), it is this second method the one that's telling us what we want to know, the shared drift with the source populations as a whole. Which is what matters to us to understand the historical ancestry.

And this has nothing to do with having more or less columns (though again, the more relevant columns, the more accurate the result). It's the concept itself that matters.

Rob said...

Ariele

Dave used all 3 Remedello samples aggregatively. They date from 3200 - 2600 BC, approx. (although not individual carbon dated). So they're earlier than out BB sample set (2600 - 2400 BC). So BB -> Remedello is ahistoric, and I suspect you're just detecting shared ancestry, or some common input into both.

And contrary to what you earlier wrote, the idea that Remedello is a "kurgan culture' is outdated & simplistic, but rather, it just shows signs of the Copper Age revolution with widespread impact.
I think we need more formal testing & better quality samples to make any definite conclusions.

Ariele Iacopo Maggi said...

Rob
"I think we need more formal testing & better quality samples to make any definite conclusions."
Understood...

Davidski said...

Please don't run the Remedello samples. They don't have enough markers, and especially high quality markers, for this type of analysis.

German Dziebel said...

@Shaikorth

"what if there is no major archaic admixture in Mbuti though? Also if there is, do you suggest Bedouins and Natufians but not other Near Easterners have the same stuff too?"

Without aDNA from SSA we won't know for sure but considering that we don't find any African uniparental lineages (Y-DNA A, B; mtDNA L1, L2, etc.) along the putative migration routes out of Africa (either marginally in modern populations or in ancient Eurasian DNA) - while we have tons of them coming to America as part of well-attested slave migrations post-1492 suggests that those very divergent lineages got absorbed into modern human populations in Africa from African-specific archaics. They never left Africa until slave trade times.

"Instead it varies which makes Natufian affinity to SSA's stable but not vice versa."

The mosaic pattern of affinities between ancient West Asian samples and SSA that you are bringing up can be explained as the effects of varying rates of drift affecting different Pleistocene populations differently.

Balaji said...

Davidski noted a “ heavy migration edge from near the base of the East Eurasian Dai/Han branch to the early Neolithic farmers from Iran”. This can be explained by a migration from India to Iran by agriculturalists who introduced agriculture to the Iranians. I think Jaydeepsinh Rathod has suggested something like this. There has long been a connection between India and Iran which makes sense since they are neighbors though separated by formidable deserts and mountains. MtDNA U7 is fairly common in both. The following paper studied this.

http://bmcgenet.biomedcentral.com/articles/10.1186/1471-2156-5-26

They found, “Moreover, the South- and West Asian-specific sub-branches of haplogroup U7 predate the last glacial maximum. Therefore, deep autochthonous history of these haplogroups in the region remains to be the most parsimonious explanation.”

The recent Lazaridis preprint found that ANI could by modeled as a combination of Iran_Neolithic and Steppe. Many have leaped to concluding that this has proved that agriculture was brought to India by Iran_Neolithic and that later invaders from the Steppe brought in Indo-European languages. Not so. Actually the results are quite consistent with OIT. There may well have been structure within the Subcontinent with a more CHG-like population and a more ANE-like population. The more CHG-like population migrated to Iran and gave rise to Iran_Neolithic and the more ANE-like moved to the Steppe spreading Indo-European languages. The analysis of the Rakhigarhi samples will probably soon resolve this issue.

Davidski said...

@Balaji

Actually the results are quite consistent with OIT. There may well have been structure within the Subcontinent with a more CHG-like population and a more ANE-like population. The more CHG-like population migrated to Iran and gave rise to Iran_Neolithic and the more ANE-like moved to the Steppe spreading Indo-European languages.

Ridiculous.

huijbregts said...

@ Alberto

"Removing the columns that are important would obviously reduce the accuracy of the result."

It depends on what you consider important. If you mean statistically important, I completely agree.
But I do not agree if you mean something like genealogically interesting.
In advanced statistical workflows a stage of dimensionality reduction is quite common. A quote from the wikipedia lemma on this subject:
"In some cases, data analysis such as regression or classification can be done in the reduced space more accurately than in the original space".

Chad Rohlfsen said...

@ Sam Andrews

"Natufians didn't have an exceptionally high amount of Basal. Neolithic Iran had more but aren't showing any signals of African ancestry.

The stat D(Chimp, Natufian)(Mbuti, East Asian) is signifcantlly less positive than the stat D(Chimp, any other Middle Easterner)(Mbuti, East Asian).

Natufian's are as distant from East Asians as heavily African admixed Egyptians. What other explanation is there but African admixture in Natufians?"

Wow, you're very wrong all over here. Iran is not more basal than Natufians. Read Table S4.9. Iran Hotu could even be significantly less basal than Natufians with a higher SNP count. Also, if you want to compare populations it should be direct comparisons. Not this nMonte Dstat stuff, with differing SNP coverage and a host of issues that pop up using Chimp. Which is probably why you pulled out some bogus Egyptian crap. Egyptians and Africans form a clade to the exclusion of Natufians and Asians, Natufians and WHG, and Natufians and UP Europeans. So, there's that.

result: Mbuti Onge Natufian Egyptian -0.0200 -5.647 13257 13799 277983
result: Mbuti Han Natufian Egyptian -0.0153 -4.586 13406 13821 277985
result: Mbuti Loschbour Natufian Egyptian -0.0377 -8.227 13304 14347 275085
result: Mbuti Villabruna Natufian Egyptian -0.0508 -10.192 12345 13666 258219
result: Mbuti Bichon Natufian Egyptian -0.0406 -8.225 13353 14484 277703
result: Mbuti Kostenki14 Natufian Egyptian -0.0399 -8.368 12413 13446 262200
result: Mbuti GoyetQ116-1 Natufian Egyptian -0.0475 -9.036 10558 11611 223349

Shaikorth said...

To get various HG's and Ust-Ishim to fit the model Iran_N has to be more basal than Natufians (by almost 10%) as per figure S4.11.

Iran_N is also less Neanderthal than Natufians, another indicator of it being the most basal sample. Natufians show extra African affinities compared to it in Treemix and relative fst-distances (while Iran_N behaves like Anatolia_N, but that seems to be related to Africa. It is not as SSA as modern North Africans though.

Rob said...

Yep;
Gulf oasis

Grey said...

If for the sake of argument OoA started from the equatorial east coast somewhere around Kenya/Tanzania and was later disappeared by a mixture of back migration and Bantu expansion the tribes that first displaced them might contain traces of it:

Kikuyo, Luhya, Luo, Kalenjin, Kamba, Maasai etc

https://en.wikipedia.org/wiki/Demographics_of_Kenya

or pushed south into Tanzania like the Sandawe

https://en.wikipedia.org/wiki/Sandawe_people

Chad Rohlfsen said...

Shaikorth,

That's not how Iosif feels. He told me they're about even. If Iran were so much more basal and less Neandertal, they would be much further from Ust_Ishim, but they're looking minimally closer. They're also significantly closer to ENA, which can make basal look higher in f4ratio.

Jaydeepsinh Rathod said...

Dear Ryukendo,

"you say that Basal Eurasian split from the proto-ANI-ASI population outside India, then the proto-ANI-proto-ASI population traveled to India, and then the ANI population spread out to Europe while the ASI population spread to East Asia."

No I did not say so. Perhaps I wrote in haste and did not express myself clearly. What I mean is the following :-

Once Out of Africa, there must have been a meta-population somewhere between South Asia & SE Asia, from which the native Australians & Oceanians may have already split up. This meta-population must have split up into Ancestral South Asian & Ancestral SE Asians. The split may have happened in SE Asia, since the more basal clades of mtDNA M & N as well as the y-dna paragroup F (Vietnam) & the presence of subclades of paragroup K are found there. This split between Ancestral South Asians & Ancestral SE Asians must have happened quite early and even before the split of Papuans & East Asians from the Ancestral SE Asians. The Ancestral SE Asians were likely the ancestors of all East Eurasians including Papuans.

The Ancestral South Asians after the split from Ancestral SE Asians, must have entered South Asia from the South East. The Peninsular Indians have more ancestry from the ASI component since they are geographically closer to the SE Asians and the greater percentage of ASI component in them, IMHO, reflects greater degree of relatedness to SE Asians.

---------

As far as basal Eurasian is concerned, I do not have much understanding of it but since Iranian Mesolithic has the highest basal Eurasian among ancient Near Easterners and since it is also the easternmost, basal Eurasian seems to come from the East. Perhaps basal Eurasian is that group that split off before the splitting up of Ancestral South Asians & Ancestral SE Asians. Its origins may lie somewhere in SE Asia or in South Asia.

---------

After the entry of Ancestral South Asians in Peninsular India, some of them must have started migrating North into North India & SC Asia, with some continuing the migration even further North. Oase1 & Ust Ishim are perhaps two of those ancient migrants into the North that belonged to the Ancestral South Asian group. They both belonged to y-dna K2a* and the closest modern human to them is a Telugu individual who is K2a1*. On David's own PCA of Ice Age Europeans, Ust Ishim & Oase1 cluster with ASI heavy groups like Paniya & Pulliyar.

http://www.open-genomes.org/analysis/PCA/Eurogenes_Ice_Age_Eurasians_PC_plot_1-2-3.html

You have yourself accepted that Paniya shares quite a lot of drift with Ust Ishim.

----------
(continued...)

Jaydeepsinh Rathod said...

(...continued)


With continuous migration of Ancestral South Asians from Peninsular India into North India & SC Asia, a sizeable population may have formed in North India/SC Asia which between 45000 YBP to 36000 YBP, must have generated some drift away from the Peninsular Indians. This drifted North Indian/SC Asian group must have then already become ANI heavy as it is today.

Kostenki14, GoyetQ116 & the Vestonice cluster seem to be migrants from this North Indian/SC Asian group which had developed some drift in relation to Peninsular Indians. Kostenki14 clusters with North Indians & Central Asians in the PCA (both in Seguin Orlando et al paper as well as in the 3d PCA of David I mentioned above). Kostenki14 was y-dna C1b & mtDNA U2, both the clades being found in modern South Asians. Moreover, in the 3D PCA of Ice Age Europeans, generated by OpenGenomes, GoyetQ116 & the Vestonice group also clusters with North Indians & SC Asians. GoyetQ116 was mtDNA M and so was Ostuni1 of the Vestonice group, and both of them cluster tightly with North Indians & SC Asians. So Ust Ishim & Oase1 cluster with South Indians like Paniya & Pulliyar while Kostenki12, GoyetQ116 & the Vestonice group clusters with North Indians & SC Asians. Its as if they lied on the ANI-ASI cline. Is this just co-incidence ? Or does this reflect very ancient population structure in South Asians ?

You have yourself argued that Kostenki14 also shows a cline of relatedness from North India to South India, with Kostenki14 being closer to North Indians than to South Indians.

According to me, Ancestral South Asians split up from Ancestral SE Asians first and then entered Peninsular India. This was before the split of South Asians & Western Eurasians. To my mind, Ancestral South Asians are also ancestors of West Eurasians, including the early palaeolithic Europeans. After its entry into Peninsular India, some members of Ancestral South Asians must have started moving Northward and formed a sizeable population in North India/SC Asia. This group over time (between 45000 YBP & 36000 YBP) drifted away from Peninsular Indians. From this drifted North Indian/SC Asian group, some members migrated North and into Europe and these were the Early Europeans like Kostenki14, GoyetQ116 & the Vestonice group. It would explain the Kostenki14 cline that we observe in modern South Asians. Therefore the North Indians/SC Asians show greater relatedness to Western Eurasians in comparison to Peninsular Indians because Western Eurasians are a group of North Indians/SC Asians that migrated North & perhaps West.

---------------

Going forward, there is MA-1. MA-1 is an extremely drifted sample and it is an outgroup to all pre-Neolithic Early Europeans like Kostenki14. However it forms a clade with Ust Ishim & Oase1 and is not an outgroup to them and pre-Neolithic Early Europeans. This, in combination with the fact that MA-1 was mtDNA U & y-dna R*, suggests that even MA-1's origins might lie further South towards South Asia. The largest ancestry component of MA-1 is ofcourse the South Asian one. South Asia also harbours both R1 & R2 clades as well as Q1b.

The ANE component, that is so high in many South Asian groups, could be therefore of South Asian origin as well. MA-1 could well be a group that migrated Northward around 30,000 YBP after the formation of the R clade, just after the migration of Early Palaolithic Europeans. However, it failed to contribute to the ancestry of pre-Neolithic Europeans. It could only contribute its ancestry to later Siberians like AG & EHG.

-----------------------------------------

(continued...)

Jaydeepsinh Rathod said...

(...continued)

Coming to post-LGM population movements, it seems to me very likely that in comparison to the Neat East as well as Iran, South Asia would have been much more suitable as a vast LGM population refugium. So while during the LGM, most populations across the Northern hemisphere would have suffered bottlenecks, South Asia would have fared significantly better. Post-LGM, we can therefore envisage a scenario in which groups from South Asia began migrating Northward & Westward into sparsely inhabited regions, sensing better weather & vast empty spaces. This maybe how the CHG reached the Caucasus & the Iranian Mesolithic sample reached the Hotu cave South of the Caspian. Incidentally, in the modified 3D PCA, Iranian Hotu sample clusters most closely with modern South Asians, even more than Iran Neolithic.

http://www.open-genomes.org/analysis/PCA/Eurogenes_Lazaridis_(2016)_Near_Eastern_aDNA_PC_plot_1-2-3.html

-----

Iran_N is descending perhaps from one of those migrants from South Asia. The Gallego_Lorento paper clearly detected the ASI component in the Ganj Dareh sample. You can check the admixture graph & the author's comments in the Supplementary Material. As it is, The Iranian Neolithic group from the Zagros looks quite out of place in the Near East as they are quite distant from the Levantine & Anatolian farmers. On the other hand they are very close to modern South Asians and even the cultural assemblage of Zagros Neolithic groups is similar not to Levantine & Anatolian Neolithic but to South Asian Neolithic as represented by Mehrgarh.

As far as the steppe-related ancestry in South Asians, I won't make any comments right now. Let the aDNA from Rakhigarhi come. We shall be in much better position to comment on it then. But for the situation during Early Neolithic & even the Palaeolithic, South Asia looks more like a source of West Eurasian populations rather than a sink.

-------------------------

Admittedly I am being very South Asian centric but I feel my theory looks quite plausible. What do you have to say ?

ryukendo kendow said...

Jaydeep,

Here's an introduction to Basal Eurasian, read both the post and the paper it references:
http://eurogenes.blogspot.sg/2013/12/ancient-human-genomes-suggest-three.html

I think this idea, that East and West Eurasians came from the South, splitting at the boundary of India and Southeast Asia, looks plausible on first sight, but its becoming more and more untenable by the day.

First, South Asians are outliers when it comes to Denisovan ancestry; populations to the North, West and East of them all have less Denisova than Indians, except for Papuans. This makes India look more like a sink than a source.

Second, South Asians look like they preserve some Crown Eurasian ancestry that is neither East nor West Eurasian, but an outgroup to both. I.e. ancestry that split off after Basal Eurasian, but before East and West Eurasian split from each other. Other scenarios are also possible: there's some kind of mystery there. This means that East and West Eurasians must have split off elsewhere other than India.

Lastly, Oceania preserves ancestry from a pre-OoA event, which may also have affected South, Southeast and East Asia judging from Haplotype-based analysis, and its also looking like South, Southeast and East Asia may preserve some type of basal ancestry which we don't know yet how to define, which somehow results in similarities with East Africans. This may mean that their latest wave of ancestry is intrusive to the region.

I'm coming, more and more, to favour a split of East and West Eurasian ancestry in Siberia, or Central Asia, and will honestly be quite surprised if the original idea of a South-centric branching order is proven to be correct.

Shaikorth said...

"If Iran were so much more basal and less Neandertal, they would be much further from Ust_Ishim, but they're looking minimally closer. They're also significantly closer to ENA, which can make basal look higher in f4ratio."

I'd say Iran_N having less Neanderthal than Natufians is one of the more solid results in the paper since they used more ancient SNP's and a high coverage Mbuti sample to test it. The difference is clear, and Natufian has about as much Neanderthal as Iran_recent.

Natufians having something from Africa and more crown eurasian can explain why they are more distant from U-I but more Neanderthal, while Iran_N has more "real" basal.

Jared Knows said...

The Natufians represent a population that diverged mostly within Africa and contributed significantly to some West Eurasian populations & North Africa. The shared alleles are not from admixture with Sub-Saharan populations.

German Dziebel said...

@

"I'm coming, more and more, to favour a split of East and West Eurasian ancestry in Siberia"

America is a far better option but you are getting close to become another enlightened one. :)

Davidski said...

@Jaydeep

Whenever I read your posts, the first thing that comes into my mind if that you're trying to explain away the latest data to fit Out of India, as opposed to interpreting it as best you can.

You really need to be more objective, less fixated on irrelevant things like ambiguous signals in ADMIXTURE output, and more in tune with current developments.

For instance, you saying that Indians don't have any Basal Eurasian admixture or that Neolithic Iranians are part ASI because of a bit of something that they show in an ADMIXTURE run, immediately tells me that the rest of your posts aren't worth reading. And you know what, that's absolutely correct, currently your posts are not worth reading.

Snap out of it. Make a real contribution.

Chad Rohlfsen said...

Shaikorth,

If Natufians were less Basal and more Neandertal than Iran, they would be significantly closer to Ust Ishim. They're actually further from Ust Ishim than Iran and it approaches significance with Hotu. You're thinking backwards. Natufians having Neandertal closer to Ust Ishim means that it is potentially more Basal Eurasian in Natufians to not only pull even with Iran, but even further away.

Chad Rohlfsen said...

It is possibly ENA in Iran making them look more basal, artificially.

Chad Rohlfsen said...

ANE has the same effect, to a third of the degree of ENA in f4 ratios with Ist Ishim, Kostenki, and Loschbour as your 3 other pops.

Shaikorth said...

"If Natufians were less Basal and more Neandertal than Iran, they would be significantly closer to Ust Ishim. They're actually further from Ust Ishim than Iran and it approaches significance with Hotu. You're thinking backwards. Natufians having Neandertal closer to Ust Ishim means that it is potentially more Basal Eurasian in Natufians to not only pull even with Iran, but even further away. "

SSA combined with more crown in Natufians and Levant_N (also more distant from Ust-Ishim than Iran_N and even more Neanderthal than Natufians) will have the effect of pulling away from Ust-Ishim while adding Neanderthal. Their more exhaustive admixturegraphs also supported Iran_N being the most basal.

If Iran_N had less basal than Natufians and Levant_N and had noteworthy ENA to boot it would have more Neanderthal than them, not clearly less. All ENA populations have relatively high Neanderthal.

Chad Rohlfsen said...

Good lord, Iosif told me himself they model them as equally basal. SSA and Crown? There's no stat to suggest this. Fantasy land... You can't just throw shit at the wall and hope something sticks.

Shaikorth said...

No stat? Those treemix edges David gets for instance can explain it that way, and TreeMix is based on formals.

The models Lazaridis et al. included modeled them as equally basal or Iran_N as more basal, especially when they included more populations in the fit like in S4.11. And to properly reflect reality a model has to account for every population so models with more of them should be given more weight.

This is all to find out a model that fits the data. If Iran_N and Natufians had similar "basal" ancestry and rest from the crown we would not see what we are seeing now, they'd have same level of neanderthal and be equally close to U-I and their fst-position between papuans and mbuti should prolly be equal too. Iran_N has clearly less Neanderthal than Levant_N and Natufians. ENA or any Euro HG won't explain it being closer to Ust-Ishim because those are high neanderthal. Natufians having more neanderthal-rich admixture from a crown branch combined with something that is more distant from Ust-Ishim than Iran_N's Basal could. Something like that can be found from Africa.

Chad Rohlfsen said...

Treemix is not as solid as you think. Why do you think no paper bases conclusions on what is shown there? All pops that are West or East Eurasian need have much more Neandertal than others. For instance, Villabruna was close to 5% and Loschbour at 3%, still more than East Asians. qpGraph is not fool proof either. Remember the one that had CHG as just ANE plus Basal? That wasn't true and neither Iosif or Nick thought it was either. So, why don't you just trust what I'm telling you and what Iosif said to me?

Again, here it is.

Dear Chad,
We do not estimate that Iran Neolithic is 20% more Basal Eurasian than Natufians.
The mixture proportions are in Supplementary Table S4.9 (column when dropping MA1) and are 45+/-8 for Natufians and 48+/-6 for Iran_N.
These are not significantly different from each other, so there is no evidence that Natufians have more/less Basal Eurasian than Iran Neolithic.

Best,

Iosif

So, you can either believe Iosif and what I'm saying, or go your own way. If the author himself doesn't claim Iran is much more basal than Natufians, why do you?

Poise n Pen said...

Basal eurasians aren't even a real thing in the first place. If they were they would show up in UP europe, but they don't. Remember these are supposedly some group that all non african moderns descend from. Pure bunk.... The names chosen for many of these groups are incredibly loaded. Just make up whatever crap they like. Basal Europeans belong in the same rubbish bin as 'steppe ancestry'.

Karl_K said...

@Poise

"The names chosen for many of these groups are incredibly loaded. Just make up whatever crap they like."

Absolutely. Who are these people? Do you know any of them? I've never heard of them. It is just crazy that anyone even spends time on this rubbish.

And what's up with 'Neanderthals' and 'Denisovans'? If they even ever existed, then you'd see them in the Bronze Age. Totally manufactured BS! Show me a Denisovan skeleton. You can't because they never existed.

Poise n Pen said...

Wow, 99 comments in and you reply to that and with this as your response? lol

I am pretty sure I gave specific examples of what I am talking about.

Naming something "mediterranean component" or worse yet "steppe component" or "basal european" are all incredibly loaded. It's even worse when the word ancestry is used, basically the whole highly speculative art of ancient dna analysis is presented as firmly established fact when the reality is there is not even circumstancial evidence of ancestry.

For example basal eurasians is probably the worse offender. Suddenly these basal eurasians are "found" in ancient dna according to some. Suddenly everyone uses this term like its fact, as if we really all evolved from these guys!

Now basal europeans are meant to be the people who mythically came out of africa and, replacing neanderthals and denisovans, then split off into all the races of eurasia somehow.

In reality, this component shows up at 0% levels in paleolithic europe now that we have autosomal DNA to play with. Not only does it turn out this component is nothing to do with basal eurasians but the results show there was never such thing as basal eurasians in the first place. But people go on as if this is still a real thing.

Similarly we have r1b y-dna found in italy 14k years ago and we still have people talking about r1b y-dna coming into europe from the steppe 5k-7k years ago. What a joke.

Davidski said...

@Poise n Pen

You can't push shit up a steep hill.

Eventually you'll realize what the problem is when the shit rolls back and buries you.

Simon_W said...

@ Rob

The so called Remedello samples are RISE486, RISE487 and RISE489. According to Allentoft et al. 2015 Supplementary Table 1 only the latter two predate 2500 BC. RISE486 has calibrated C14 dates of 1773 BC - 2134 BC, thus it's post Beaker.

MIND Domine said...

I had a question. If all Europeans have Basal Eurasian in them since all Basal Eurasians had a few % of SSA in them (Out of Africa) then why does SSA show up for some Europeans on the Gedrosia K3 calculator but none for others?

On my Gedrosia K6 I had no SSA. Just Natufian but when I ran the K3 it came back with 1.58% SSA. This makes no sense.

Why doesn't the Basal SSA show up for everyone?

Davidski said...

all Basal Eurasians had a few % of SSA in them (Out of Africa)

Wrong assumption on your part.

On my Gedrosia K6 I had no SSA. Just Natufian but when I ran the K3 it came back with 1.58% SSA. This makes no sense.

I didn't design those tests, but it does make sense.

What seems to be happening is that the SSA-related ancestry that you have is subsumed within the Natufian or other clusters that appear at higher K and are formed partly from the SSA-related alleles.

MIND Domine said...

I'm confused. Don't all Europeans have a bit of SSA in them if you consider the last 50k years through Basal Eurasians? Out of Africa.

Also, is that old SSA different from modern SSA?

MIND Domine said...

Do you have any facts to back up that my assumption is wrong?

MIND Domine said...

Your wrong, look at the Description from the Ancient K6.

"Natufian: This was an Epipaleolithic culture that existed from 12,500 to 9,500 BC in the area of Israel.

"They were derived about 50% from an original Out-of-Africa population, referred to as Basal Eurasians."

If you are a European and show Natufian admixture, this does not imply that Natufians interacted with your ancestors. All it means is that Natufian like admixture was mediated to you via intermediaries, such as the early European Farmers from the Near East."

Davidski said...

Yes, you're confused. You're mixing up three very different populations.

- Out of Africa migrants

- Basal Eurasians

- Modern-day Africans

These populations are related in various ways, but very different nonetheless.

I can't help you any more than that. You need to do more reading.