The Yassitepe challenge

This is about the only successful qpAdm model that I can find for the pair of Early Bronze Age (EBA) females from Yassitepe, Turkey, using a decent set of outgroups and markers. I wouldn't take it too literally, but it does suggest a potentially significant level of European ancestry, including some steppe ancestry, in these Yassitepe individuals.

TUR_Aegean_Yassitepe_EBA
AZE_Caucasus_lowlands_LN 0.565±0.054
ROU_N 0.387±0.041
RUS_Progress_En 0.048±0.022
P-value 0.103248
Full output

If anyone reading this can find a better, more convincing solution then I'd love to see it. Feel free to share it in the comments below.

Obviously, both of the Yassitepe samples are from the recent Lazaridis, Alpaslan-Roodenberg et al. paper. Their EBA dating suggests that they might be relevant to the debate over the origins of Anatolian speakers, such as the Hittites and Luwians.

See also...

Dear Iosif, about that ~2%

The precursor of the Trojans

Dear Iosif, about that ~2%

The debate over the location of the so called Indo-Anatolian homeland won't be decided by the persistence of any type of genetic ancestry in ancient Anatolia.

It'll be decided by a multidisciplinary study on the interactions between the ancient peoples of the North Pontic steppe, the eastern Balkans, and western Anatolia.

If such a study finds a pulse of steppe-related gene flow from the Balkans into Anatolia sometime during the early metal ages, it'll corroborate the linguistic hypothesis that a language ancestral to Hittite, Luwian and related tongues moved into Anatolia from Eastern Europe.

Why do we only need a pulse of gene flow, you might ask? Obviously, because:

- language and genetic ancestry can start with a strong association but, since they're not linked, they can eventually follow very different trajectories

- the dilution of genetic ancestry is an important factor, especially in ancient West Asia, and it must be taken into account in models of language spread, rather than ignored in favor of simple, elegant models that do not reflect reality.

Here's my favorite quote from the recent Lazaridis, Alpaslan-Roodenberg et al. paper, because, probably unbeknownst to the authors, it's exceptionally revealing about the spread of a wide range of Indo-European speakers into Anatolia.

However, in individuals from Gordion, a Central Anatolian city that was under the control of Hittites before becoming the Phrygian capital and then coming under the control of Persian and Hellenistic rulers, the proportion of Eastern hunter-gatherer ancestry is only ~2%, a tiny fraction for a region controlled by at least four different Indo-European–speaking groups.

Indeed, this is exactly what the Lazaridis, Alpaslan-Roodenberg et al. paper should've been about. That is, the authors should've given us a painstaking account of the spread of different ancient Indo-European speaking groups into Anatolia and explained how, overall, their DNA was rapidly diluted to a trace amount.

However, instead they treated us to a make-believe tale about a so called Indo-Anatolian homeland in what is now Armenia.

See also...

Dear Iosif...Yamnaya

But Iosif, what about the Phrygians?

Dear Iosif...

Dear Iosif #2

Dear Iosif #3

But Iosif, what about the Phrygians?

A paper in Science co-authored by around 200 scientists from some of the world's top academic institutions surely must mean something, right? Not necessarily.

In this short blog post I'll try to explain, as simply as I can, why the Lazaridis, Alpaslan-Roodenberg et al. paper doesn't get us any closer to solving the riddle of the so called Indo-Anatolian homeland.

However, it must be said that the paper does include many interesting and valuable samples. I'll be using six of these samples, labeled TUR_C_Gordion_Anc, to argue my case.

The TUR_C_Gordion_Anc sample set is from Gordion, the capital of ancient Phrygia, and thus, in all likeliness, it represents Phrygian speakers.

Phrygian is an Indo-European language and the leading hypothesis is that it originated in the Balkans.

In terms of fine scale ancestry, TUR_C_Gordion_Anc can be reliably divided into two genetic clusters. In the Principal Component Analysis (PCA) below these clusters are labeled TUR_C_Gordion_Anc1 and TUR_C_Gordion_Anc2.

Note that TUR_C_Gordion_Anc1 is obviously pulling away from TUR_C_Gordion_Anc2 towards samples from the Balkans. I've used ancient samples from what is now North Macedonia, labeled MKD_Anc, to represent the Balkans. To see an interactive version of the plot, paste the PCA coordinates from here into the relevant field here.

Visually, this is not an especially dramatic outcome, but it's an incredible result nonetheless, because it shows that even a few ancient samples can help to solve an age old mystery.

Across many dimensions of genetic variation, the shift in the PCA from TUR_C_Gordion_Anc1 to TUR_C_Gordion_Anc2 represents about 20% admixture from the Balkans, and about 8% from the Eastern European steppe. That's plenty enough to corroborate the linguistic hypothesis that the Phrygians originated in the Balkans, and that some of their ancestors came from the steppe. The mixture models below were done with the tools here.

Target: TUR_C_Gordion_Anc1
Distance: 1.6634% / 0.01663373
40.6 Kura-Araxes_ARM_Kaps
22.2 Anatolia_Barcin_N
21.8 MKD_Anc
13.6 Levant_PPNB
1.4 IRN_Ganj_Dareh_N
0.4 Han

Target: TUR_C_Gordion_Anc1
Distance: 1.7109% / 0.01710904
40.2 Kura-Araxes_ARM_Kaps
37.8 Anatolia_Barcin_N
12.4 Levant_PPNB
8.0 Yamnaya_RUS_Samara
1.2 IRN_Ganj_Dareh_N
0.4 Han

Target: TUR_C_Gordion_Anc2
Distance: 2.0293% / 0.02029339
51.0 Kura-Araxes_ARM_Kaps
26.8 Anatolia_Tepecik_Ciftlik_N
17.6 Anatolia_Barcin_N
4.6 Levant_PPNB

Surprisingly, Lazaridis, Alpaslan-Roodenberg et al. didn't have much to say about this topic. This quote basically sums it up:

However, in individuals from Gordion, a Central Anatolian city that was under the control of Hittites before becoming the Phrygian capital and then coming under the control of Persian and Hellenistic rulers, the proportion of Eastern hunter-gatherer ancestry is only ~2%, a tiny fraction for a region controlled by at least four different Indo-European–speaking groups.

I have no doubt that Lazaridis, Alpaslan-Roodenberg et al. can run a very decent PCA, and then blow it up to a size big enough to show that the Gordion samples represent two genetically somewhat distinct groups. I'm also sure that, if they really try, they can locate significant levels of proximate and relevant European ancestry in some of these samples.

They don't have to use my methods; they can use any methods they like. My point is that they won't find much if they're just looking for genetic signals from the Upper Paleolithic or Mesolithic.

Now, considering the way that the Phrygian question was treated by Lazaridis, Alpaslan-Roodenberg et al., despite the fact that they managed to sequence a few likely Phrygian speakers from none other than the Phrygian capital, let's not pretend that their paper brought us any closer to understanding the genetic origins of Anatolian speakers or pinpointing their ancestral homeland.

In order to even try to solve these problems with ancient DNA, we need a wide range of samples from Hittite, Luwian and other key sites where Anatolian languages were spoken. And then we must analyze them properly.

I'm guessing that Lazaridis, Alpaslan-Roodenberg et al. went out of their way to get such samples, but for one reason or another they failed. If so, that's OK, but I have a feeling that even if they got them, they wouldn't know what to do with them, because at best these samples would only show ~2% Eastern hunter-gatherer ancestry. Haha.

For what it's worth, I believe that the ancient data in the Lazaridis, Alpaslan-Roodenberg et al. paper point to the North Pontic steppe as the Indo-Anatolian homeland, and I'll lay out my arguments in an upcoming blog post.

See also...

Dear Iosif...Yamnaya

Dear Iosif, about that ~2%

Dear Iosif...

Dear Iosif #2

Dear Iosif #3

Dear Iosif...

Update 29/08/22: Dear Iosif #2

...


I'm skimming through the Lazaridis, Alpaslan-Roodenberg et al. paper that just came out at Science. And I feel like someone punched me in the face.

Nevertheless, I'll try to be diplomatic. Suffice to say, for now, that there's some rather strange stuff in this paper.

The main problem is that the authors are attempting to study fine scale ancestry with a somewhat rough distal model. As a result, they miss important details.

For instance, this quote is from the paper's supplementary PDF file, freely available here.

However, the complete lack of association of R-haplogroup descendants and EHG ancestry in either Armenia or Iran is consistent with either a massive dilution of EHG ancestry in these populations resulting in the dissociation of Y-chromosome lineages from autosomal ancestry over time, or with a scenario in which R-M269 was not associated with substantial EHG ancestry to begin with.

Obviously, EHG means Eastern European Hunter-Gatherer. But why focus on EHG? Surely, this makes little sense when looking at the genetic prehistory of West Asia, because no one ever argued that this region was settled by EHG populations. It was widely settled by Yamnaya-related groups, with already heavily diluted EHG ancestry, during the metal ages.

OK, so the authors are actually aware about the potential dilution of EHG ancestry, but they don't really do anything about it.

If we're looking at the origins of West Asian R1b-M269, and using its association with autosomal DNA components as a guide, then we should be focusing on Yamnaya-related ancestry.

For instance, here's a fine scale ancient ancestry model based on Principal Component Analysis (PCA) data. It shows the ancestry proportions of two relatively high coverage Iron Age males from two different sites in Iran from the Lazaridis, Alpaslan-Roodenberg et al. dataset. Both belong to R1b-M269 and both show significant Yamnaya-related ancestry.

Target: IRN_HajjiFiruz_IA:I2327_all
Distance: 2.2930% / 0.02292994
39.6 Kura-Araxes_ARM_Kaps
24.2 IRN_Ganj_Dareh_N
18.2 Levant_PPNB
12.4 Yamnaya_RUS_Samara
4.4 Anatolia_Tepecik_Ciftlik_N
1.2 Han

Target: IRN_Hasanlu_IA:I4232_all
Distance: 2.5179% / 0.02517895
26.0 IRN_Ganj_Dareh_N
25.6 Kura-Araxes_ARM_Kaps
24.4 Anatolia_Tepecik_Ciftlik_N
15.8 Yamnaya_RUS_Samara
7.6 Levant_PPNB
0.6 IRN_Shahr_I_Sokhta_BA2

As a control, here's an earlier, Chalcolithic sample bearing Y-haplogroup J2b from the same region. Not surprisingly, this individual totally lacks the Yamnaya-related signal.

Target: IRN_HajjiFiruz_ChL:I4241_all
Distance: 2.7938% / 0.02793782
32.6 Kura-Araxes_ARM_Kaps
25.6 IRN_Ganj_Dareh_N
23.6 Anatolia_Tepecik_Ciftlik_N
18.2 Levant_PPNB

Overall, these results make perfect sense. I could probably locate very minor signals of EHG ancestry in the Iron Age samples, but that would be more difficult and much less certain, so I won't bother.

Soon I'll be able to rerun these analyses with Bronze Age samples from Dagestan and surrounds. That should bump up the levels of Yamnaya-related ancestry and improve the statistical fits (wink, wink, nudge, nudge).

Disappointingly, Lazaridis, Alpaslan-Roodenberg et al. go so far as to suggest that R1b-M269 may have originated in West Asia.

However, considering the scores of ancient Eastern European populations rich in R1b-M269 and many near and far related subclades of R1b, this makes no sense whatsoever.

Indeed, contemplating nowadays that R1b-M269 might be native to West Asia, where R1b only starts showing up in the ancient DNA record during the Copper Age, is about as stupid as claiming that gravity doesn't exist.

Largely due to their distal model approach, Lazaridis, Alpaslan-Roodenberg et al. also argue that the Indo-Anatolian homeland was located in what is now Armenia and surrounds. I'm far from convinced that this solution will stand the test of time.

In terms of the more widely accepted theory that the Indo-Anatolian homeland was located on the Pontic-Caspian steppe in Eastern Europe, the most important samples in the paper are the three Bronze Age individuals from Yassitepe in western Anatolia. That's because they're from a region that is traditionally seen as the entry point of Indo-Anatolian speakers into Anatolia from the European steppe via the Balkans.

Interestingly, individual I5737, dated to 2035-1900 calBCE or the Middle Bronze Age, belongs to Y-chromosome haplogroup I2a-P78, which surely must be a signal of European ancestry. I see this as a significant result.

Here's how the trio from Yassitepe look in my fine scale ancient ancestry model. Minor Yamnaya-related ancestry does show up, although, admittedly, it might just be noise in individual I5735.

Target: TUR_Aegean_Izmir_Yassitepe_MBA:I5737
Distance: 2.7507% / 0.02750748
58.4 Anatolia_Barcin_N
20.4 Kura-Araxes_ARM_Kaps
9.2 Anatolia_Tepecik_Ciftlik_N
5.6 IRN_Ganj_Dareh_N
3.8 Yamnaya_RUS_Samara
2.6 Levant_PPNB

Target: TUR_Aegean_Izmir_Yassıtepe_EBA:I5733
Distance: 2.7969% / 0.02796887
52.0 Anatolia_Barcin_N
27.2 Kura-Araxes_ARM_Kaps
8.6 Levant_PPNB
6.2 Yamnaya_RUS_Samara
6.0 IRN_Ganj_Dareh_N

Target: TUR_Aegean_Izmir_Yassıtepe_EBA:I5735
Distance: 3.1270% / 0.03127009
36.0 Kura-Araxes_ARM_Kaps
32.4 Anatolia_Tepecik_Ciftlik_N
26.0 Anatolia_Barcin_N
2.8 IRN_Shahr_I_Sokhta_BA2
1.2 Yamnaya_RUS_Samara
1.0 Levant_PPNB
0.6 MAR_Taforalt

This isn't much, especially considering it's already late 2022, but it's better than nothing. Fortunately, more samples from Bronze Age western Anatolia are on the way (wink, wink, nudge, nudge).

However, I'm not done with the Lazaridis, Alpaslan-Roodenberg et al. dataset yet. I'm planning to spend much more time on this blog in the coming weeks and months and will be using their samples in a wide range of analyses.

Citation...

Iosif Lazaridis, Songül Alpaslan-Roodenberg et al., The genetic history of the Southern Arc:A bridge between West Asia and Europe, Science 377, eabm4247 (2022)

See also...

Dear Iosif #3

But Iosif, what about the Phrygians?

Dear Iosif, about that ~2%

Dear Iosif...Yamnaya

The precursor of the Trojans

Who remembers kum4 from Omrak et al. 2016? I'm pretty sure now that this individual packs a lot of ancestry from the Pontic-Caspian (PC) steppe.

If so, that's a big deal, because her Chalcolithic (or Late Neolithic?) burial was located at Kumtepe. That is, in the same part of Anatolia as the later settlement of Troy, which may have been founded by early Anatolian speakers from Eastern Europe (see here).

The qpAdm mixture models below, featuring kum4 and the likely older kum6, also from Kumtepe, are based on qpfstats output. qpfstats is a new program from the David Reich Lab specifically designed to help analyze low coverage ancients (see here). And kum4 is certainly that.

TUR_Kumtepe_N_kum4
RUS_Progress_En 0.383±0.114
TUR_Barcin_N 0.617±0.114
chisq 7.868
tail prob 0.247957
Full output

TUR_Kumtepe_N_kum4
IRN_Seh_Gabi_C 0.325±0.150
TUR_Barcin_N 0.675±0.150
chisq 14.736
tail prob 0.0224096
Full output

TUR_Kumtepe_N_kum6
RUS_Progress_En 0.121±0.042
TUR_Barcin_N 0.879±0.042
chisq 21.790
tail prob 0.00132149
Full output

TUR_Kumtepe_N_kum6
IRN_Seh_Gabi_C 0.283±0.059
TUR_Barcin_N 0.717±0.059
chisq 6.289
tail prob 0.391566
Full output

Indeed, kum4 and kum6 offer just ~10,000 and ~100,000 "valid SNPs", respectively (see here). However, if nothing else, the results are clearly not random.

For one, because they fit the expected pattern, with the likely older individual lacking ancestry from the PC steppe (her model with RUS_Progress_En shows a weak statistical fit). Moreover, the qpAdm mixture ratios align almost perfectly with the results in my Principal Component Analysis (PCA) of ancient West Eurasian genetic variation. Coincidence?

See also...

Perhaps a hint of things to come

Perhaps a hint of things to come

It's still a mystery how the Hittites and other Anatolian speakers ended up in the Near East. However, the leading theory is that their ancestors migrated from the steppes of Eastern Europe to western Anatolia via the Balkans sometime during the Copper Age.

Consider the qpAdm mixture models below, made possible thanks to some of the ancient samples published recently along with Skourtanioti et al. 2020. The key ancients are described in a text file available here.

TUR_Barcin_C
AZE_Caucasus_lowlands_LN 0.471±0.094
RUS_Vonyuchka_En 0.148±0.040
TUR_Barcin_N 0.381±0.069
chisq 12.874
tail prob 0.116261
Full output

TUR_Barcin_C
RUS_Vonyuchka_En 0.107±0.029
TUR_Buyukkaya_EC 0.893±0.029
chisq 12.107
tail prob 0.207331
Full output

I'd say it's quite clear now that TUR_Barcin_C harbors minor ancestry from the Pontic-Caspian (PC) steppe. The reason this isn't widely accepted yet is because demonstrating it convincingly hasn't been possible without a proximate Anatolian ancestry source for TUR_Barcin_C, precisely like TUR_Buyukkaya_EC.

Admittedly, though, the statistical fits in my models aren't all that great. I suspect the problem lies with RUS_Vonyuchka_En, which is likely to be a rather poor stand in for the people who brought steppe ancestry, and possibly early Anatolian speech, to western Anatolia.

So let's see what happens when I try a more proximate reference for the steppe ancestry in TUR_Barcin_C. How about Yamnaya_BGR, an individual of mixed Balkan and steppe origin from what is now Bulgaria?

TUR_Barcin_C
AZE_Caucasus_lowlands_LN 0.518±0.075
TUR_Barcin_N 0.203±0.056
Yamnaya_BGR 0.279±0.067
chisq 10.602
tail prob 0.225269
Full output

TUR_Barcin_C
TUR_Buyukkaya_EC 0.749±0.058
Yamnaya_BGR 0.251±0.058
chisq 9.687
tail prob 0.376414
Full output

That's a little better. Unfortunately, the problem now is that the models are anachronistic, because TUR_Barcin_C is about a thousand years older than Yamnaya_BGR. Clearly, we need more Copper Age samples from the western edge of the PC steppe, the eastern Balkans, and especially northwestern Anatolia.

The Principal Component Analysis (PCA) below effectively illustrates why my qpAdm models work. It was produced with Global25 data using the Vahaduo PCA tools freely available here. Note that TUR_Barcin_C is shifted away from the essentially perfect cline formed by AZE_Caucasus_lowlands_LN, TUR_Barcin_N and TUR_Buyukkaya_EC towards samples from ancient Eastern Europe, including Yamnaya_BGR.

See also...

Steppe invaders in the Bronze Age Balkans