PDA

View Full Version : Comparing different sources of Neolithic ancestry across the Balkans and Italy



XXD
12-02-2020, 02:04 PM
Hey guys. I was thinking that one way of tracking Slavic admixture and possibly earlier Neolithic/Bronze Age migrations could be achieved by tracking down the different sources of Neolithic ancestry that they carried. I am new to using G25, so apologies if I have made some technical errors.

However, when I try different models, it is clear that several populations of the North West Balkans (Serbians, Albanians, Croatians) carry a lot of Neolithic ancestry from the Serbian Neolithic Starcevo culture (or something similar). Areas of Greece also affected by the Slavic migrations also carry significant amounts of Starcevo ancestry, while the islands considerably less so, which makes sense, as upcoming studies suggest that the invading slavs were already genetically very similar to a Serbian/southern shifted population.

I have not used Anatolian Neolithic populations, as they tend to suck up all the Neolithic ancestry from the tested populations, presumably because it is ancestral. Interestingly, the Neolithic of Sicily or Sardinia fail to capture any ancestry in the modern Italian populations I used.

It is very interesting to see that in all models used, an Iron Age "Illyrian?" from Croatia (HRV_IA) derives all of his Neolithic ancestry from Starcevo, under all models I have tested. An allegedly Greek woman from 500 AD Bavaria (DEU_MA_o) gets most of her ancestry from Neolithic Peloponnese, but also a smaller portion from Starcevo, which suggests that she was a Greek mixed with a north-western Balkan population (e.g. Illyrian/Moesian).

Although I am sure the models can be improved considerably, I think two patterns are clear:

1) The western Balkans (both ancient and modern) derive most of their Neolithic ancestry from a Western Balkan source

2) Mainland Italy appears to derive most of its current Neolithic ancestry from the Greek-like source, although Central and North-eastern Italy derives a large chunk of its ancestry from a Starcevo-like source

2) Starcevo ancestry (and Bulgarian Varna) can be used to track the movements of Slavic and Albanian people further South, to the extent that modern Greek Neolitchic ancestry appears as a mixture of Peloponnese Neolithic, Serbian Starcevo and Bulgarian Varna.

Removing the Varna culture from the mix assigns all non-Peloponnese Neolithic ancestry to Starcevo (third image). This shows that both Greek and Italian populations derive their Neolithic ancestry from several populations.

What do you guys think, is there some accuracy to these models, and what could they indicate about the history and prehistory of these areas?

How can we improve these models?

41525


41526

XXD
12-02-2020, 02:06 PM
41527

41528

TonyC
12-02-2020, 05:21 PM
I think your model makes sense but I'm just a casual observer. I posted my own results using your model.

Distance: 3.2031% / 0.03203123
Target: Anthony_C_scaled
42.3 SRB_Starcevo_N
28.2 Yamnaya_RUS_Samara
21.3 GRC_Peloponnese_N
4.9 GEO_CHG
3.3 Levant_Natufian


Distance to: Anthony_C_scaled
0.08764657 BGR_Varna_C:ANI159-ANI181
0.10794745 SRB_Starcevo_N
0.11267346 GRC_Peloponnese_N
0.13963340 ITA_Grotta_Continenza_N_o
0.21887987 Yamnaya_RUS_Samara
0.24303958 Levant_Natufian
0.25102023 GEO_CHG
0.35851605 Baltic_LVA_HG

dosas
12-02-2020, 05:50 PM
You need Iran Neo. in that list. A lot of the fits are bad, over 2-3%.

lacreme
12-02-2020, 06:02 PM
My friend's and his mother's results using all the populations that you've posted

Strangely,for my friend mostly, he doesn't score any Starcevo or Varna even though he is at least a quarter Peloponnesian (from Elis ).
His mother is quite close to Crete 2. What kind of populations comprise this average ?

Distance: 2.2648% / 0.02264827
Target: Chris_scaled
58.2 GRC_Peloponnese_N
29.0 Yamnaya_RUS_Samara
6.5 Levant_Natufian
6.3 GEO_CHG


Distance: 3.3129% / 0.03312855
Target: ChrisMom_scaled
52.2 GRC_Peloponnese_N
24.4 Yamnaya_RUS_Samara
12.1 GEO_CHG
11.3 Levant_Natufian

@dosas
by Iran Neo do you mean Iran Gang Dareh N ? If so, their results change to the following with marginally improved fits but Starcevo (and Varna) appear which based on OPs explanation look more probable.

Distance: 1.9292% / 0.01929185
Target: Chris_scaled
50.0 GRC_Peloponnese_N
27.1 Yamnaya_RUS_Samara
8.0 SRB_Starcevo_N
7.4 IRN_Ganj_Dareh_N
4.4 Levant_Natufian
2.0 BGR_Varna_C
1.1 GEO_CHG

Distance: 3.1522% / 0.03152243
Target: ChrisMom_scaled
44.1 GRC_Peloponnese_N
22.8 Yamnaya_RUS_Samara
9.6 Levant_Natufian
9.4 SRB_Starcevo_N
7.8 GEO_CHG
6.3 IRN_Ganj_Dareh_N

dosas
12-02-2020, 06:43 PM
My friend's and his mother's results using all the populations that you've posted



The West Asian admix-ed Greeks require Iran_Neo to be modelled with good fits (sub 2% distances), Kotias GEO_CHG acts as proxy but with a worse fit, for obvious reasons. From what we've seen so far, GEO_CHG exists in Pontic Greeks who require it to be modelled properly as a (South) Caucasus derived population, it's not particularly relevant to the rest of the Greek groups who model better with Iran_N and Levant_PPNB.

XXD
12-02-2020, 06:53 PM
I will try to find some time later and try to improve the models. However, it does seem that there is some genuine signal going on. Even Mycenaeans don't score as pure Neolithic Peloponnese, as they require Varna as a source, which could suggest they themselves were mixed prior or soon after their entry to Greece.

I will be interested to see your further comments!

XXD
12-02-2020, 09:46 PM
The model gets considerably better for most groups if I add Kura Araxes for the Caucasian ancestry. I also added two Anatolian Neolithic sources (Barcin, Tepecik), and although they suck out most of Peloponnese Neolithic (likely artificially), the results for Serbia Starcevo and Varna remain largely unchanged. This could mean that the pattern is actually real.

I removed WHG and Baltic HG, as they were 0 everywhere. I also added Romanian Neolithic, as it works better for modern Bulgarians.

However, modern South Slavs still require additional components, as the fits are quite bad.

41534

Sorcelow
12-02-2020, 10:20 PM
The reality is that all Balkanites have HG ancestry from the Baltic region through their Slavic ancestry. It's not showing up because the WHG and Baltic_LVA_HG samples are too distant and lack modern Balto-Slavic drift.

You can use this as a proxy, it mimics Narva HG's but carries modern drift:

Baltic_HG_Simulated,0.149868802,0.136380306,0.2097 30848,0.190840117,0.156106048,0.072437149,0.038002 007,0.063688237,0.046382364,-0.072430537,-0.000902502,-0.05876715,0.106677275,0.144928109,-0.071019374,-0.005004616,0.021966607,0.003931878,-0.009757506,0.005758381,0.003759276,-0.024595198,0.007626296,-0.077723948,0.014953698

23abc
12-03-2020, 02:06 AM
Scoring more Starcevo_N than Peloponnese_N is certainly a hint of a northern shift but at the same time you cannot measure fairly the Anatolian admixture in the Greek islands with such a model.

Here's me and my mother with the same model (but with added Iran_N as dosas suggested).

Distance: 2.5934% / 0.02593441
Target: 23abc
55.5 GRC_Peloponnese_N
16.3 Yamnaya_RUS_Samara
10.5 IRN_Ganj_Dareh_N
9.0 Levant_Natufian
8.7 GEO_CHG

Distance: 3.6090% / 0.03609029
Target: 23abc_Mother
33.9 SRB_Starcevo_N
19.5 GRC_Peloponnese_N
17.4 GEO_CHG
10.1 IRN_Ganj_Dareh_N
10.0 Yamnaya_RUS_Samara
9.1 Levant_Natufian

As you can see, my mother scores much less Steppe than me and yet still scores more Starcevo. This makes sense to me, because I noticed she must have a significant mainland shift (on top of a significant Eastern-Anatolian shift) due to the fact that she gets improved fits with Balkan/Northern Italian sources and gets a few distant relative matches from deep within the Balkans (Albania, North Macedonia, Bulgaria) which are totally absent from my father's side.

On the same note, you cannot use this to measure the actual % of impact. I score no Starcevo even though my mother has a lot.

Greekscholar
12-03-2020, 03:05 PM
Scoring more Starcevo_N than Peloponnese_N is certainly a hint of a northern shift but at the same time you cannot measure fairly the Anatolian admixture in the Greek islands with such a model.

Here's me and my mother with the same model (but with added Iran_N as dosas suggested).

Distance: 2.5934% / 0.02593441
Target: 23abc
55.5 GRC_Peloponnese_N
16.3 Yamnaya_RUS_Samara
10.5 IRN_Ganj_Dareh_N
9.0 Levant_Natufian
8.7 GEO_CHG

Distance: 3.6090% / 0.03609029
Target: 23abc_Mother
33.9 SRB_Starcevo_N
19.5 GRC_Peloponnese_N
17.4 GEO_CHG
10.1 IRN_Ganj_Dareh_N
10.0 Yamnaya_RUS_Samara
9.1 Levant_Natufian

As you can see, my mother scores much less Steppe than me and yet still scores more Starcevo. This makes sense to me, because I noticed she must have a significant mainland shift (on top of a significant Eastern-Anatolian shift) due to the fact that she gets improved fits with Balkan/Northern Italian sources and gets a few distant relative matches from deep within the Balkans (Albania, North Macedonia, Bulgaria) which are totally absent from my father's side.

On the same note, you cannot use this to measure the actual % of impact. I score no Starcevo even though my mother has a lot.

Good point to make. I normally don't use Barcin and Tepecik in the same model for this reason.

Target: GS_family(Fourni/Ikaria/Samos):GS(Fourni/Ikaria/Samos)
Distance: 2.4745% / 0.02474513
70.2 Anatolia_Tepecik_Ciftlik_N
25.2 Yamnaya_RUS_Samara
2.4 Levant_Natufian
1.8 IRN_Ganj_Dareh_N
0.4 GEO_CHG

Target: GS_family(Fourni/Ikaria/Samos):YiaYia(Fourni/Ikaria)
Distance: 2.3467% / 0.02346744
42.0 Anatolia_Barcin_N
21.4 Yamnaya_RUS_Samara
20.2 Anatolia_Tepecik_Ciftlik_N
8.6 IRN_Ganj_Dareh_N
4.6 Levant_Natufian
1.8 Baltic_LVA_HG
1.4 GEO_CHG

It's not possible my grandmother has 42% Barcin ancestors while I have none. Using both sources together can be useful in picking up island vs. mainlander ancestry as Tepecik is more likely to be assigned to islander like genetic profiles in my experience, but it can't be taken literally.

Here are some family samples with the two sources being discussed.


Target: GS_family(Fourni/Ikaria/Samos):YiaYia(Fourni/Ikaria)
Distance: 2.4291% / 0.02429118
60.2 GRC_Peloponnese_N
17.8 Yamnaya_RUS_Samara
7.8 Levant_Natufian
6.4 IRN_Ganj_Dareh_N
3.8 Baltic_LVA_HG
3.8 GEO_CHG
0.2 RUS_Devils_Gate_Cave_N


Target: GS_family(Fourni/Ikaria/Samos):GS_Dad(Fourni/Ikaria/Samos)
Distance: 2.9728% / 0.02972806
58.4 GRC_Peloponnese_N
24.8 Yamnaya_RUS_Samara
10.8 Levant_Natufian
3.0 GEO_CHG
2.4 IRN_Ganj_Dareh_N
0.4 Baltic_LVA_HG
0.2 RUS_Devils_Gate_Cave_N

Target: GS_family(Fourni/Ikaria/Samos):GS(Fourni/Ikaria/Samos)
Distance: 2.9698% / 0.02969768
40.6 GRC_Peloponnese_N
21.0 Yamnaya_RUS_Samara
17.2 SRB_Starcevo_N
11.2 Levant_Natufian
8.4 GEO_CHG
1.6 IRN_Ganj_Dareh_N

dosas
12-03-2020, 04:16 PM
I don't want to come across as a party-pooper, I really don't, but all this seems like mental gymnastics to dance around the recorded historical facts of the Medieval migrations.

Northern input in Greek groups come in all sorts and sizes, but the majority is either

a) Albanian - Arvanite, mass migrations during 13-14th century,
b) Aromanian - Latin Roman,
c) Bulgarian, 6th-7th century,

or a combination of those three.

XXD
12-03-2020, 07:50 PM
Thank you all for your replies! So, the bottom line is...We cannot say much about the medieval Slavic migrations until we have more samples from that period...Plus preinvasion Greeks and other Balkaners!

XXD
12-03-2020, 07:52 PM
Using Tepecik Neolithic also sucks up most of my Neolithic ancestry. However, under all other models, most my Neolithic ancestry comes from Starcevo, while I score zero from Peloponnese Neolithic!

Greekscholar
12-03-2020, 08:38 PM
Using Tepecik Neolithic also sucks up most of my Neolithic ancestry. However, under all other models, most my Neolithic ancestry comes from Starcevo, while I score zero from Peloponnese Neolithic!

You can model the other EEF samples using Barcin + the rest of the neolithic model. Peloponnese_N is slighly West Asian shifted, Tepecik is more West Asian shifted, Starcevo is north shifted. This causes the samples to act in different ways with the other modeling components. Tepecik, the more West Asian shifted EEFs, are likely the ancestral group for Greeks and Southern Italians. Using it in models does suck up the other components though, so it is important to understand when you see that Source sample it already includes Levant+Iran_N/GEO_CHG ancestry in it.

Target: Anatolia_Tepecik_Ciftlik_N
Distance: 2.2553% / 0.02255260
78.8 Anatolia_Barcin_N
10.4 Levant_Natufian
8.6 GEO_CHG
2.2 IRN_Ganj_Dareh_N

Target: SRB_Starcevo_N
Distance: 2.3148% / 0.02314842
91.8 Anatolia_Barcin_N
6.0 WHG
2.2 Yamnaya_RUS_Samara

Target: GRC_Peloponnese_N
Distance: 1.3464% / 0.01346409
93.8 Anatolia_Barcin_N
4.2 IRN_Ganj_Dareh_N
2.0 GEO_CHG

Greekscholar
12-03-2020, 08:44 PM
Thank you all for your replies! So, the bottom line is...We cannot say much about the medieval Slavic migrations until we have more samples from that period...Plus preinvasion Greeks and other Balkaners!

True. You can make some interesting models using Roman Imperial samples + Hun_Avar though. The Roman Imperial samples are still very genetically close to Island Greeks, especially Kos in G25. Adding this to those Dark Age era samples creates a solid idea of how much various Greek regional populations have changed since Roman times.

Some users (Michalis and Agamemnon come to mind) have seperated out the Roman Imperial samples by natural groupings, outliers, Aegean-like, Levant-like, etc. If you search their posts you will find these data sets. IMO, they are probably the best matches we have for modern Island Greeks except maybe the Roopkund B samples dating from Ottoman times.

XXD
12-07-2020, 05:55 PM
I have plotted all major European Neolithic groups on the Eurasian PCA. This neatly shows the relationships between the different sources of Neolithic ancestry in modern Europeans, and might be useful when one decides which sources they will use for their Vahaduo models.

Added a few Minoans and ancient Greeks for comparison. Note that I added a HG by mistake.

41598