Did ancient Greeks leave a genetic impact on West Anatolia? qpAdm and G25 analysis

Click this to see modern Greeks modelled with these samples: Attempting to improve the qpAdm models from the new study: A genetic history of the Balkans from Roman frontier to Slavic migrations

See also: Testing the Roman era sample from Marathon, Greece with qpAdm, f4 statistics and G25

Post by Davidski about the study: https://eurogenes.blogspot.com/2024/01/romans-and-slavs-in-balkans-olalde-et.html

Answer: Yes, they did. The Roman era West Anatolians 2000 years ago had around 28-40% Mycenaean Greek ancestry.

1) On both tools the Roman era West Anatolians are a mix of 28-40% Mycenaean + South East Anatolian + North Levantine + West-Central Anatolian. The qpAdm rotation rejects any model without Mycenaean.

2) When running Fst/f2 the Roman era West Anatolians are closer genetically to Mycenaeans, South East Anatolians, North Levantines than Iron age West Anatolians because they were mainly a direct mix of those first 3. But this part could be faulty due to problematic samples, it still does not disprove 1) or 3).

3) When rotating we can rank P-values. The model with Mycenaean gets higher P-value than Thracian. Therefore Mycenaean/Greek is the most likely source.

I took the reference/right populations as this study but expanded upon them. 
DNA samples from the Version v54.1.p1 Harvard dataset: https://reichdata.hms.harvard.edu/pub/datasets/amh_repo/curated_releases/

For context:
Turkey_Archaic_SubGeometric = Iron age West Anatolian from Mugla
Turkey_EBA_Isparta = Bronze age West Anatolia

Turkey_Alalakh_MLBA = Bronze age North Levant/South East Anatolia

Turkey_IA_Batman = Iron age South East Turkey

If you have ideas of how to improve the models then post them in the comments. I will attempt to improve the models even more in the future.

qpAdm:

Only the model with Mycenaean + South East Anatolian/North Levantine is plausible.
This is a rotation:

right = c('Mbuti.DG', 'Russia_Samara_EBA_Yamnaya', 'Russia_Karelia_HG', 'Serbia_IronGates_Mesolithic', 'Turkey_N', 'Iran_GanjDareh_N', 'Spain_IA', 'Greece_Minoan_Lassithi', 'Croatia_MLBA', 'Netherlands_EIA', 'Netherlands_MBA', 'Russia_IA_Ingria.SG', 'Latvia_BA', 'Turkey_EBA_II.SG', 'Israel_C', 'Armenia_EBA_KuraAraxes', 'Mongolia_North_N', 'Croatia_EIA', 'Albania_BA_IA', 'Russia_MLBA_Sintashta', 'Iran_C_SehGabi', 'Turkey_EBA_Isparta', 'Lebanon_MBA.SG', 'Turkey_Archaic_SubGeometric', 'Macedonia_Classical_Hellenistic', 'Bulgaria_EIA', 'Turkey_Alalakh_MLBA', 'Turkey_IA_Batman', 'Turkey_OldHittitePeriod.SG', 'Greece_BA_Mycenaean')

right = c('Mbuti.DG', 'Russia_Samara_EBA_Yamnaya', 'Russia_Karelia_HG', 'Serbia_IronGates_Mesolithic', 'Turkey_N', 'Iran_GanjDareh_N', 'Spain_IA', 'Greece_Minoan_Lassithi', 'Croatia_MLBA', 'Netherlands_EIA', 'Netherlands_MBA', 'Russia_IA_Ingria.SG', 'Latvia_BA', 'Turkey_EBA_II.SG', 'Israel_C', 'Armenia_EBA_KuraAraxes', 'Mongolia_North_N', 'Croatia_EIA', 'Albania_BA_IA', 'Russia_MLBA_Sintashta', 'Iran_C_SehGabi', 'Turkey_EBA_Isparta', 'Lebanon_MBA.SG', 'Turkey_Archaic_SubGeometric', 'Macedonia_Classical_Hellenistic', 'Bulgaria_EIA', 'Turkey_Alalakh_MLBA', 'Turkey_SoutheastByzantine_Mardin')

right = c('Mbuti.DG', 'Russia_Samara_EBA_Yamnaya', 'Russia_Karelia_HG', 'Serbia_IronGates_Mesolithic', 'Turkey_N', 'Iran_GanjDareh_N', 'Spain_IA', 'Greece_Minoan_Lassithi', 'Croatia_MLBA', 'Netherlands_EIA', 'Netherlands_MBA', 'Russia_IA_Ingria.SG', 'Latvia_BA', 'Turkey_EBA_II.SG', 'Israel_C', 'Armenia_EBA_KuraAraxes', 'Mongolia_North_N', 'Bulgaria_EIA', 'Croatia_EIA', 'Albania_BA_IA', 'Russia_MLBA_Sintashta', 'Iran_C_SehGabi', 'Turkey_EBA_Isparta', 'Lebanon_MBA.SG', 'Turkey_Archaic_SubGeometric', 'Turkey_OldHittitePeriod.SG', 'Greece_BA_Mycenaean', 'Turkey_Alalakh_MLBA')

Rotating between Mycenaean and samples from North Macedonia. Mycenaean is still the most likely source.

right = c('Mbuti.DG', 'Russia_Samara_EBA_Yamnaya', 'Russia_Karelia_HG', 'Serbia_IronGates_Mesolithic', 'Turkey_N', 'Iran_GanjDareh_N', 'Spain_IA', 'Greece_Minoan_Lassithi', 'Croatia_MLBA', 'Netherlands_EIA', 'Netherlands_MBA', 'Russia_IA_Ingria.SG', 'Latvia_BA', 'Turkey_EBA_II.SG', 'Israel_C', 'Armenia_EBA_KuraAraxes', 'Mongolia_North_N', 'Croatia_EIA', 'Albania_BA_IA', 'Russia_MLBA_Sintashta', 'Iran_C_SehGabi', 'Turkey_EBA_Isparta', 'Lebanon_MBA.SG', 'Turkey_Archaic_SubGeometric', 'Bulgaria_EIA', 'Macedonia_Classical_Hellenistic', 'Greece_BA_Mycenaean')

Assessing the performance of qpAdm: a statistical tool for studying population admixture

G25:

G25 models also show the exact same thing. Roman era West Anatolians and Iron age West Anatolians formed from different completely admixture events.

Source pops: https://pastebin.com/xycjBLAe

Source pops: https://pastebin.com/qJzcJZLg

Even when not using an Armenian-Upper Mesopotamian like source to force West Anatolian to inflate, the minimum Mycenaean is still 28%. 

Extra qpAdm runs:

right = c('Mbuti.DG', 'Russia_Samara_EBA_Yamnaya', 'Russia_Karelia_HG', 'Serbia_IronGates_Mesolithic', 'Turkey_N', 'Iran_GanjDareh_N', 'Spain_IA', 'Greece_Minoan_Lassithi', 'Croatia_MLBA', 'Netherlands_EIA', 'Netherlands_MBA', 'Russia_IA_Ingria.SG', 'Latvia_BA', 'Turkey_EBA_II.SG', 'Israel_C', 'Armenia_EBA_KuraAraxes', 'Mongolia_North_N', 'Croatia_EIA', 'Albania_BA_IA', 'Russia_MLBA_Sintashta', 'Iran_C_SehGabi', 'Turkey_EBA_Isparta', 'Lebanon_MBA.SG', 'Turkey_Archaic_SubGeometric', 'Macedonia_Classical_Hellenistic', 'Bulgaria_EIA', 'Turkey_Alalakh_MLBA', 'Turkey_AssyrianColonyPeriod.SG', 'Turkey_OldHittitePeriod.SG')

Fst population differentiation (like genetic distance):

They are closer to Mycenaeans and South East Anatolians/North Levantines than Carians, probably because they have mostly direct admix from those 2 instead of Carians or something is wrong with the Carian samples which is a common problem for F2 and Fst. Even if the Fst results are problematic, the qpAdm models above shouldn't have a problem with this.

I will post F4 too another time.