Archaeogenetic Data Mining Supports a Uralic–Minoan Homeland in the Danube Basin ^†

Abstract

Four types of archaeogenetic data mining are used to investigate the origin of the Minoans and the Uralic peoples: (1) six SNP mutations related to eye, hair, and skin phenotypes; (2) whole-genome admixture analysis using the G25 system; (3) an analysis of the history of the U5 mitochondrial DNA haplogroup; and (4) an analysis of the origin of each currently known Minoan mitochondrial and y-DNA haplotypes. The uniform result of these analyses is that the Minoans and the Uralic peoples had a common homeland in the lower and middle Danube Basin, as well as the Black Sea coastal regions. This new result helps to reconcile archaeogenetics with linguistics, which have shown that the Minoan language belongs to the Uralic language family.

1. Introduction

Archaeogenetic and linguistic studies give contradictory results regarding the origins of the Minoan civilization, which flourished on the island of Crete in the Bronze Age. An influential whole-genome archaeogenetic study by Lazaridis et al. [1] concluded that

“Minoans and Mycenaeans were genetically similar, having at least three-quarters of their ancestry from the first Neolithic farmers of western Anatolia and the Aegean, and most of the remainder from ancient populations related to those of the Caucasus and Iran.”

In contrast, recent linguistic studies by Revesz [2,3,4] have indicated that the Minoan language belongs to the Uralic language family, which had a homeland near the Ural Mountains [5], the Northern Black Sea region [6], or the Carpathian Basin [7]—or more generally, the Danube Basin [8]. According to the traditional view, the Uralic language family originated about 7000 to 10,000 years ago [9]. It contains both Finno-Ugric and Samoyedic languages, and the Finno-Ugric languages have two main branches: the Finno-Permic branch—which includes Finnish, Estonian, Saami, and other languages—and the Ugric branch—which includes Hungarian, Khanty, and Mansi [10].

Revesz [2,3,4] identified Minoan as an extinct member of the Ugric branch based on translating thirty-one Minoan inscriptions as Proto-Ugric language documents. This linguistic classification was strengthened recently [11] (pp. 208–212) by showing regular sound changes between Pre-Greek origin Greek words identified by Beekes [12] and Proto-Uralic, Proto-Finno-Ugric, and Proto-Ugric words reconstructed by Rédei [13]. The overwhelming number of Pre-Greek words shows a uniformity that implies borrowings from a single source [14] (p. 45). Since the Minoan culture preceded the Greek-speaking Mycenaean culture on the islands of Crete and Santorini [15], the Pre-Greek words are likely to be borrowings from the Minoan language. Furthermore, demonstrating regular sound changes is the primary way to prove linguistic relationships among languages and is also used in Indo-European linguistics [16]. Bernal [17], Best [18], Campbell-Dunn [19], Gordon [20], Kvashilava [21], La Marle [22], and other authors who proposed a different linguistic affiliation of the Minoan language did not show regular sound changes.

The aim of the present paper is to reconcile the archaeogenetic and linguistic data and to show that they are compatible with a Danube Basin homeland of the Uralic languages. The Danube Basin includes the Danube Delta area from which the Minoans could have sailed south to the Aegean Sea via the Bosporus strait, while the rest of the Uralic language speakers could have migrated eastward along the Northern Black Sea coast and then northward along the major rivers as described by Wiik [6]. Figure 1 shows the hypothetical dispersal of the Uralic language family based on Krantz [7] and extended by an Ugric-Minoan link by Revesz [8]. Hence, the primary focus of the reconciliation proposed in this paper is to show that the archaeogenetic data support a Uralic-Minoan homeland in the Danube Basin.

This rest of this paper is organized as follows. Section 2 discusses SNP mutations related to eye, hair, and skin phenotypes. Section 3 discusses whole-genome G25 admixture analysis. Section 4 presents an analysis of U5 mtDNA haplotypes. Section 5 presents an analysis of the origin of all Minoan mtDNA and y-DNA haplotypes. Section 6 summarizes the results of the analyses given in Section 2, Section 3 and Section 4 and provides a discussion of the results. Finally, Section 7 gives some conclusions and directions for further work.

2. Method and Experiment 1: SNP Mutations Related to Eye, Hair and Skin Phenotypes

2.1. Method of Analyzing Archaeogenetic Phenotype Data

Human eye, hair, and skin phenotypes are genetically determined by various alleles. The lighter eye, hair, and skin phenotypes have some selective advantages at higher latitudes. Hence, they are spread widely among Eurasian populations.

Table 1. Genetic loci associated with lighter eye, hair, or skin color.

Gene	Loci	Allele Mutation	Phenotype
HERC2	rs12913832	A > G	Blue Eye
SLC24A4	rs2402130	G > A	Light Hair
SLC24A5	rs1426654	G > A	Light Skin
SLC45A2	rs28777	C > A	Low Melanin
SLC45A2	rs16891982	C > G	Light Skin
TYR	rs1042602	C > A	Light Skin

shows some data regarding six alleles that affect eye, hair, or skin pigmentation.

Table 2. Archaeological cultures with their periods based on the estimates or sample dates in the references.

Name	Abbreviation	Period	References
Caucasian Hunter-Gatherers	CHG	15,000–8000 BP	[23]
Eastern-European Hunter-Gatherers	EHG	10,000–7000 BP	[23,24]
Western-European Hunter-Gatherers	WHG	15,000~5000 BP	[23]
Lower-Danube Mesolithic Hunter-Gatherers	L_Danube_ME	9075–8435 BP	[25]
Fertile Crescent Neolithic Farmers	FertileC_NE	8300–7800 BP	[23,25,26,27]
Aegean Early Neolithic Farmers	Aegean_NE	8438–8030 BP	[28]
Körös Neolithic	Körös_NE	7800–7300 BP	[29]
Hungarian Middle Neolithic Farmers	Hungary_MN	7310–6950 BP	[29]
Hungarian Bronze Age	Hungary_BA	3900–3450 BP	[30]
Minoan Civilization (Early and Middle)	Minoan	5100–3450 BP	[1,31]
Mycenaean Civilization	Mycenaean	3750–3050 BP	[1,31]

lists the eleven archaeological cultures that we considered in this paper. The first nine archaeological cultures preceded the Minoan civilization. Hence, these nine archaeological cultures were considered possible ancestors of the Minoan civilization. The Mycenaean civilization largely followed the Minoan civilization and was included as a comparison with the Minoan civilization. The comparison would reveal whether the Minoans and Mycenaeans had different ancestors.

After collecting loci mutation and allele data from samples from these archaeological cultures, we computed the percentage of the various alleles. Then, we computed the root mean square error (RMSE) between every pair of archaeological cultures x and y, where x ≠ y using the following formula:

$$RMSE\left(x,y\right)=\sqrt{{\displaystyle \frac{{\sum }_{k=1}^{k=n}{{(p}_{x,k}-p_{y,k})}^2}{n}}}$$

(1)

where n is the number of genetic loci considered, and p_x,k is the percentage of the kth allele associated with lighter eye, hair, or skin pigmentation among the samples from x. A lower RMSE value indicates a greater overall genotypic similarity between the populations of two cultures.

2.2. Experiment with Archaeogenetic Phenotype Data

Table 3. Six eye, hair, skin alleles found in the samples from eleven ancient cultures.

Culture	Sample	Location	Ref.	rs12913832	rs2402130	rs1426654	rs28777	rs16891982	rs1042602
CHG	Kotias	Kotias, Georgia	[23]	AA	AA	AA	CC	CC	CC
CHG	Satsurblia	Satsurblia, Georgia	[23]	AG	AA	AA	CC	CC	CC
CHG				0.25 G	1.00 A	1.00 A	0.00 A	0.00 G	0.00 A
EHG	Ukr_HG1	Vasil’evka, Ukraine	[24]	AA	-	AA	-	GG	-
EHG	SVP44	Samara Oblast, Russia	[23]	GG	AA	AA	-	GG	CC
EHG	UzOO77	Yuzhnyy Oleniy, Russia	[23]	AA	AA	AA	AC	CG	CC
EHG				0.33 G	1.00 A	1.00 A	0.50 A	0.83 G	0.00 A
WHG	Braña	Braña, Spain	[23]	GG	AG	GG	CC	CC	CC
WHG	Loschbour	Loschbour, Luxemburg	[23]	GG	AG	GG	CA	CC	CC
WHG	Bichon	Bichon, France	[23]	AG	AA	GG	CC	CC	CC
WHG	Villabruna	Villabruna, Italy	[23]	GG	AG	GG	CC	CC	CC
WHG				0.88 G	0.63 A	0.00 A	0.13 A	0.00 G	0.00 A
L_Danube_ME	SC1	Schela Cladovei, Romania	[25]	AG	AA	GG	CA	CC	CC
L_Danube_ME	SC2	Schela Cladovei, Romania	[25]	AA	AA	GG	AA	CC	CC
L_Danube_ME	OC1	Ostrovul Corbului, Romania	[25]	AG	GA	GG	CC	CC	CC
L_Danube_ME				0.33 G	0.83 A	0.00 A	0.50 A	0.00 G	0.00 A
FertileC_NE	WC1	Wezmeh cave, Iran	[32]	AG	AA	GA	CC	CC	CC
FertileC_NE	AH1	Tepe Abdul Hosein, Iran	[32]	-	AA	GA	-	-	CC
FertileC_NE	AH2	Tepe Abdul Hosein, Iran	[32]	AA	AA	GG	CC	-	-
FertileC_NE	AH4	Tepe Abdul Hosein, Iran	[32]	-	AA	AA	-	-	-
FertileC_NE	GD13a	Ganj Dareh, Iran	[26]	AA	GG	AA	CC	CC	CC
FertileC_NE	Bon004	Boncuklu, Turkey	[27]	AA	-	AA	-	CC	CC
FertileC_NE	Bon014	Boncuklu, Turkey	[27]	AG	-	AA	-	CC	-
FertileC_NE	Bon001	Boncuklu, Turkey	[27]	GG	-	AA	-	CC	CC
FertileC_NE	Bon002	Boncuklu, Turkey	[23]	AG	-	AG	-	-	-
FertileC_NE				0.36 G	0.80 A	0.72 A	0.00 A	0.00 G	0.00 A
Aegean_NE	Bar8	Barcin, Turkey	[28]	AA	GA	AA	CA	CG	CC
Aegean_NE	Bar31	Barcin, Turkey	[28]	AA	AA	AA	AA	CG	CC
Aegean_NE	Rev5	Revenia, Greece	[28]	AA	-	AA	-	CC	CC
Aegean_NE				0.00 G	0.25 A	1.00 A	0.75 A	0.33 G	0.00 A
Körös_NE	KO1	Tiszaszőlős, Hungary	[29]	GG	GA	AG	CC	CC	CC
Körös_NE	KO2	Berettyóújfalu, Hungary	[29]	AG	AA	AG	CA	CG	CC
Körös_NE				0.75 G	0.75 A	0.50 A	0.25 A	0.25 G	0.00 C
Hungary_MN	NE1	Polgár-Ferenci-hát, Hungary	[29]	AG	AA	AG	CA	CC	CC
Hungary_MN	NE2	Debrecen Tócópart, Hungary	[29]	AA	GA	AA	CC	CC	CA
Hungary_MN	NE3	Garadna, Hungary	[29]	AG	GG	AA	CA	CG	CA
Hungary_MN	NE4	Polgár-Ferenci-hát, Hungary	[29]	GG	GA	AA	CA	CG	CC
Hungary_MN	NE5	Kompolt-Kigyósér, Hungary	[29]	AG	AA	AA	CA	CC	CA
Hungary_MN	NE6	Apc-Berekalja I., Hungary	[29]	GG	GA	AA	CA	CC	CC
Hungary_MN				0.58 G	0.58 A	0.92 A	0.42 A	0.17 G	0.25 A
Hungary_BA	BR1	Kompolt-Kigyósér, Hungary	[29]	AG	GA	AA	CA	CG	CC
Hungary_BA	BR2	Ludas-Varjú-dűlő, Hungary	[29]	AG	GA	AA	AA	GG	CC
Hungary_BA	S11	Balatonkeresztúr, Hungary	[30]	GG	GG	-	-	CC	CC
Hungary_BA	S14	Balatonkeresztúr, Hungary	[30]	AG	GG	-	AA	-	CC
Hungary_BA	S21	Balatonkeresztúr, Hungary	[30]	GG	-	AA	-	-	AA
Hungary_BA				0.70 G	0.25 A	1.00 A	0.84 A	0.50 G	0.20 A
Minoan	Pta08	Petras Siteia, Greece	[31]	AA	GG	AA	-	CG	-
Minoan	I0070	Charalambos Cave, Greece	[1]	AA	AA	AA	-	GG	AA
Minoan	I0071	Charalambos Cave, Greece	[1]	AA	AA	AA	CC	CC	CA
Minoan	I0073	Charalambos Cave, Greece	[1]	AA	AA	AA	AA	GG	-
Minoan	I0074	Charalambos Cave, Greece	[1]	AA	AG	AA	AA	GG	-
Minoan	I9005	Charalambos Cave, Greece	[1]	AA	AA	AA	-	GG	CA
Minoan	I9130	Odigitria, Greece	[1]	-	-	-	-	GG	CC
Minoan				0.00 G	0.75 A	1.00 A	0.67 A	0.79 G	0.50 A
Mycenaean	I9006	A. Kyriaki, Salamis, Greece	[1]	AA	GG	AA	-	GG	AA
Mycenaean	I9010	Galatas Apatheia, Greece	[1]	-	GG	AA	-	GG	AA
Mycenaean	I9033	Peristeria Tryfilia, Greece	[1]	AA	GG	AA	-	CG	AA
Mycenaean	I9041	Galatas Apatheia, Greece	[1]	AG	AG	AA	AA	CG	CA
Mycenaean	Log02	Logkas Elati, Greece	[31]	AA	GA	AA	CA	CC	-
Mycenaean				0.13 G	0.20 A	1.00 A	0.75 A	0.60 G	0.88 A

records the SNP variations data we could collect from 48 samples from the eleven archaeological cultures in Table 2. The 48 samples are listed in the second column of Table 3, and their locations are shown in Figure 2. For each culture, the boldface row gives the percentages for each of the six alleles of Table 1 that were associated with a lighter eye, hair, or skin pigmentation.

We calculated the RMSE between each pair of archaeological cultures using Equation (1), with n = 6. A lower RMSE value indicates a greater overall genotypic similarity between the populations of two cultures. For example, Figure 3 shows an RMSE value of 0.15 between the CHG and the FertileC_NE cultures. This indicates that the two cultures had similar genotypes and presumably also had similar phenotypes regarding eye, hair, and skin pigmentation. For more discussion of the results, see Section 6.

3. Method and Experiment 2: A G25 Admixture Analysis of Archaeogenetic Data

3.1. Method of Identifying the Ancestors of the Minoans Using the G25 System

Given a set of archaeological cultures S₁, S₂, …, S_n, and T, an admixture analysis finds an apportionment among the S₁, S₂, …, S_n cultures, which are called the source cultures, that seems to best explain the archaeological culture T, which is called the test culture. For example, Lazaridis et al. [1] used the popular qpADM admixture analysis system with S₁ = CHG, S₂ = Anatolia_N, and T = Minoan_Odigitria, that is, all the samples from the Minoan site of Moni Odigitria. The qpADM admixture analysis system returned the result that the Minoan_Odigitria culture is composed of 14.4 percent CHG and 85.6 percent Anatolia_N.

Unfortunately, the qpADM system is limited to 2 or 3 possible sources in most archaeogenetic publications. This creates a severe limitation, because hundreds of archaeological cultures located on the coastal areas of the Mediterranean Sea, the Black Sea, and the Atlantic Ocean could be possible genetic sources of the Minoans to some extent. Hence, we would need to simultaneously compare hundreds of possible sources for a completely fair apportionment among all those archaeological cultures. Luckily, the G25 genome admixture analysis system, which is available at https://www.dnagenics.com/products/g25studio (accessed on 30 July 2021), can compare hundreds of possible sources. The G25 system describes each archaeological culture by a numerical vector of length 25, which summarizes thousands of SNPs.

3.2. Experiment with the G25 System

The G25 system listed 271 different Neolithic, Mesolithic, and Paleolithic cultures in Africa, Asia, and Europe. We considered these 271 cultures as potential sources. While the Minoan civilization flourished on the island of Crete, many of the other Aegean islands were part of the Cycladic culture (c. 5100–3000 BP) [15], which we included in this experiment for comparison. We separately tested the Cycladic samples from Koufonisia island (Kou01 and Kou03) and the Minoan samples from the Charalambos Cave (I0070, I0071, I0073, I0074, I9005), Moni Odigitria (I9129, I9130, I9131), and Petras (Pta08).

Figure 4 shows the G25 admixture analysis results based on [33] with a listing of only those rows that had some non-zero value for at least one of the eleven tested samples. Figure 4 presents the data by grouping together the archaeological culture sources into five main regions: (1) Africa, (2) Greece and Macedonia, (3) Danube Basin, (4) Caucasus, Russia and Ukraine, and (5) Fertile Crescent and Iran.

The G25 analysis found a previously completely overlooked genetic connection between the Odigitria I9129 sample and a hunter-gatherer from the Shum Laka rock shelter in Cameroon about 8000 years ago. One hypothesis to explain the connection is that a common source population once lived in the Sahara. When the Sahara dried up, some people moved north into Europe and reached the island of Crete, while others from the same group moved south to Cameroon. These hypothetical movements could explain the linguistic connections between African and European mountain names [34].

The G25 analysis also reveals large differences among the various Cycladic and Minoan groups that were overlooked by previous admixture analysis publications. Figure 4 shows that there are great differences between the Charalambos Cave and the Moni Odigitria samples. The Charalambos Cave’s primary source is the Greek Neolithic (62.2 percent), and the secondary sources are the Danube Basin (19.4 percent), the Caucasus (15.1 percent), and the Fertile Crescent (3.2 percent). In contrast, the Moni Odigitria’s primary source is the Danube Basin (72.1 percent), and the secondary sources are the Fertile Crescent (15.1 percent), the Greek Neolithic (9.5 percent), the Caucasus (3.1 percent), and Africa (0.2 percent) on average. Hence, the two cultures are very different in origin and possibly came to Crete at different times as well. The Cyclades’ primary source is the Greek Neolithic (42 percent). Hence, the Cyclades and the Charalambos Cave samples form a natural cluster. The Petras’ primary source is the Danube Basin. Hence, the Moni Odigitria and the Petras samples also form a natural cluster.

Figure 5 shows some hypothetical population movements based on the above analysis. The map suggests that while the Anatolian and Fertile Crescent Neolithic culture reached Crete, major population movements to the island only happened later.

Figure 4. The sources (rows) for various Cycladic samples (columns 2–3) and Minoan samples (columns 4–12), as well as some averages (columns 13–15) according to the G25 admixture analysis system. The sources have been grouped into five regions (column 1).

Figure 4. The sources (rows) for various Cycladic samples (columns 2–3) and Minoan samples (columns 4–12), as well as some averages (columns 13–15) according to the G25 admixture analysis system. The sources have been grouped into five regions (column 1).

Figure 5. The Greek and Macedonian Neolithic cultures are the primary sources of the Cycladic and the Minoan Charalambos samples (red), while the Danube Basin Neolithic cultures are the primary sources of the Minoan Odigitria and Petras samples (blue) according to the G25 admixture analysis. The red and dark blue lines show hypothetical migrations.

Figure 5. The Greek and Macedonian Neolithic cultures are the primary sources of the Cycladic and the Minoan Charalambos samples (red), while the Danube Basin Neolithic cultures are the primary sources of the Minoan Odigitria and Petras samples (blue) according to the G25 admixture analysis. The red and dark blue lines show hypothetical migrations.

One of the major population movements was from the Peloponnese Peninsula. This movement reached both the Cyclades and Central Crete, including the Lassithi Plateau, where the Charalambos Cave is located. Another major movement was from the Danube Basin. This movement reached eastern Crete, including the town of Petras and southern Crete, where Moni Odigitria can be found near Phaistos.

Figure 6 shows on the x and y axes the 1st and 2nd principal components of the principal component analysis generated by the G25 system. The principal component analysis shows that the Moni Odigitria samples (the red triangle) are directly below the Middle Neolithic Linear Pottery culture samples from Hungary (HUN_LBK_MN). The Petras sample (GRC_Minoan_EBA) is located directly below the Moni Odigitria samples. In contrast, the Hagia Charalambos samples from the Lassithi Plateau (the green pentagon), the two Cycladic samples (the dashed brown line), and the Mycenean samples (purple quadrangle) are located below the Moni Odigitria samples and most to its left or right. Hence, the principal component analysis in Figure 6 supports the analysis in Figure 4.

4. Method and Experiment 3: Analysis of the Origin of the Minoan U5a1 Haplogroup

4.1. Method of Analyzing the Origin of the Minoan U5a1 Haplogroup

The method of analyzing the origin of the Minoan U5a1 Haplogroup relies on the available data from the Ancient DNA Database [35]. We found where U5, U5a1 and U5a1d2b samples are found at times before the arrival of Indo-Europeans to Europe. If these are found exclusively in Europe, then these haplogroups cannot originate from other continents and could not have been brought to Europe by Indo-Europeans.

The motivation to focus on the U5 haplogroup is that it is known to be associated with Uralic speakers. Table 4 shows that the U5 mtDNA haplogroup percentages are almost always higher among the Uralic speakers than among their Indo-European neighbors according to Simoni et al. [36], who examined more than 2600 mtDNA sequences from current European populations. For example, the U5 mtDNA haplogroup percentage is 48 percent among the Saami, while it is only 11.4 percent among the Norwegians, their Indo-European speaking neighbors.

While some caution is warranted because there are still relatively few ancient mtDNA samples from Greece, they show the same trend, with the U5a1 haplotype reported for 2 out of 11 of Minoan samples by Lazaridis et al. [1] (sample I0071 from the Charalambos Cave and sample I9123 from the Late Minoan cemetery at Armenoi, Crete), and the U5 haplotype reported for 1 out of 40 Mycenaean samples by Skourtanioti et al. [37]. This suggests that Proto-Uralic speakers had a high percentage of the U5 mtDNA haplogroup, and Proto-Indo-European speakers had no U5 mtDNA haplogroup initially. However, the percentage decreased among the Uralic speakers and increased among their Indo-European neighbors due to millennia of genetic admixture.

Table 4. mtDNA haplogroup U5 within Uralic language speaking and neighboring populations.

Uralic Speakers	U5 Percent	Indo-European Neighbors	U5 Percent	Source
Saami	48	Norway	11.4	Simoni et al. [36]
Finland	20.7	Sweden	12.1	Simoni et al. [36]
Moksha	18.9	Russia	10.4	Bramanti et al. [38]
Minoan	18.2	Mycenaean	2.5	[1], Skourtanioti et al. [37]
Mordovians	15.7	Russia	10.4	Simoni et al. [36]
Mari	14	Russia	10.4	Simoni et al. [36]
Estonia	13.3	Latvia	10	Simoni et al. [36]
Basques	11.7	Spain	8.1	Simoni et al. [36]
Udmurt	8.9	Russia	10.4	Simoni et al. [36]
Hungary	7.4	Romania	7.2	Simoni et al. [36]

Table 4. mtDNA haplogroup U5 within Uralic language speaking and neighboring populations.

Uralic Speakers	U5 Percent	Indo-European Neighbors	U5 Percent	Source
Saami	48	Norway	11.4	Simoni et al. [36]
Finland	20.7	Sweden	12.1	Simoni et al. [36]
Moksha	18.9	Russia	10.4	Bramanti et al. [38]
Minoan	18.2	Mycenaean	2.5	[1], Skourtanioti et al. [37]
Mordovians	15.7	Russia	10.4	Simoni et al. [36]
Mari	14	Russia	10.4	Simoni et al. [36]
Estonia	13.3	Latvia	10	Simoni et al. [36]
Basques	11.7	Spain	8.1	Simoni et al. [36]
Udmurt	8.9	Russia	10.4	Simoni et al. [36]
Hungary	7.4	Romania	7.2	Simoni et al. [36]

4.2. Experiment with U5, U5a1, and U5a1d2b Haplogroup Data

We used the Ancient DNA Database [35] to map all the U5 haplogroup samples before 8000 BP (Figure 7 top), the U5a1 haplogroup samples before 6000 BP (Figure 7 middle), and the U5a1d2b haplogroup samples before 5500 BP (Figure 7 bottom).

Figure 7 (top) shows (purple symbols) that the U5 haplogroup appeared first among Gravettian hunter-gatherers in present-day Dolní Věstonice, Czechia, around 30,800 BP [39] spread to other areas of Europe but not to other continents by 8000 BP.

Figure 7 (middle) shows (dark blue symbol) that the U5a1 mtDNA haplogroup appeared first in the Iron Gates gorges area in the lower Danube Basin around 10,530 BP. The U5a1 haplogroup was concentrated in the Danube Basin, as well as some areas that are considered to have been long inhabited by Uralic speakers such as the Baltic Sea region and the middle Volga region.

Figure 7 (bottom) shows the distribution of the even more specific haplogroup U5a1d2b before 5500 BP. The Baltic Sample, Tamula22, is from the Combed Ware culture in Estonia, and the other sample, MUR019, from Murzikhinsky II, Russia, is associated with the Eneolithic Volga-Kama culture. These cultures are commonly associated with early Uralic speakers. In addition, three ‘early Hungarian’ U5a1d2b samples, BAL23.6B, NTHper1, and VPBoer51, from 1000 to 1000 CE are also known [35].

The U5, U5a1, and U5a1d2b haplogroups do not have an Indo-European origin because these haplogroups are native to Europe, while the Indo-Europeans came to Europe only around 5300 BP according to the Kurgan Hypothesis of Indo-European origin [16]. However, these haplogroups were absorbed by the Mycenaeans and other Indo-Europeans after their arrival to Europe, as shown by the Mycenaean U5a1d2b sample from Aidonia, northern Greece, that was found by Skourtanioti et al. [37].

5. Method and Experiment 4: Analysis of Maximal Minoan Haplogroups

The previous experiment showed that the U5a1 haplogroup, which was found in two Minoan samples, had to come from the Danube Basin from where it originated. The next experiment investigates the origin of all the mtDNA and y-DNA haplogroups that were observed in the Minoan samples.

5.1. Method of Analyzing Maximal Minoan Haplogroups

We start by giving a definition to better describe the experiment in this section.

Definition 1. 

Given a set S = {S₁, …, S_n} of mtDNA (or y-DNA) haplogroups, any S_i is a maximal mtDNA (or y-DNA) haplogroup in S if there is not another S_j ∈ S such that S_i is a prefix or beginning of S_j.

For example, if S = {U5, U5a1, U5a1d2b} is a set of mtDNA haplogroups, then U5 is not a maximal haplogroup in S, because U5 is the prefix or beginning of U5a1, which is also in S. However, U5a2d2b is a maximal mtDNA haplogroup.

Clearly, the maximal haplogroups carry the most valuable information, because they can be found in fewer places than the non-maximal haplogroups. Hence, to make our search efficient, it is enough to focus on identifying the origin of those Minoan mtDNA and y-DNA haplogroups that are maximal.

Hence, the method is to find for each maximal Minoan mtDNA and y-DNA from which the following three regions it could possibly come from: (1) the western Mediterranean coastal regions, (2) the Black Sea coastal regions, or (3) the Fertile Crescent. We searched for samples with the same haplogroup or even a more specific haplogroup from these three regions from a time before 3700 BP. If we found several samples, then we picked the one that was closest to Crete. If we found no samples, then we wrote down N/A for ‘not available’. At the end of the process, we found the total number of N/As for each of these three regions. The higher the number of N/As for a region, the less likely that the Minoans came from that region.

5.2. Experiment with Maximal Minoans Haplogroups

We only analyzed the Early Minoan and Middle Minoan samples from Aposelemis, Charalambos, Odigitria, and Petras because some later samples from the other sites could be Mycenaean samples, meaning that due to the Mycenaeans’ occupation of the island Crete during the Late Minoan period [15]. All the Minoan samples that we considered are dated to 3700 BP or earlier.

Table 5 lists and Figure 8 shows the Minoan mtDNA and y-DNA samples and their closest matches to Mediterranean, Black Sea region, and Fertile Crescent samples.

The second column of Table 5 gives the sample ID and reference to the source for each y-DNA and mtDNA haplogroup sample. If there are no 3700 BP samples in the database, then a ‘not available’ (N/A) is shown.

We revised some of the reported mtDNA haplogroups of Hughey et al. [40], because their earlier classifications were sometimes not as specific as possible. The revisions are indicated by ‘(rev)’ in the first column of Table 5. Some revisions were already given by Revesz [41], but the ones listed in Table 6 are further improvements.

The revisions use Hughey et al. [40]’s reported mutations with respect to the rCRS reference sequence. The reported mutations are shown in black in the last column of Table 6. Unfortunately, the technique used by Hughey et al. [40] can miss many mutations, because only fragments of the mtDNA are scanned. It appears that the mutations shown in red were missed, because they were needed on a path from the root to the new haplogroup in the PhyloTree [42] used for mtDNA haplogroup classification. For example, the reported mutation 14055T implies that the haplogroup classification should be H41a1a, even though the red ones are missing in the second row.

Table 5. mtDNA and y-DNA (blue) haplotypes shared by DNA samples (1st–3rd columns) from Minoan, Mediterranean, Black Sea, and Fertile Crescent locations (4th–7th columns). N/A means there are no non-Minoan samples from at least 3700 BP.

mtDNA/y-DNA	Sample ID	BP	Minoan	Mediterranean Region	Black Sea Region	Fertile Crescent
G2a2b2a1a1	Pta08 [43]	4685	Petras
				N/A
G2a2b2a1a1c2	PIE015 [44]	6534			Pietrele, Romania
						N/A
H1bm	I0073 [35]	4000	Charalambos
				N/A
H1bm	I8531 [35]	5050			Geoksyur, Turkmenistan
						N/A
H2a2a1d (rev)	8H [40]	3700	Charalambos
				N/A
					N/A
H2a2a1d	CCH290 [35]	8590				Çatalhöyük, Turkey
H4a1	HGC005 [37]	4178	Charalambos
H4a1a	CRE14 [35]	6302		Béziers, France
H4a1	PIE048 [44]	6586			Pietrele, Romania
						N/A
H5	HGC017 [37]	EMBA	Charalambos
H5	I4565 [35]	4915		Galls Carboners, Cat., Spain
H5	I0679 [35]	7617			Krepost, Bulgaria
						N/A
H7	12AH [40]	3700	Charalambos
H7c	I5072 [35]	7551		Kargadur, Croatia
H7	PIE014 [44]	6455			Pietrele, Romania
						N/A
H13a1a	I0070 [1]	4000	Charalambos
				N/A
H13a1a1	BOY009 [44]	4799			Boyanovo, Bulgaria
						N/A
H41a1a (rev)	6AH [40]	3700	Charalambos
				N/A
H41a	BOL003 [35]	4408			Bolshnevo, Tver, Russia
						N/A
H102	HGC041 [37]	EMBA	Charalambos
H102	I14689 [45]	4568		Çinamak, Albania
					N/A
						N/A
HV-b	HGC018 [37]	EMBA	Charalambos
				N/A
HV-b	PIE057 [44]	6421			Pietrele, Romania
						N/A
I1	HGC040 [37]	4134	Charalambos
I1	Neolithic 5 [35]	5200		Camí de Can Grau, Spain
I1a1	MAJ008 [44]	6110			Majaky, Ukraine
						N/A
I5a	HGC024 [37]	3700	Charalambos
				N/A
I5a	PIE063 [44]	6460			Pietrele, Romania
						N/A
J1a2a1a2~	HGC001 [37]	EMBA	Charalambos
				N/A
J1a2a1a2d2b2b2~	I16120 [45]	3390			Dzori Gekh, Armenia
J1a2a1a2d2b2	ETM012 [35]	4470				Ebla, Syria
J2a1a1a2b1b2	HGC006 [37]	EMBA	Charalambos
				N/A
J2a1a1a2b1b	ART020 [45]	5177			Arslantepe, Turkey
						N/A
J2b1a1	ERS1770867 [1]	3895	Odigitria
J2b1a1	I8153 [35]	4650		Sima del Ángel, Luc., Spain
J2b1a1b	I23210 [43]	3900			Vojvodina, Serbia
						N/A
K1a2	ERS1770871 [1]	3895	Odigitria
K1a2a	CB13 [35]	7345		Cova Bonica, Cat., Spain
K1a2	I2532 [35]	7614			Coțatcu, Romania
						N/A
K1a4	HGC027 [37]	EMBA	Charalambos
				N/A
K1a4	PIE065 [44]	6568			Pietrele, Romania
K1a4	Ash129 [35]	10093				Aşıklı, Turkey
K2b1	APO023 [37]	3558	Aposelemis
K2b1	I4065 [35]	6815		Fossato di Stretto Partana, IT
K2b1c	POP06 [43]	6450			Popova, Croatia
						N/A
T1	9H [40]	3700	Charalambos
				N/A
T1a	VAR016 [44]	6452			Varna, Bulgaria
T1a2	I1727 [35]	10050				‘Ain Ghazal, Jordan
T2b25	HGC008 [37]	4219	Charalambos
T2b	584 [35]	4950		Treilles cave, France
T2b	PIE008 [44]	6422			Pietrele, Romania
						N/A
T2c1d	HGC020 [37]	EMBA	Charalambos
T2c1d1	I15946 [35]	5968		Anghelu Ruju, Sardinia
T2c1d1	PIE030 [44]	6259			Pietrele, Romania
						N/A
T2e6 (rev)	21H [40]	3700	Charalambos
T2e	Bar10 [35]	4710		Barranc dˈen Rifà, Spain
T2e	I0700 [35]	7912			Malak Preslavets, Bulgaria
T2e	CCH311 [35]	8520				Çatalhöyük, Turkey
U1a1a-a	HGC010 [37]	EMBA	Charalambos
				N/A
U1a1a3a*	PG2002 [43]	4361			Progress 2, Russia
						N/A
U3b3	I9130 [35]	3895	Odigitria
				N/A
U3b3	KTL005 [44]	4905			Kartal, Ukraine
						N/A
U5a1f1 (rev)	4H [40]	3700	Charalambos
				N/A
U5a1f1	MAJ020 [44]	5871			Majaky, Ukraine
						N/A
U7b	HGC053 [37]	EMBA	Charalambos
				N/A
U7b*	SA6001 [43]	5444			Sharakhalsun 6, Russia
						N/A
U8b1b4 (rev)	M4 [40]	3700	Charalambos
				N/A
U8b1b4	I2378 [35]	7050			Hejőkürt, Hungary
						N/A
W	6H [40]	3700	Charalambos
				N/A
W3b	PIE022 [44]	6392			Pietrele, Romania
W1c4*	MK308703.1 [43]	8365				Çatalhöyük, Turkey
X2b (rev)	M8 [40]	3700	Charalambos
X2b	Rev5 [35]	8316		Revenia, Greece
X2b	USV005 [44]	5588			Usatove, Ukraine
X2b4*	MK308702.2 [43]	8365				Çatalhöyük, Turkey
--	Total N/As	--		17	2	21
--	Percent N/As	--		58.6	6.9	72.4

Table 5. mtDNA and y-DNA (blue) haplotypes shared by DNA samples (1st–3rd columns) from Minoan, Mediterranean, Black Sea, and Fertile Crescent locations (4th–7th columns). N/A means there are no non-Minoan samples from at least 3700 BP.

mtDNA/y-DNA	Sample ID	BP	Minoan	Mediterranean Region	Black Sea Region	Fertile Crescent
G2a2b2a1a1	Pta08 [43]	4685	Petras
				N/A
G2a2b2a1a1c2	PIE015 [44]	6534			Pietrele, Romania
						N/A
H1bm	I0073 [35]	4000	Charalambos
				N/A
H1bm	I8531 [35]	5050			Geoksyur, Turkmenistan
						N/A
H2a2a1d (rev)	8H [40]	3700	Charalambos
				N/A
					N/A
H2a2a1d	CCH290 [35]	8590				Çatalhöyük, Turkey
H4a1	HGC005 [37]	4178	Charalambos
H4a1a	CRE14 [35]	6302		Béziers, France
H4a1	PIE048 [44]	6586			Pietrele, Romania
						N/A
H5	HGC017 [37]	EMBA	Charalambos
H5	I4565 [35]	4915		Galls Carboners, Cat., Spain
H5	I0679 [35]	7617			Krepost, Bulgaria
						N/A
H7	12AH [40]	3700	Charalambos
H7c	I5072 [35]	7551		Kargadur, Croatia
H7	PIE014 [44]	6455			Pietrele, Romania
						N/A
H13a1a	I0070 [1]	4000	Charalambos
				N/A
H13a1a1	BOY009 [44]	4799			Boyanovo, Bulgaria
						N/A
H41a1a (rev)	6AH [40]	3700	Charalambos
				N/A
H41a	BOL003 [35]	4408			Bolshnevo, Tver, Russia
						N/A
H102	HGC041 [37]	EMBA	Charalambos
H102	I14689 [45]	4568		Çinamak, Albania
					N/A
						N/A
HV-b	HGC018 [37]	EMBA	Charalambos
				N/A
HV-b	PIE057 [44]	6421			Pietrele, Romania
						N/A
I1	HGC040 [37]	4134	Charalambos
I1	Neolithic 5 [35]	5200		Camí de Can Grau, Spain
I1a1	MAJ008 [44]	6110			Majaky, Ukraine
						N/A
I5a	HGC024 [37]	3700	Charalambos
				N/A
I5a	PIE063 [44]	6460			Pietrele, Romania
						N/A
J1a2a1a2~	HGC001 [37]	EMBA	Charalambos
				N/A
J1a2a1a2d2b2b2~	I16120 [45]	3390			Dzori Gekh, Armenia
J1a2a1a2d2b2	ETM012 [35]	4470				Ebla, Syria
J2a1a1a2b1b2	HGC006 [37]	EMBA	Charalambos
				N/A
J2a1a1a2b1b	ART020 [45]	5177			Arslantepe, Turkey
						N/A
J2b1a1	ERS1770867 [1]	3895	Odigitria
J2b1a1	I8153 [35]	4650		Sima del Ángel, Luc., Spain
J2b1a1b	I23210 [43]	3900			Vojvodina, Serbia
						N/A
K1a2	ERS1770871 [1]	3895	Odigitria
K1a2a	CB13 [35]	7345		Cova Bonica, Cat., Spain
K1a2	I2532 [35]	7614			Coțatcu, Romania
						N/A
K1a4	HGC027 [37]	EMBA	Charalambos
				N/A
K1a4	PIE065 [44]	6568			Pietrele, Romania
K1a4	Ash129 [35]	10093				Aşıklı, Turkey
K2b1	APO023 [37]	3558	Aposelemis
K2b1	I4065 [35]	6815		Fossato di Stretto Partana, IT
K2b1c	POP06 [43]	6450			Popova, Croatia
						N/A
T1	9H [40]	3700	Charalambos
				N/A
T1a	VAR016 [44]	6452			Varna, Bulgaria
T1a2	I1727 [35]	10050				‘Ain Ghazal, Jordan
T2b25	HGC008 [37]	4219	Charalambos
T2b	584 [35]	4950		Treilles cave, France
T2b	PIE008 [44]	6422			Pietrele, Romania
						N/A
T2c1d	HGC020 [37]	EMBA	Charalambos
T2c1d1	I15946 [35]	5968		Anghelu Ruju, Sardinia
T2c1d1	PIE030 [44]	6259			Pietrele, Romania
						N/A
T2e6 (rev)	21H [40]	3700	Charalambos
T2e	Bar10 [35]	4710		Barranc dˈen Rifà, Spain
T2e	I0700 [35]	7912			Malak Preslavets, Bulgaria
T2e	CCH311 [35]	8520				Çatalhöyük, Turkey
U1a1a-a	HGC010 [37]	EMBA	Charalambos
				N/A
U1a1a3a*	PG2002 [43]	4361			Progress 2, Russia
						N/A
U3b3	I9130 [35]	3895	Odigitria
				N/A
U3b3	KTL005 [44]	4905			Kartal, Ukraine
						N/A
U5a1f1 (rev)	4H [40]	3700	Charalambos
				N/A
U5a1f1	MAJ020 [44]	5871			Majaky, Ukraine
						N/A
U7b	HGC053 [37]	EMBA	Charalambos
				N/A
U7b*	SA6001 [43]	5444			Sharakhalsun 6, Russia
						N/A
U8b1b4 (rev)	M4 [40]	3700	Charalambos
				N/A
U8b1b4	I2378 [35]	7050			Hejőkürt, Hungary
						N/A
W	6H [40]	3700	Charalambos
				N/A
W3b	PIE022 [44]	6392			Pietrele, Romania
W1c4*	MK308703.1 [43]	8365				Çatalhöyük, Turkey
X2b (rev)	M8 [40]	3700	Charalambos
X2b	Rev5 [35]	8316		Revenia, Greece
X2b	USV005 [44]	5588			Usatove, Ukraine
X2b4*	MK308702.2 [43]	8365				Çatalhöyük, Turkey
--	Total N/As	--		17	2	21
--	Percent N/As	--		58.6	6.9	72.4

Another problem with Hughey et al. classifications [40] is that since they reported the mutations with respect to rCRS, which belongs to haplogroup H2a2a1, some mutations with respect to RSRS cannot be expected to be reported if there is an agreement on these rCRS mutations between the analyzed sample and the rCRS. These not-expected-to-be-reported mutations could be assumed to be present in the third sample because it also has the mutation 16172C, which indicates that its haplogroup classification is most likely H2a2a1d, although H66a is also a possible classification based on the 2706A and 16172C mutations.

Table 6. Updating the mtDNA haplogroup classifications of Hughey et al. [40]. The sample IDs are from the European Nucleotide Archive database, which is available online: https://www.ebi.ac.uk/ena/browser/home (accessed on).

	mtDNA
Sample ID	Old	New	Mutations with Respect to RSRS
HM022275	H	H41a1a	15617A (H41), 262T, 5460A, 10124C, 14118G (H41a), 16362T! (H41a1), 14055T (H41a1a)
HM022291	U5a	U5a1f1	16192T, 16270T (U5), 3197C, 9477A, 13617C (U5a’b), 14793G, 16256T (U5a), 15218G, 16399G (U5a1), 6023A (U5a1f), 5585A, 7569G, 16311C! (U5a1f1)
HM022294	H	H2a2a1d	2706A, 7028C (H), 1438A (H2), 4769A (H2a), 750A (H2a2), 8860A, 15326A (H2a2a), 263A (H2a2a1), 16172C (H2a2a1d)
HM022303	T	T2e6	11812G, 14233G, 16296T (T2), 150T (T2-a), 16153A (T2e), 16240C (T2e6)
HM022308	U	U8b1b4	9698C (U8), 3480G (U8b’c), 9055A, 14167T (U8b), 195C!, 16189C!, 16234T (U8b1), 1811A!, 5165T, 16324C (U8b1b), 16290T (U8b1b4)
HM022312	X	X2b	6221C, 6371T, 13966G, 14470C, 16189C!, 16278T! (X), 153G (X1’2’3), 195C!, 1719A (X2), 225A (X2-a), 13708A (X2b’d), 8393T, 15927A (X2b)

Table 6. Updating the mtDNA haplogroup classifications of Hughey et al. [40]. The sample IDs are from the European Nucleotide Archive database, which is available online: https://www.ebi.ac.uk/ena/browser/home (accessed on).

	mtDNA
Sample ID	Old	New	Mutations with Respect to RSRS
HM022275	H	H41a1a	15617A (H41), 262T, 5460A, 10124C, 14118G (H41a), 16362T! (H41a1), 14055T (H41a1a)
HM022291	U5a	U5a1f1	16192T, 16270T (U5), 3197C, 9477A, 13617C (U5a’b), 14793G, 16256T (U5a), 15218G, 16399G (U5a1), 6023A (U5a1f), 5585A, 7569G, 16311C! (U5a1f1)
HM022294	H	H2a2a1d	2706A, 7028C (H), 1438A (H2), 4769A (H2a), 750A (H2a2), 8860A, 15326A (H2a2a), 263A (H2a2a1), 16172C (H2a2a1d)
HM022303	T	T2e6	11812G, 14233G, 16296T (T2), 150T (T2-a), 16153A (T2e), 16240C (T2e6)
HM022308	U	U8b1b4	9698C (U8), 3480G (U8b’c), 9055A, 14167T (U8b), 195C!, 16189C!, 16234T (U8b1), 1811A!, 5165T, 16324C (U8b1b), 16290T (U8b1b4)
HM022312	X	X2b	6221C, 6371T, 13966G, 14470C, 16189C!, 16278T! (X), 153G (X1’2’3), 195C!, 1719A (X2), 225A (X2-a), 13708A (X2b’d), 8393T, 15927A (X2b)

We explain some of the regional classifications as follows. The rare mtDNA H1bm sample from Geoksyur, Turkmenistan was counted as ‘Black Sea region’ for regional classification in Table 5 because of likely ancient migration from the Caspian Sea area via the Volga and the Don to the Black Sea area. These regions were connected by trade routes since at least the Bronze Age [46].

The rare mtDNA H41a sample from Bolshnevo, Tver, Russia, which is part of the Fatyanovo culture, was also counted as ‘Black Sea region’ because the Fatyanovo culture resulted from a migration from the Middle-Dnieper culture [47].

Finally, the rare y-DNA J2a1a1a2b1b sample from Arslantepe, Turkey was counted as the ‘Black Sea region’ because the Kura-Araxes culture moved south from the north of the Caucasus to Arslantepe around 3000 BCE, when there was widespread burning and destruction, after which Kura–Araxes culture pottery appeared in the area [48]. Moreover, this J2a1a1a2b1b sample, ART20, had blue eyes according to Lazaridis et al. [45] (supplement Table 5), further implying that this individual was part of the Kura–Araxes migration to the south.

6. Results and Discussion

6.1. Summary of the Results

Section 2 showed that six SNP mutations associated with light eyes, hair, and skin phenotypes and originating in Europe or Eurasia have a very high presence among the Minoan samples. The SNP mutations analysis showed that the Eastern-European Hunter-Gatherer culture (EHG) was closest to the Minoans, because Minoans had the lowest distance, 0.27, to the EHGs according to the root mean square error measure. Since the area of EHGs included present day Ukraine, the proximal genetic source of the Minoans was likely the northern Black Sea coastal region.

Section 3 presented a G25 admixture analysis that showed that the Minoan samples from Odigitria and Petras likely have a Danube Basin origin, while the Minoan Charalambos samples likely have a Greek mainland origin.

Section 4 showed that the U5, U5a1, and U5a1d2b mtDNA haplogroups are native to Europe. The U5 mtDNA haplogroup is frequent among Uralic speakers and could be found among several Minoans, whereas the U5 haplogroup was absent in Neolithic Anatolia.

Section 5 traced back the origins of each known Minoan mtDNA and y-DNA haplogroup. The analysis involved looking at the most specific haplogroups that can be identified based on the current PhyloTree [42] classification. Table 5 listed 29 different Minoan mtDNA and y-DNA haplotypes. The bottom row showed that out of these 29 Minoan haplotypes 17, or 58.6 percent, could not have come from the Mediterranean region, and 21, or 72.4 percent, could not have come from the Fertile Crescent, while only 2, or 6.9 percent, could not have come from the Black Sea region. This also seems to suggest that the proximal genetic source of the Minoans was overwhelmingly the Black Sea region.

6.2. Discussion on the Phenotypes

Blue eyes, which are associated with the HERC2 rs12913832-A allele, likely originated about 42,000 years ago among the WHGs, where it has the highest frequency [49]. This allele spread widely and can be found among hunter-gatherers, as well as farmers, except the Aegean early Neolithic farmers and the Bronze Age Minoans.

Light skin color is associated with the SLC45A2 rs28777-A and the SLC45A2 rs16891982-G alleles among other alleles. Both alleles can be found among the EHGs with a high percentage, and the second allele was also found in a Paleolithic hunter-gatherer (Kostenki14) in the Don River area [23]. Hence, these alleles seem ancient in the area where EHGs also lived. These alleles apparently spread from the EHGs to the WHGs at Loschbour, to the Lower Danube Mesolithic hunter-gatherers, and to European Neolithic and Bronze Age cultures, except the Fertile Crescent Neolithic culture [23].

The SLC24A4 rs2402130-A allele is associated with light hair, and the SLC24A5 rs1426654-A allele is associated with light skin. These alleles are present in all CHG and EHG samples. Hence, these alleles likely originated in a common ancestor of these hunter-gatherers around the Caucasus area and spread to other regions, except the rs1426654-A allele did not reach the Mesolithic Lower Danube and the WHG cultures.

The TYR rs1042602-A allele is also associated with light skin and a lower occurrence of freckles. This allele is absent from both hunter-gatherers and the early farmers of the Fertile Crescent, the Aegean, and the Körös River area. It seems to first have occurred in the Hungarian Middle Neolithic culture around 5000 BC in three samples (NE2, NE3, and NE5) [29]. This allele continued in the Hungarian Bronze Age and has been found among the Minoans and the Mycenaeans. This allele apparently spread from the Danube Basin southward to the Aegean area.

Hence, the SLC45A2 rs28777-A, SLC45A2 rs16891982-G, SLC24A4 rs2402130-A, and SLC24A5 rs1426654-A alleles suggest a genetic connection between the EHGs and the Minoans. The TYR rs1042602-A allele suggests another genetic connection between the Hungarian Middle Neolithic farmers and the Minoans. Furthermore, the lack of the HERC2 rs12913832-A allele makes it unlikely that either WHGs or FertileC_NE farmers reached Crete in significant numbers. While farming spread to Crete during the early phase of the Neolithic, the lack of the HERC2 rs12913832-A allele suggests that farming reached Crete from the Aegean_NE culture, which also lacks this allele, rather than directly from the FertileC_NE, where the allele is present in a significant percentage of the samples. While the Aegean_NE culture learned farming from the FertileC_NE culture, a likely genetic admixture with local Aegean hunter-gatherers who lacked the HERC2 rs12913832-A allele may have diluted this allele to an insignificant percentage before reaching Crete.

The genetic admixture between hunter-gatherers and early farmers is most noticeable in the Körös_NE culture. In fact, KO1 did not cluster together with early European farmers according to a study by Gamba et al. [29]. The genetic admixture between hunter-gatherers and farmers is also well-documented at the Iron Gates gorges area [50], where the Danube crosses the Carpathian Mountains. Most early farmers likely passed through the Iron Gates before entering the Carpathian Basin, where the early Neolithic Körös_NE and the Middle Neolithic Hungary_MN cultures also flourished.

A common problem in archaeogenetics is the low sample size, which may cause statistical errors. For example, we have only twelve Minoan allele samples (two allele samples from six individuals) regarding the rs12913832 genetic locus. While none of these allele samples had the mutation that causes blue eye color, there is still a certain probability that a blue-eyed Minoan sample will be found later.

6.3. Discussion on the G25 System

The results of the G25 admixture analysis system need to be handled with caution, because the reliability of the system is not yet well-tested. However, the main result of a movement from the Danube Basin to Crete is also supported by some studies on climate change and the spread of agriculture.

The exact time and reason for these population movements shown in Figure 5 is unknown currently. However, the second movement may be related to the 4.2 kiloyear BP aridification event [51] that dried out the Danube Basin and made agriculture infeasible there. That may have caused the Danube Basin farmers to move to the Messara Plain in southern Crete, which may have provided better agricultural and fishing opportunities. The distinguishing of these two major population movements has major implications regarding the languages spoken in different areas of Crete and the decipherment of the Minoan scripts.

These hypothetical migrations are also supported by the presence of millet grains at the Minoan sites such as Chania, Knossos and Zominthos starting from the Neopalatial period [52]. Livarda and Kotzamani [52] speculate that millet had reached Crete from Bronze Age Central Europe, where it was commonly cultivated.

6.4. Discussion on U5, U5a1, and U5a1d2b Haplogroups

Some of the Gravettian hunter-gatherers found refuge in the lower Danube Basin and the northern Pontic coastal areas during the subsequent Ice Age. The Proto-Uralic language likely developed in this refuge area during the Ice Age and broke up sometime during the Mesolithic period when some of the Uralic speakers went northeast. These early Uralic speakers may have followed the mammoth herds, which also moved from this area north to the Baltic and southern Scandinavia, where remains were found between 17 and 12 thousand years ago [53].

The U5 mtDNA haplogroup is strongly associated with Uralic language speakers, because the Uralic language speakers had matrilineal cultures in the past. Since the husband moves to the village of the wife in a matrilineal society, their children will speak the mother’s language. Hence, in matrilineal societies the mother’s language is passed on in parallel to the mother’s mtDNA. In contrast, the Yamnaya and other early Indo-European cultures were patrilineal. Since the wife moves to the village of the husband in a patrilineal culture, their children will speak the father’s language. Hence, Indo-European language speakers are more commonly associated with the R1 y-DNA haplogroup [54].

6.5. Discussion on the Minoan Maximal Haplogroups

While the U haplogroup was the dominant European hunter-gatherer haplogroup, other haplogroups arrived from the Fertile Crescent and the Caucasus. Despite the new haplogroups, the Neolithic Old European Civilization (Gimbutas [55,56]) or the Danube Civilization (Haarmann [57]) was likely Uralic speaking, because the neolithization of the Danube Basin was a slow process taking place over thousands of years. Hence, those who came earlier from the Anatolia may have learned the local Uralic language, and they and their descendants may have taught it to those who came later from Anatolia. Hence, while the incoming Anatolian famers’ total genetic effect on the local population was considerable after several millennia, their linguistic effect may have been small.

The geography of the Carpathian Basin may have helped in the process of unifying the spoken language there. If one follows the Danube River, then entering the Carpathian Basin requires going through the Iron Gates gorge, which is a natural geographic barrier. The Iron Gates barrier likely slowed down the inward movement of any wave of newcomers. Since it is a defensible barrier, passing through it may have required cultivating friendly relations with the locals, and that likely resulted in intermingling between the local population and the newcomers.

Brami et al. [50] found evidence of this intermingling studied at the site of Lepenski Vir, near the Iron Gates gorge. Brami at al. [50] found one individual with only hunter-gatherer genes, three individuals with some genetic admixture between hunter-gatherers and Aegean farmers, and two individuals with only Aegean farmer genes between 6100 and 6000 BCE. In addition, two individuals had only hunter-gatherer genes, and three individuals had only Aegean farmer genes before this transition period. Furthermore, out of the three admixed individuals, two belonged to the U5 and one to the H mtDNA haplogroup. These two haplogroups were already present in the hunting-gathering period before 7400 BCE. This suggests that Aegean farmers moving into the Lepenski Vir community married local hunter-gatherer women. Moreover, if such an intermingling happened at the Iron Gates area between hunter-gatherers and Aegean farmers, then it likely happened with even greater ease later between the already neolithic local Iron Gates population and later Aegean farmer newcomers. This suggests that the local hunter-gatherer language did not change with the arrival of Aegean farmers. Therefore, the Old European mtDNA and y-DNA haplogroups can be associated with Uralic languages.

7. Conclusions and Further Work

Four different experiments of Section 2, Section 3, Section 4 and Section 5 suggest that the proximal sources of the Minoans were the Danube Basin and the Black Sea coastal area, which overlap in the Danube Delta area, providing easy migration opportunities between them. Future work needs to look at the rapidly increasing ancient DNA data to be able to make a statistically stronger conclusion and to further narrow down the proximal source of the Minoans.

Lazaridis et al. [1] presented the first whole-genome sequences for Minoan samples. That is a lasting contribution to archaeogenetics, but their data analysis is flawed, because they overlooked the Danube Basin and the Black Sea coastal area as a possible proximal source. Hence, their claim that the proximal source of the Minoans was Anatolia or the Fertile Crescent has to be abandoned. This correction regarding the origin of the Minoans helps to reconcile archaeogenetics with linguistic work that links the Minoan language to the Uralic language family. The reconciliation of archaeogenetics and linguistics would follow, because the Danube Basin and the northern Black Sea coastal areas were identified by some researchers as potential Uralic speaking areas before the arrival of the Yamnaya people, who are believed to have spoken an Indo-European language [6,7]. The arrival of the Yamnaya may have prompted the Minoans to sail to Crete. It also may have prompted other Uralic peoples to move away from the Steppe, although the precise route and timing of their migrations remains an open problem.

Unfortunately, furthering the incorrect view that the proximal source of the Minoans was Anatolia or the Fertile Crescent would continue to lead linguists to suspect that the Minoan language is related to Near Eastern or African languages [17,18,19,20,21,22]. A search in those regions for language connections with Minoan did not yield any result for over a century. Both the sequencing and the data mining of archaeogenetic data have to be correct to aid instead of hinder linguistic discoveries.