Non-West Eurasian ancestry in the Southern Arc
We first examined the ADMIXTURE output(1)to identify a set of individuals with at least
10% of their ancestry not from the four West Eurasian-related components which dominate the
ancestry of Southern Arc individuals. Low levels of such admixture might be due to noise for the
ancient DNA samples, and our 10% threshold is intended to identify a smaller subset for further
investigation. We also excluded low coverage (marked _lc ) samples.
A total of 47 Southern Arc individuals were identified according to this criterion, and we
show their ADMIXTURE proportions inFig. S4. The African-maximized “black” component is
found in Levantine individuals as early as the Natufians and should thus not be interpreted as
evidence of recent African influence in West Eurasia. A likely explanation is the partial
derivation of the Natufians from Paleolithic Iberomaurusian(48)North African-related ancestors
as suggested in(49)Indeed, the average proportion of this component in all Natufian individuals
(including those for which it is less than the detection threshold of 10%) is 9.1%, while in
Taforalt from Morocco it is 41.4%, thus suggesting ~22% of North African influence, similar to
the ~27% inferred using an admixture graph framework in(49)
The remaining outliers are driven primarilyby eastern non-African ancestry, the main
component of which is the “dark blue” component maximized in present-day East Asians. This
includes samples from Moldova and Romania of likely “steppe nomad” derivation from the
medieval period, as well as an undated sample from Moldova (I20071) of putative Middle
Bronze Age, another one (I20086) of putative Eneolithic to Early Bronze Age, and an undated
sample fromKalehöyük(50)of the Iron Age (MA2196).
Individuals fromShahr-i Sōkhtafrom southeastern Iran also show eastern non-African
admixture, and are also notable for their possession of some “gray” (Australasian-maximized)
component. This component is probably related to neighboring South Asian populations, as it is
also found in the ADMIXTURE analysis in virtually all ancient samples from Pakistan.(23)
We highlight two other sets of outliers with non-West Eurasian ancestry.
Three outliers from Albania are also outliers of the 5-way admixture model(1). Their East
Asian-related ancestry suggest that they may have Central Asian ancestry (which could be
consistent with them being of Turkic Central Asian derivation, at least in part, given their post-
medieval time frame). The city ofKorҫa (just 1.2 km from the village of Barҫi) was already
invaded by the Ottomans in the 15th century CEand thus the individuals sampled there could
very well be descendants of the Ottomans.
A set of samples from ÇapalıbağatMuğla(1300-1650CE; average of C14 dates of
1480CE)are also likely to be of Turkic ancestry as they postdate the establishment of the Seljuq
dynasty in Anatolia and also have substantial East Asian-related ancestry which could have been
mediated via Central Asian ancestors. We dated the admixture timing of this population using
DATES(51)using the Anatolian Byzantine population as one source and a diverse set of samples
from Central Asia between 500-1500 years ago as the second source. The obtained date is
12.2±1.4 generations or, assuming a generation length of 28 years,(51)a 95% confidence
interval of ~267-418 years (mean 342 years) prior to their time which would be consistent with
either the admixture taking place prior to or after the arrival of Seljuqs in Anatolia. We also
estimated the admixture date for present-day Turkish people from Anatolia genotyped on the
Human Origins array(34)using the same sources, finding it to be 30.6±1.9 generations or ~755-
958 years (mean 856 years). The estimated admixture time using the Çapalıbağand present-day
Turkish data coincides with the early centuries of the 2ndmillennium CE, roughly the period in
which the Seljuqs and Ottomans gained control of Anatolia from the Romans (Byzantines)
before the final capture of the Imperial capital city of Constantinople in 1453CE.
The problem of the sources and mode of arrival of Turkic speakers in Anatolia is complex
and is beyond the scope of this paper, but we hope that the data presented here would be useful
to future studies of the topic, as they establish the “Roman-Byzantine” baseline population across
large parts of Anatolia on which the Turkic population influence could be studied in the future,
as well as the first known sample ancient DNA population of plausibly Turkic descendants from
Anatolia.
Mycenaean-like ancestry in southeastern Europe and beyond
A recent study identified the presence of individuals resembling the Bronze Age Mycenaean
population of Greece(4)in a time transect of the Spanish site of Empúries(Greek
Ἐμπόριον).(20)The approach adopted there was to identify the outliers in the site and find that
they were most similar to the Mycenaean population.
Here we try to perform the converse operation: beginning with the Mycenaean Greeks (to
the sampling of which our study adds many new individuals), can we identify other individuals
in our total dataset that could be drawn from the same population. Such an operation would
hopefully identify the known individuals fromEmpúries but might also disclose other such
individuals in either the new data of our paper or the literature at large.
Our approach to data mine the dataset is as follows. We estimate the Mahalanobis distance
of each sample to the total Mycenaean population and order samples according to this distance
using a p-value cutoff of 0.01 to identify samples that are not significantly different than the
Mycenaeans. To ensure that we identify samples that are genuinely within the Mycenaean
genetic variation (to the limits of our ability), we perform this operation in three types of data:
ADMIXTURE coefficients, F4admix coefficients (excluding the SRB_Iron_Gates_HG
component for numerical stability as this is zero in most individuals), and the first 10 principal
components of the West Eurasian PCA(1).
In Table S1we list the individuals that are indistinguishable from Mycenaeans according to
our procedure in all three tests. This does indeed identify two individuals from Empúries(I8215
andI8208) as highly similar to the Mycenaean population. The strong similarity of these two
Classical and Hellenistic individuals (4th-3rdcentury BCE) to the Mycenaeans of a 1,000 years
earlier has interesting implications beyond their local Iberian setting and underscores the
importance of “Big Picture” studies to produce a framework through which the analysis of local
populations can be better interpreted:
The western Mediterranean Greek colonists in this site in Spain were derived from 6thc.
BCE Massaliotes(AncientΜασσαλία,modern Marseilles in France) who themselves were
derived from Phocaeans (AncientΦώκαια,modernFoçain Turkey) who themselves were
colonists from Phokis (Φωκίς) in mainland Greece with Ionian kings who traced descent from
Codrus (and thus from Attica).1Whatever the origin of the specific individuals unearthedat
Empúries, their genetic similarity to the Mycenaean population suggests that no major admixture
had occurred in their ancestry from the Bronze Age to their own time, e.g., in either Asia Minor
(during the founding of Phocaea) or western Europe, which would have introduced ancestry
more prevalent in either region (e.g., CHG or WHG) compared to mainland Greece. We do not
have all the links in this long chain of transmission of the Aegean ancestry into the western
Mediterrenan, yet we do have samples of Myceanean age from the site of Kastrouli near Delphi
in Phokis, two Archaic sample from Phokis (I17962;773-544 calBCE, andI17959; 800-500
BCE) closer to the time of the foundation of Phocaea, and Mycenaean samples from Attica and
can thus confirm that the population of the putative ancestors of the Western Mediterranean
Greeks were indeed similar to that of the Mycenaeans in general on the basis of I17962 which
appears to by Mycenaean-like according to our procedure (I17959 is ranked #76 and is not listed
in the Table, but we do not ascribe any importance to this as this is a lower coverage sample with
only ~15k SNPs covered).
From Greece itself there is another post-Mycenaean (Proto-Geometric/Early Iron Age)
individual (I19368) from the vicinity of the Palace of Nestor at Pylos which is also confirmed by
our procedure to be Mycenaean-like and thus similar to the people that lived in Greece a few
centuries earlier across the LBA to Iron Age transition.
Another sample which resembles Mycenaeans genetically isASH068an Iron Age
“Philistine” from the Levant, also identified as resembling the Late Bronze Age population of
southern Greece in the original publication.(21)
Two other samples from the literature were identified:
SZ19 is a Langobard-era sample fromSzólád, Hungary from the 5th-6thc. CE. SZ19 was a
young female of 17-25 years old who was also a genetic outlier in the group of individuals
buried there, had a distinct burial type, and also had a “stylistically distinct (possibly
Roman)”(52)artifact associated with her burial. Quite possibly she was related to the population
of the Aegean and the southern Balkans given the similarity to Mycenaeans detected here.
I20257is an ancient adolescent female fromDeğirmendereinMuğlafrom the Aegean
region of Turkey (750-480 BCE).Her similarity to the Mycenaean population is not surprising
given the proximity to Greece and her time postdating the colonization of the coast of Anatolia.
Two other samples from the same site are more distant (I20229andI20233). Thus only 3 of 10
samples from this site are similar to Mycenaeans. We cannot speak of a general similarity here,
but rather that the “Carian” population atDeğirmendereincluded Mycenaean-like individuals
while being generally distinct. Thus, the previously plausible theory that culturally Greek people
in the classical period and earlier did not mix with locals—suggested by the patterns at
Empúries—is not supported by the data.
I5737, a Middle Bronze Age sample from Yassıtepe(Izmir, AncientΣμύρνα/ Smyrna) is
also identified,predating the Mycenaean samples (2033-1920 calBCE). Its EHG ancestry is
2.9±2.6% so we cannot be certain that it was present here as in most Mycenaean samples, but its
overall genetic makeup appears to be similar. This individual also had Y-chromosome I-P58
linking him to southeastern Europe. We cannot speak of the population in general here, but this
sample provides the earliest direct evidence of human migration from the Balkans to Anatolia, a
pattern that recurs more than a millennium later at Değirmendere and provides evidence of a
long history of genetic interchange across the Aegean. Two Roman/Byzantine samples from the
Basilica at Nicaea are the remaining Mycenaean-like samples from Anatolia (I8366andI8368).
Overall, however, our procedure only identified a very small number of individuals from
Anatolia as being genetically similar to Mycenaeans, which is notable given the colonization of
Anatolia by Ionian Greeks and the later incorporation of it to the Hellenistic Kingdoms and
Roman Empire which used koineGreek as its language in the east. Possible explanations for this
are either that our sampling bias—that our dataset has few samples derived from contexts
specific to ancient colonists—or that the colonists of Anatolia intermarried with the local
population as suggested in ancient times by Herodotus for Ionian colonists from Athens who
intermarried with local Carian women (again, different from the pattern seen at Empúrieswhere
many in the culturally Greek population retained their genetic affinity to Greece despite a long
history of serial colonization.2The same could be true for individuals sampled from Samsun
(AncientἈμισός/ Amisos) and Bodrum (AncientἉλικαρνασσός/Halikarnassos)which were
certainly places of ancient settlement and were the colonists may have intermarried with locals
which would have modified their ancestry in a more “eastern” direction.
To the west of Greece, 1 sample from Italy, a Punic sample from Sardinia (MSR002) is
identified as Mycenaean-like.(53)We note that the samples from Italy do not include Sicily and
Southern Italy at the time or postdating Greek colonization, but they do include a large set of
samples from Imperial Rome which we infer to be mostly of Anatolian rather than Aegean or
southeastern European origin.
Many samples from southeastern Europe north of Greece are identified as being genetically
similar to the Mycenaean population of southern Greece.
From North Macedonia, a sample (I7233;897-811calBCE) is quite early in time. We
cannot speak of a general Mycenaean-like population here as the remaining samples from the 1st
millennium BCE do not bear this close resemblance to the Mycenaean population.
By far, the greatest number of Mycenaean-like individuals in our dataset outside Greece
itself is found in neighboring Bulgaria where 10 such samples (from several sites) areidentified.
A Neolithic outlier individual from Krepost(7)is the earliest. This individual has no EHG
ancestry according to our estimation, but is a mixture of mainly Anatolian Neolithic and CHG-
related ancestry. Thus, it may somewhat resemble Mycenaeans, but it would be difficult to speak
of continuity since its 6thmillennium BCE date on its basis, especially as this pattern is not
supported by other Neolithic/Chalcolithic era samples from Bulgaria or Greece, someof which
post-date the Krepost individual.
More convincing are several 1stmillennium BCE individuals from Rozovo (I19500),
Diamandievo (I19481),Dzhulyunitsa(I5769), and Kapitan Andreevo (about half of the samples
here). As these sites are inland, they should not be attributed to maritime contacts and the
foundation of colonies in the Thracian coast by Greek settlers, but may better suggest a similarity
of population in the southern Balkans with the Aegean.
Future studies of intermediate regions between southern Greece, North Macedonia, and
Bulgaria will be important in further mapping the extent of the Mycenaean-like population and
its relationships to those further north in the Balkans.
Supplementary Materials for
A genetic probe into the ancient and medieval history of Southern Europe and West
Asia
Iosif Lazaridis, Songül Alpaslan-Roodenberget al. 2022
file:///C:/Users/18509/Downloads/science.abq0755_sm.pdf
We first examined the ADMIXTURE output(1)to identify a set of individuals with at least
10% of their ancestry not from the four West Eurasian-related components which dominate the
ancestry of Southern Arc individuals. Low levels of such admixture might be due to noise for the
ancient DNA samples, and our 10% threshold is intended to identify a smaller subset for further
investigation. We also excluded low coverage (marked _lc ) samples.
A total of 47 Southern Arc individuals were identified according to this criterion, and we
show their ADMIXTURE proportions inFig. S4. The African-maximized “black” component is
found in Levantine individuals as early as the Natufians and should thus not be interpreted as
evidence of recent African influence in West Eurasia. A likely explanation is the partial
derivation of the Natufians from Paleolithic Iberomaurusian(48)North African-related ancestors
as suggested in(49)Indeed, the average proportion of this component in all Natufian individuals
(including those for which it is less than the detection threshold of 10%) is 9.1%, while in
Taforalt from Morocco it is 41.4%, thus suggesting ~22% of North African influence, similar to
the ~27% inferred using an admixture graph framework in(49)
The remaining outliers are driven primarilyby eastern non-African ancestry, the main
component of which is the “dark blue” component maximized in present-day East Asians. This
includes samples from Moldova and Romania of likely “steppe nomad” derivation from the
medieval period, as well as an undated sample from Moldova (I20071) of putative Middle
Bronze Age, another one (I20086) of putative Eneolithic to Early Bronze Age, and an undated
sample fromKalehöyük(50)of the Iron Age (MA2196).
Individuals fromShahr-i Sōkhtafrom southeastern Iran also show eastern non-African
admixture, and are also notable for their possession of some “gray” (Australasian-maximized)
component. This component is probably related to neighboring South Asian populations, as it is
also found in the ADMIXTURE analysis in virtually all ancient samples from Pakistan.(23)
We highlight two other sets of outliers with non-West Eurasian ancestry.
Three outliers from Albania are also outliers of the 5-way admixture model(1). Their East
Asian-related ancestry suggest that they may have Central Asian ancestry (which could be
consistent with them being of Turkic Central Asian derivation, at least in part, given their post-
medieval time frame). The city ofKorҫa (just 1.2 km from the village of Barҫi) was already
invaded by the Ottomans in the 15th century CEand thus the individuals sampled there could
very well be descendants of the Ottomans.
A set of samples from ÇapalıbağatMuğla(1300-1650CE; average of C14 dates of
1480CE)are also likely to be of Turkic ancestry as they postdate the establishment of the Seljuq
dynasty in Anatolia and also have substantial East Asian-related ancestry which could have been
mediated via Central Asian ancestors. We dated the admixture timing of this population using
DATES(51)using the Anatolian Byzantine population as one source and a diverse set of samples
from Central Asia between 500-1500 years ago as the second source. The obtained date is
12.2±1.4 generations or, assuming a generation length of 28 years,(51)a 95% confidence
interval of ~267-418 years (mean 342 years) prior to their time which would be consistent with
either the admixture taking place prior to or after the arrival of Seljuqs in Anatolia. We also
estimated the admixture date for present-day Turkish people from Anatolia genotyped on the
Human Origins array(34)using the same sources, finding it to be 30.6±1.9 generations or ~755-
958 years (mean 856 years). The estimated admixture time using the Çapalıbağand present-day
Turkish data coincides with the early centuries of the 2ndmillennium CE, roughly the period in
which the Seljuqs and Ottomans gained control of Anatolia from the Romans (Byzantines)
before the final capture of the Imperial capital city of Constantinople in 1453CE.
The problem of the sources and mode of arrival of Turkic speakers in Anatolia is complex
and is beyond the scope of this paper, but we hope that the data presented here would be useful
to future studies of the topic, as they establish the “Roman-Byzantine” baseline population across
large parts of Anatolia on which the Turkic population influence could be studied in the future,
as well as the first known sample ancient DNA population of plausibly Turkic descendants from
Anatolia.
Mycenaean-like ancestry in southeastern Europe and beyond
A recent study identified the presence of individuals resembling the Bronze Age Mycenaean
population of Greece(4)in a time transect of the Spanish site of Empúries(Greek
Ἐμπόριον).(20)The approach adopted there was to identify the outliers in the site and find that
they were most similar to the Mycenaean population.
Here we try to perform the converse operation: beginning with the Mycenaean Greeks (to
the sampling of which our study adds many new individuals), can we identify other individuals
in our total dataset that could be drawn from the same population. Such an operation would
hopefully identify the known individuals fromEmpúries but might also disclose other such
individuals in either the new data of our paper or the literature at large.
Our approach to data mine the dataset is as follows. We estimate the Mahalanobis distance
of each sample to the total Mycenaean population and order samples according to this distance
using a p-value cutoff of 0.01 to identify samples that are not significantly different than the
Mycenaeans. To ensure that we identify samples that are genuinely within the Mycenaean
genetic variation (to the limits of our ability), we perform this operation in three types of data:
ADMIXTURE coefficients, F4admix coefficients (excluding the SRB_Iron_Gates_HG
component for numerical stability as this is zero in most individuals), and the first 10 principal
components of the West Eurasian PCA(1).
In Table S1we list the individuals that are indistinguishable from Mycenaeans according to
our procedure in all three tests. This does indeed identify two individuals from Empúries(I8215
andI8208) as highly similar to the Mycenaean population. The strong similarity of these two
Classical and Hellenistic individuals (4th-3rdcentury BCE) to the Mycenaeans of a 1,000 years
earlier has interesting implications beyond their local Iberian setting and underscores the
importance of “Big Picture” studies to produce a framework through which the analysis of local
populations can be better interpreted:
The western Mediterranean Greek colonists in this site in Spain were derived from 6thc.
BCE Massaliotes(AncientΜασσαλία,modern Marseilles in France) who themselves were
derived from Phocaeans (AncientΦώκαια,modernFoçain Turkey) who themselves were
colonists from Phokis (Φωκίς) in mainland Greece with Ionian kings who traced descent from
Codrus (and thus from Attica).1Whatever the origin of the specific individuals unearthedat
Empúries, their genetic similarity to the Mycenaean population suggests that no major admixture
had occurred in their ancestry from the Bronze Age to their own time, e.g., in either Asia Minor
(during the founding of Phocaea) or western Europe, which would have introduced ancestry
more prevalent in either region (e.g., CHG or WHG) compared to mainland Greece. We do not
have all the links in this long chain of transmission of the Aegean ancestry into the western
Mediterrenan, yet we do have samples of Myceanean age from the site of Kastrouli near Delphi
in Phokis, two Archaic sample from Phokis (I17962;773-544 calBCE, andI17959; 800-500
BCE) closer to the time of the foundation of Phocaea, and Mycenaean samples from Attica and
can thus confirm that the population of the putative ancestors of the Western Mediterranean
Greeks were indeed similar to that of the Mycenaeans in general on the basis of I17962 which
appears to by Mycenaean-like according to our procedure (I17959 is ranked #76 and is not listed
in the Table, but we do not ascribe any importance to this as this is a lower coverage sample with
only ~15k SNPs covered).
From Greece itself there is another post-Mycenaean (Proto-Geometric/Early Iron Age)
individual (I19368) from the vicinity of the Palace of Nestor at Pylos which is also confirmed by
our procedure to be Mycenaean-like and thus similar to the people that lived in Greece a few
centuries earlier across the LBA to Iron Age transition.
Another sample which resembles Mycenaeans genetically isASH068an Iron Age
“Philistine” from the Levant, also identified as resembling the Late Bronze Age population of
southern Greece in the original publication.(21)
Two other samples from the literature were identified:
SZ19 is a Langobard-era sample fromSzólád, Hungary from the 5th-6thc. CE. SZ19 was a
young female of 17-25 years old who was also a genetic outlier in the group of individuals
buried there, had a distinct burial type, and also had a “stylistically distinct (possibly
Roman)”(52)artifact associated with her burial. Quite possibly she was related to the population
of the Aegean and the southern Balkans given the similarity to Mycenaeans detected here.
I20257is an ancient adolescent female fromDeğirmendereinMuğlafrom the Aegean
region of Turkey (750-480 BCE).Her similarity to the Mycenaean population is not surprising
given the proximity to Greece and her time postdating the colonization of the coast of Anatolia.
Two other samples from the same site are more distant (I20229andI20233). Thus only 3 of 10
samples from this site are similar to Mycenaeans. We cannot speak of a general similarity here,
but rather that the “Carian” population atDeğirmendereincluded Mycenaean-like individuals
while being generally distinct. Thus, the previously plausible theory that culturally Greek people
in the classical period and earlier did not mix with locals—suggested by the patterns at
Empúries—is not supported by the data.
I5737, a Middle Bronze Age sample from Yassıtepe(Izmir, AncientΣμύρνα/ Smyrna) is
also identified,predating the Mycenaean samples (2033-1920 calBCE). Its EHG ancestry is
2.9±2.6% so we cannot be certain that it was present here as in most Mycenaean samples, but its
overall genetic makeup appears to be similar. This individual also had Y-chromosome I-P58
linking him to southeastern Europe. We cannot speak of the population in general here, but this
sample provides the earliest direct evidence of human migration from the Balkans to Anatolia, a
pattern that recurs more than a millennium later at Değirmendere and provides evidence of a
long history of genetic interchange across the Aegean. Two Roman/Byzantine samples from the
Basilica at Nicaea are the remaining Mycenaean-like samples from Anatolia (I8366andI8368).
Overall, however, our procedure only identified a very small number of individuals from
Anatolia as being genetically similar to Mycenaeans, which is notable given the colonization of
Anatolia by Ionian Greeks and the later incorporation of it to the Hellenistic Kingdoms and
Roman Empire which used koineGreek as its language in the east. Possible explanations for this
are either that our sampling bias—that our dataset has few samples derived from contexts
specific to ancient colonists—or that the colonists of Anatolia intermarried with the local
population as suggested in ancient times by Herodotus for Ionian colonists from Athens who
intermarried with local Carian women (again, different from the pattern seen at Empúrieswhere
many in the culturally Greek population retained their genetic affinity to Greece despite a long
history of serial colonization.2The same could be true for individuals sampled from Samsun
(AncientἈμισός/ Amisos) and Bodrum (AncientἉλικαρνασσός/Halikarnassos)which were
certainly places of ancient settlement and were the colonists may have intermarried with locals
which would have modified their ancestry in a more “eastern” direction.
To the west of Greece, 1 sample from Italy, a Punic sample from Sardinia (MSR002) is
identified as Mycenaean-like.(53)We note that the samples from Italy do not include Sicily and
Southern Italy at the time or postdating Greek colonization, but they do include a large set of
samples from Imperial Rome which we infer to be mostly of Anatolian rather than Aegean or
southeastern European origin.
Many samples from southeastern Europe north of Greece are identified as being genetically
similar to the Mycenaean population of southern Greece.
From North Macedonia, a sample (I7233;897-811calBCE) is quite early in time. We
cannot speak of a general Mycenaean-like population here as the remaining samples from the 1st
millennium BCE do not bear this close resemblance to the Mycenaean population.
By far, the greatest number of Mycenaean-like individuals in our dataset outside Greece
itself is found in neighboring Bulgaria where 10 such samples (from several sites) areidentified.
A Neolithic outlier individual from Krepost(7)is the earliest. This individual has no EHG
ancestry according to our estimation, but is a mixture of mainly Anatolian Neolithic and CHG-
related ancestry. Thus, it may somewhat resemble Mycenaeans, but it would be difficult to speak
of continuity since its 6thmillennium BCE date on its basis, especially as this pattern is not
supported by other Neolithic/Chalcolithic era samples from Bulgaria or Greece, someof which
post-date the Krepost individual.
More convincing are several 1stmillennium BCE individuals from Rozovo (I19500),
Diamandievo (I19481),Dzhulyunitsa(I5769), and Kapitan Andreevo (about half of the samples
here). As these sites are inland, they should not be attributed to maritime contacts and the
foundation of colonies in the Thracian coast by Greek settlers, but may better suggest a similarity
of population in the southern Balkans with the Aegean.
Future studies of intermediate regions between southern Greece, North Macedonia, and
Bulgaria will be important in further mapping the extent of the Mycenaean-like population and
its relationships to those further north in the Balkans.
Supplementary Materials for
A genetic probe into the ancient and medieval history of Southern Europe and West
Asia
Iosif Lazaridis, Songül Alpaslan-Roodenberget al. 2022
file:///C:/Users/18509/Downloads/science.abq0755_sm.pdf