Ancient Rome: A genetic crossroads of Europe and the Mediterranean

[XenophobicPrussian]

So I've been working on a new admixture tool like nMonte which essentially produces similar results but much faster, maybe I'll release it later on if I see it working well, this allows me to seek parsimonious models automatically over a space of many possible sources, which using nMonte or Vahaduo would be too slow, given that the amount of possible subsets of k different elements of a set with n elements grows exponentially. This I think is useful to avoid the overfitting that happens when using nMonte or Vahaduo putting many sources all at once. In that case the algorithm having so many possibilities I think can get quite "imaginative" so to speak in order to get as small as possible error.
In any case, I was trying something for Bergamo, with the following set of possible sources: Anatolia_IA, GRC_Mycenaean, ITA_Ardea_Latini_IA, ITA_Boville_Ernica_Latini_IA, ITA_Etruscan, ITA_Grotta_Continenza_CA, ITA_Prenestini_tribe_IA, ITA_Proto-Villanovan, ITA_Rome_Imperial, ITA_Rome_Latini_IA, ITA_Villanovan, Levant_LBN_Roman, SWE_IA, SWE_Viking_Age_Sigtuna, Celt_I8206, Lombard_CL146.
The last two are samples with ID which are respectively a (likely?) Celtic proxy from Empuries in Roman era Spain and a Lombard from Collegno which seems fairly representative.

Here are the results for 3, 4 and 5 sources:

X	Distance	ITA_Etruscan	ITA_Rome_Imperial	SWE_Viking_Age_Sigtuna
Italian_Bergamo	0.01186796	68.37509	18.03989	13.58502

X	Distance	ITA_Boville_Ernica_Latini_IA	ITA_Etruscan	ITA_Rome_Imperial	SWE_Viking_Age_Sigtuna
Italian_Bergamo	0.01024404	22.75319	42.76607	19.61784	14.8629

X	Distance	ITA_Ardea_Latini_IA	ITA_Boville_Ernica_Latini_IA	ITA_Etruscan	ITA_Rome_Imperial	SWE_Viking_Age_Sigtuna
Italian_Bergamo	0.009178887	21.74669	16.1414	23.88893	22.6091	15.61389

If you have any ideas so that I may check how good the algorithm is, feel free to propose something.

Use DEU_MA or Norwegian, 15 is too low for Bergamo most likely. Viking Age Sigtuna average has Balt/Finn/Slav/Briton outliers.

Also, overfitting is a myth and something completely made up by Davidski. Overfitting is a -good- thing.

[Calpurnius]

Use DEU_MA or Norwegian, 15 is too low for Bergamo most likely. Viking Age Sigtuna average has Balt/Finn/Slav/Briton outliers.

Also, overfitting is a myth and something completely made up by Davidski. Overfitting is a -good- thing.

DEU_MA climbs somewhat higher and gets around ~20%, though I preferred to choose one of the northwest European-like Lombard samples from collegno.

I wouldn't say overfitting is a myth, especially if you are dealing with many closely related sources. If the algorithm has many sources chosen randomly during each cycle, it will pretty much pick up whatever it takes to minimize the distance regardless of how realistic the model is. The monte carlo algorithm can't make miracles with just 25 dimensional data when attempting to resolve exactly the precise source within European ones, the results should be taken with a grain of salt.

[MinervaItalica]

In Sicily there was a very important N.Italian repopulation though aside with some norman impact in specific cities.

Agree

The House of Hohenstaufen and their successors (Capetian House of Anjou and Aragonese House of Barcelona) gradually "Latinized" Sicily over the course of two centuries, and this social process laid the groundwork for the introduction of Latin (as opposed to Byzantine) Catholicism. The process of Latinization was fostered largely by the Roman Church and its liturgy. The annihilation of Islam in Sicily was completed by the late 1240s, when the final deportations to Lucera took place.[34] By the time of the Sicilian Vespers in 1282 there were no Muslims in Sicily and the society was completely Latinized.

https://en.wikipedia.org/wiki/Emirat...cily#Aftermath

lol 41 of 48 Imperial Roman samples are over 50% recent Middle-Eastern(Latins and Etruscans 0% except for 2 outliers), 66% Middle-Eastern overall and you have the audacity to post this on a thread where the paper is linked.

I was talking about Middle Ages immigration to Italy of Germanics, Byzantine and Arabs not the mutt that was Imperial Rome.

Imperial Rome was like nowadays US.

In any case there are new studies almost every year and this one will be debunked like previous ones. Just wait and you'll see.

And...

Conversely, the foreign population of Rome was very small. Scientific evidence has shown that non-locals were in the minority, and most came from other parts of Italy or nearby provinces in Southern/Central Europe. Biochemical analysis of 166 skeletons from three non-elite imperial-era cemeteries in the vicinity of Rome revealed that only 1 individual definitely came from outside of Europe (North Africa), and another 2 possibly did, but results are inconclusive.[31][32] Other than that, there was no apparent net inter-regional migration in the imperial period, except perhaps a small continuous resettlement of Easterners in the West.[33]

https://en.wikipedia.org/wiki/Demogr...pire#Migration

[Creoda]

Target: Sicilian_West
Distance: 1.0170% / 0.01016960
Aggregated
20.4 Levant_ISR_Ashkelon_IA1
19.2 DEU_MA
15.4 Levant_ISR_Ashkelon_LBA
14.8 ITA_Etruscan
9.8 ITA_Boville_Ernica_Latini_IA
6.6 Anatolia_Kaman-Kalehoyuk_MLBA
4.8 ITA_Ardea_Latini_IA
3.4 Levant_ISR_Ashkelon_IA2
2.8 Morocco_Iberomaurusian
1.6 Anatolia_IA
1.0 ITA_Proto-Villanovan
0.2 England_Roman_Near_Eastern_o

Target: Sicilian_East
Distance: 0.7614% / 0.00761360
Aggregated
23.0 Anatolia_Kaman-Kalehoyuk_MLBA
17.2 Levant_ISR_Ashkelon_IA1
16.2 DEU_MA
15.2 ITA_Proto-Villanovan
13.2 ITA_Boville_Ernica_Latini_IA
9.4 Levant_Canaanite_MBA
4.2 England_Roman_Near_Eastern_o
1.2 Morocco_Iberomaurusian
0.2 GRC_Mycenaean
0.2 Levant_ISR_Ashkelon_LBA

Target: Italian_Veneto
Distance: 0.8889% / 0.00888867
Aggregated
26.0 DEU_MA
19.2 ITA_Ardea_Latini_IA
15.6 ITA_Etruscan
13.2 Anatolia_Isparta_EBA
7.8 ITA_Villanovan
6.8 DEU_Lech_EBA
5.2 Levant_ISR_Ashkelon_LBA
3.4 DEU_Lech_MBA
2.8 ITA_Proto-Villanovan

Target: Italian_Umbria
Distance: 0.7621% / 0.00762093
Aggregated
20.2 Anatolia_Isparta_EBA
17.8 DEU_Lech_EBA
16.0 ITA_Etruscan
9.0 Anatolia_Kaman-Kalehoyuk_MLBA
8.2 ITA_Ardea_Latini_IA
7.6 Levant_ISR_Ashkelon_LBA
5.6 DEU_MA
3.8 ITA_Rome_Latini_IA
3.6 ITA_Boville_Ernica_Latini_IA
3.2 ARM_LBA
3.2 ITA_Proto-Villanovan
1.0 Morocco_Iberomaurusian
0.6 Anatolia_IA
0.2 ITA_Villanovan

Target: Italian_Tuscany
Distance: 0.6005% / 0.00600529
Aggregated
16.4 Anatolia_Isparta_EBA
15.6 DEU_Lech_EBA
13.6 ITA_Etruscan
13.6 DEU_MA
10.0 Anatolia_Kaman-Kalehoyuk_MLBA
9.6 ITA_Boville_Ernica_Latini_IA
4.8 England_Roman_Near_Eastern_o
4.8 ITA_Rome_Latini_IA
4.4 ITA_Proto-Villanovan
4.2 ITA_Ardea_Latini_IA
2.6 Levant_ISR_Ashkelon_LBA
0.4 DEU_Lech_MBA

Target: Italian_South
Distance: 0.6034% / 0.00603360
Aggregated
37.2 ITA_Etruscan
15.8 ARM_LBA
14.8 Levant_Canaanite_MBA
13.6 Anatolia_Isparta_EBA
5.0 Levant_ISR_Ashkelon_IA1
4.0 DEU_MA
3.8 ITA_Boville_Ernica_Latini_IA
2.4 Anatolia_IA
1.4 GRC_Mycenaean
1.2 England_Roman_Near_Eastern_o
0.6 Morocco_Iberomaurusian
0.2 ITA_Proto-Villanovan

Target: Italian_Piedmont
Distance: 0.5874% / 0.00587406
Aggregated
21.2 DEU_MA
14.6 ITA_Ardea_Latini_IA
14.4 Anatolia_Isparta_EBA
10.6 ITA_Proto-Villanovan
8.0 ITA_Boville_Ernica_Latini_IA
7.8 DEU_Lech_EBA
7.2 Levant_ISR_Ashkelon_LBA
7.0 ITA_Etruscan
5.4 GRC_Mycenaean
2.2 DEU_Lech_MBA
1.2 Anatolia_Kaman-Kalehoyuk_MLBA
0.2 Morocco_Iberomaurusian
0.2 England_Roman_Near_Eastern_o

Target: Italian_Naples
Distance: 0.9437% / 0.00943689
Aggregated
19.8 ITA_Etruscan
15.6 Levant_ISR_Ashkelon_LBA
15.0 Anatolia_Kaman-Kalehoyuk_MLBA
10.2 Levant_ISR_Ashkelon_IA1
9.6 DEU_MA
8.6 Anatolia_Isparta_EBA
6.6 ITA_Ardea_Latini_IA
4.6 ITA_Proto-Villanovan
4.4 ARM_LBA
3.0 England_Roman_Near_Eastern_o
1.6 ITA_Boville_Ernica_Latini_IA
1.0 Morocco_Iberomaurusian

Target: Italian_Molise
Distance: 0.8052% / 0.00805230
Aggregated
20.6 ITA_Proto-Villanovan
15.4 DEU_MA
14.6 Anatolia_Kaman-Kalehoyuk_MLBA
12.6 Levant_ISR_Ashkelon_LBA
8.6 ITA_Villanovan
6.6 GRC_Mycenaean
6.0 Anatolia_Isparta_EBA
5.6 ITA_Ardea_Latini_IA
5.4 ARM_LBA
3.8 Levant_ISR_Ashkelon_IA1
0.6 Levant_ISR_Ashkelon_IA2
0.2 DEU_Lech_MBA

Target: Italian_Marche
Distance: 0.6663% / 0.00666332
Aggregated
20.4 ITA_Etruscan
18.0 Anatolia_Isparta_EBA
15.8 DEU_MA
15.2 ITA_Proto-Villanovan
10.2 Levant_Canaanite_MBA
7.8 ITA_Boville_Ernica_Latini_IA
4.4 ITA_Rome_Latini_IA
3.4 Levant_ISR_Ashkelon_LBA
3.2 ARM_LBA
0.8 England_Roman_Near_Eastern_o
0.4 Anatolia_Kaman-Kalehoyuk_MLBA
0.2 Levant_ISR_Ashkelon_IA2
0.2 ITA_Ardea_Latini_IA

Target: Italian_Lombardy
Distance: 0.7876% / 0.00787622
Aggregated
25.2 ITA_Boville_Ernica_Latini_IA
24.8 ITA_Etruscan
23.0 ITA_Proto-Villanovan
6.8 ARM_LBA
5.4 ITA_Ardea_Latini_IA
5.4 DEU_MA
4.4 ITA_Rome_Latini_IA
2.8 Levant_ISR_Ashkelon_IA1
1.6 Anatolia_Isparta_EBA
0.4 Levant_ISR_Ashkelon_LBA
0.2 DEU_Lech_MBA

Target: Italian_Liguria
Distance: 1.4769% / 0.01476950
Aggregated
31.2 ITA_Etruscan
22.0 DEU_MA
16.4 Levant_ISR_Ashkelon_IA1
9.4 ITA_Ardea_Latini_IA
7.4 ITA_Proto-Villanovan
6.2 ITA_Boville_Ernica_Latini_IA
5.8 ARM_LBA
1.6 Morocco_Iberomaurusian

Target: Italian_Lazio
Distance: 0.8826% / 0.00882571
Aggregated
21.4 Anatolia_Isparta_EBA
21.0 ITA_Ardea_Latini_IA
15.4 DEU_MA
12.0 Levant_ISR_Ashkelon_LBA
9.0 DEU_Lech_MBA
6.8 Levant_ISR_Ashkelon_IA2
5.2 ITA_Etruscan
3.8 ITA_Boville_Ernica_Latini_IA
2.6 England_Roman_Near_Eastern_o
2.0 ITA_Proto-Villanovan
0.6 DEU_Lech_EBA
0.2 ARM_LBA

Target: Italian_Jew
Distance: 0.7098% / 0.00709773
Aggregated
18.2 ITA_Etruscan
18.0 Levant_ISR_Ashkelon_LBA
11.8 Levant_Canaanite_MBA
10.0 ITA_Boville_Ernica_Latini_IA
9.0 Anatolia_Isparta_EBA
9.0 ITA_Proto-Villanovan
6.2 Anatolia_Kaman-Kalehoyuk_MLBA
5.6 England_Roman_Near_Eastern_o
3.2 Morocco_Iberomaurusian
3.0 ARM_LBA
2.4 Anatolia_IA
1.4 Levant_ISR_Ashkelon_IA1
1.2 Levant_ISR_Ashkelon_IA2
1.0 DEU_Lech_MBA

Target: Italian_Campania
Distance: 0.5753% / 0.00575286
Aggregated
23.4 Anatolia_Isparta_EBA
11.6 DEU_MA
10.6 Anatolia_Kaman-Kalehoyuk_MLBA
10.4 ITA_Etruscan
8.6 Levant_ISR_Ashkelon_LBA
7.4 ITA_Ardea_Latini_IA
6.4 Levant_Canaanite_MBA
4.8 ITA_Boville_Ernica_Latini_IA
4.6 ITA_Proto-Villanovan
2.6 ARM_LBA
2.6 England_Roman_Near_Eastern_o
2.0 DEU_Lech_MBA
1.6 Anatolia_IA
1.2 Morocco_Iberomaurusian
1.2 Levant_ISR_Ashkelon_IA1
1.0 DEU_Lech_EBA

Target: Italian_Calabria
Distance: 0.8541% / 0.00854083
Aggregated
22.0 Anatolia_Kaman-Kalehoyuk_MLBA
15.6 ITA_Etruscan
15.0 Anatolia_Isparta_EBA
9.6 DEU_MA
8.2 England_Roman_Near_Eastern_o
8.0 Levant_ISR_Ashkelon_IA1
6.4 Anatolia_IA
4.2 ITA_Ardea_Latini_IA
2.4 ITA_Villanovan
2.2 GRC_Mycenaean
2.0 Levant_ISR_Ashkelon_IA2
1.6 Morocco_Iberomaurusian
1.6 ARM_LBA
1.2 ITA_Rome_Latini_IA

Target: Italian_Bergamo
Distance: 0.7375% / 0.00737462
Aggregated
24.8 ITA_Etruscan
20.0 DEU_MA
16.2 ITA_Boville_Ernica_Latini_IA
14.4 Anatolia_Isparta_EBA
9.8 ITA_Proto-Villanovan
7.4 ITA_Ardea_Latini_IA
7.0 ITA_Villanovan
0.2 Morocco_Iberomaurusian
0.2 GRC_Mycenaean

Target: Italian_Basilica
Distance: 0.8436% / 0.00843599
Aggregated
27.0 Anatolia_Isparta_EBA
16.8 DEU_MA
10.6 ITA_Proto-Villanovan
9.4 ITA_Ardea_Latini_IA
8.8 England_Roman_Near_Eastern_o
7.4 Levant_ISR_Ashkelon_IA1
6.8 Anatolia_Kaman-Kalehoyuk_MLBA
3.6 Anatolia_IA
3.0 ITA_Etruscan
1.8 ITA_Boville_Ernica_Latini_IA
1.6 Levant_Canaanite_MBA
1.4 Levant_ISR_Ashkelon_LBA
1.2 ARM_LBA
0.4 Levant_ISR_Ashkelon_IA2
0.2 Morocco_Iberomaurusian

Target: Italian_Apulia
Distance: 0.7983% / 0.00798274
Aggregated
23.4 Anatolia_Isparta_EBA
23.2 ITA_Proto-Villanovan
11.6 Levant_ISR_Ashkelon_IA1
9.2 ITA_Etruscan
9.0 DEU_MA
6.2 Levant_Canaanite_MBA
5.4 England_Roman_Near_Eastern_o
5.2 ITA_Boville_Ernica_Latini_IA
4.4 ARM_LBA
1.2 ITA_Ardea_Latini_IA
0.8 Anatolia_IA
0.2 Morocco_Iberomaurusian
0.2 Anatolia_Kaman-Kalehoyuk_MLBA

Target: Italian_Abruzzo
Distance: 0.5375% / 0.00537469
Aggregated
21.6 Anatolia_Isparta_EBA
18.2 ITA_Etruscan
9.6 DEU_MA
9.0 ITA_Proto-Villanovan
8.6 ITA_Ardea_Latini_IA
8.4 ARM_LBA
7.8 Levant_ISR_Ashkelon_LBA
6.4 Levant_Canaanite_MBA
3.6 ITA_Boville_Ernica_Latini_IA
3.0 DEU_Lech_MBA
2.8 England_Roman_Near_Eastern_o
0.4 Morocco_Iberomaurusian
0.4 DEU_Lech_EBA
0.2 Levant_ISR_Ashkelon_IA2

[Dick]

^ those are good fits

[Voskos]

Greek ancestry tends to be overestimated by several agendas.

Here's some more reliable data using dental remains from magna graecia.

Southern Italy during pre‐colonial times was characterized by low levels of mobility and marked differences in group means and variances. During post‐colonial times, mobility increased and there were no differences in group means and variances. About 18% of the people in post‐colonial times were of Greek ancestry and lived equally distributed across Greek colonies and indigenous villages. Nevertheless, the overall biological composition and variability of southern Italy remained relatively unchanged across pre‐ and post‐colonial periods.

Discussion

Our results support a scenario in which only few Greek colonists migrated to southern Italy and lived in smaller numbers alongside indigenous people in Greek colonies as well as in indigenous villages. Our results contradict a scenario in which large numbers of Greek invaders founded biologically isolated and substantially homogeneous colonial enclaves within conquered territories.

https://onlinelibrary.wiley.com/doi/...002/ajpa.23937

[Samnium]

^ those are good fits

Too much populations can lead to overfitting. I never use more than 6 populations for that reason

Envoyé de mon ALE-L21 en utilisant Tapatalk

[Creoda]

Too much populations can lead to overfitting. I never use more than 6 populations for that reason

Envoyé de mon ALE-L21 en utilisant Tapatalk

Too many similar populations is overfitting, but that's only a problem if you're taking each ancestral component literally.

[XenophobicPrussian]

Too much populations can lead to overfitting. I never use more than 6 populations for that reason

Envoyé de mon ALE-L21 en utilisant Tapatalk

Overfitting is not a bad thing. For all intents and purposes, consider, "ARM_LBA" and "Anatolia_EBA" the same population if you must for example, but not overfitting is worse because you miss crucial populations in certain instances.

[Samnium]

Overfitting is not a bad thing. For all intents and purposes, consider, "ARM_LBA" and "Anatolia_EBA" the same population if you must for example, but not overfitting is worse because you miss crucial populations in certain instances.

You can obtain very good distances with a 6-population model, I've made one that works very well in central and southern italian regions.

It's much more "elegant" and you limit the overfit.