December 07, 2012

23andMe Ancestry Composition

23andMe has launched its new Ancestry Composition feature, the workings of which are summarized -at a very high level- in this page.

I have already received some feedback from customers who also happen to be part of my Dodecad Project and who appear to be perplexed by their results. It is unfortunate that my own rules preclude me from discussing the details of these reports. I encourage people who want to discuss their ancestry composition to do so in the comments.

Without going into details, I would first advise that 23andMe make transparent the way in which 23andMe participants were selected as part of their training data. This is explained in their writeup with the following paragraph:
Most of the reference dataset comes from 23andMe members just like you. When someone tells us that they have four grandparents all born in the same country, and the country isn't a colonial nation like the US, Canada or Australia, they become candidates for inclusion in the reference dataset. We filter out all but one of any set of closely-related people, since they can distort the results. And we remove "outliers," people whose genetic ancestry doesn't seem to match up with their survey answers.
23andMe takes a "birthplace of grandparents" approach rather than an "ethnic origin" approach. This may be reasonable when the two tend to coincide but not appropriate at all when ethnic groups of different origins co-exist in a given territory. Contrary to the implicit belief expressed in the above paragraph, ethnic complexity is not limited to "colonial nations", and an approach that disregards ethnicity, language, and religion, and limits itself to "birthplace of grandparents" is bound to miss it.

The problem with supervised learning is that the end product is only as good as the labels. If the labels aren't good, or they're ambiguous, then you end up with a mess.

Let's take an example of an individual who reports "4 grandparents from Turkey." This may mean anything ranging from a Mesopotamian Kurd within the boundaries of Turkey, a Central Anatolian Turk, a Cappadocian Greek, a Turkocretan, an Armenian from Cilicia, an ethnic Greek from European Turkey, or a Turkish-speaking Muslim from Skopje or Bulgaria. Some of these may interpret "Turkey" geographically; others ethnically. The label "Turkey" is polysemous, for a variety of reasons: it can be interpreted either geographically or ethnically, and in both these senses it has not been time-invariant.

I don't know how 23andMe built their reference populations, but I am ~100% sure that 4 grandparents from Turkey = "Middle Eastern" in their terminology. I am also fairly sure that their "Balkan" sample consists of individuals as different as Croats and Greeks. So what do these meta-population labels mean? Your guess is as good as mine: a balance of samples of different origins and different interpretations of these origins in whatever training set 23andMe assembled.

In my own project, I never include a priori labels of individuals in the inference of ancestral components. I deal with genotypes and individuals, not self-reported ancestral origins and labelled sets of individuals (populations). Components emerge from unsupervised learning over a set of individual genotypes, and it is only a posteriori that labels are assigned to the inferred components, by observation. Indeed, one could forego the assignment of labels altogether!

My amicable advice to 23andMe is to drop supervised learning altogether. It will only get worse as new customers (aka new test data) join in.

100 comments:

shenandoah said...

Is there any way that you can make your Dodecad Project analyzer more user friendly, simpler? It accepts Fasta files, doesn't it? I'm using a public computer which requires administration approval (ie passwords unavailable to me) for downloading things like that. On James Lick's mtDNA analyzer (mitohap) for example, it isn't necessary to download his program software.

pconroy said...

My father has a Greek from Crete as a Relative, and his ancestry painting shows up as:
99.7% Balkan
0.2% Eastern European
0.1% Nonspecific European

Aargau said...

I've sequenced 14 members of my family through 3 generations with the v2/v3 Illumina at 23andme. I also have my exome sequenced and published. We have good genealogical records. This data does not look very accurate to me.

The children are showing small differences in regional classifications that have no derivation from a parent or grandparent.

The amount of non-specific regional data is very high as well.

genefan said...

They did not mention which ethnicities are included in the "Balkan" sample.
Those which are probably used as a reference show like 98-99% Balkan. As I share with a very large number of Balkanians, I can see that Bulgarians, Romanians and Greeks are used as samples. Could not find any ex-Yugoslavia with very high percentage Balkan, don't know if they are included, or just show different. Well, maybe Northern Greeks are close to Macedonians and South Bulgarians and North Bulgarians close to Romanians South of the Carpathian mountains, but no way a Greek from Crete is very close to a Romanian from Transylvania.
The other thing that takes the Balkanians off course is branding Hungarian "East Europeans", while they are maybe in between East, Central and South Europeans. This way a Croat, naturally close to a Hungarian will have big amount of East European, while may be not that close to East European as defined in Dodecad project centered around Belorussia and Lithuania.
Another strange thing is the disappearing of East Asian in many samples and appearing of Sub Saharan not only in the South Balkans, where this could be plausible.
I am obviously used as a reference, too and comparing me and myself I can't be anything but 99% Balkan. This makes me, the other references and those close related to them disadvantaged.
I don't remember who spoke first about the "calculator effect", i.e all should be compared to samples outside of them to have an equal starting point. In this sense the old Global Similarity makes better picture where Balkanians stand compared to each other.

Dienekes said...

I think they have simply reified the population labels. You label something X, you use it as training data, and you get back 100% X for that training sample.

That is why Turks get 100% Middle Eastern and Russians get 100% East_European. They've been used as training samples, so basically Ancestry Composition spits out the labels it was fed. I've seen examples of real Middle Eastern people get less than 100% Middle Eastern, and you have Turks from the Balkans get 100% Middle Eastern.

It will be _very_ interesting to see how this performs on new customers.

fz said...

Comparing to what 23andme used to put up as Ancestry Painting, the new Ancestry Composition is certainly better. However, it still has problems. I had myself, my wife and my son all tested on the same platform. There are still ghost fragment popped up in my son's chromasome painting under the normal condition. However, when one switch to the most speculative mode, the allignment of ancestry segments is actually more acurate.

fz said...

There is a mysterious south asian component popped up in mine and my wife's ancestry composition, which is absence in Dodecad runs.

Fanty said...

Here are my results:

"Standard Estaminate" (Default):
95.6% French&German
2.3% Nonspecific Northern European
2.1% NOnspecific European

"Speculative Estaminate":
99.8% French&German
0.2% Nonspecific Northern European

"Conservative Estaminate":
78.9% French&Geman
16.5% Nonspecific Northern European
4.6% NOnspecific European

"French&German" bases on:
262 Germans from 23andMe
169 Dutch from 23andMe
160 French from 23andMe
73 Swiss from 23andMe
64 Belgians from 23andMe
29 French from HDGP

Dienekes said...

@Fanty

Where are the samples listed?

Fanty said...

"They did not mention which ethnicities are included in the "Balkan" sample."

They do.

"Balkan" bases on:
106 Romanians
102 Greeks
55 Bulgars

"East European" bases on:
181 Poles
169 Russians
68 Hungarians
55 Ukrainians
25 HGDP Russians

"Middle Eastern" bases on:
123 Iranians
96 Turks
47 HGDP Druze

EDIT:
@Dienekes:
One finds the samples by clicking on the INFO Icon behind the Population and then on "details".

Dienekes said...

"Middle Eastern" bases on:
123 Iranians
96 Turks
47 HGDP Druze


That explains why Assyrians end up part-European and Turks from the Balkans end up 100% Middle Eastern.

A classic case of overfitting to the training data.

Anonymous said...

I only accept the Speculative results, not because I like the actual ancestry composition better, but the Conservative and Standard results contain a high "unassigned" percentage. I find that unacceptable considering my ancestry is mostly from one small island for 600 years. I understand that my ethnic group is an isolated population and unrepresented in 23andMe's reference populations, but no one should get high unassigned percentages.

Yes, 23andMe have used two Finnish samples in their Northern European category which means that the East Asian ancestry of Finns and other Northern Europeans are converted to European. You can say the same for Southern Europeans that contain ancestry from West Asia, the Near East and Africa being masked in the European for Iberians, Italians and Balkanians. The results are only correct if you accept existing European ethnic group averages as being wholly composed of that ethnic group which of course they are not.

Onur Dincer said...

In my own project, I never include a priori labels of individuals in the inference of ancestral components. I deal with genotypes and individuals, not self-reported ancestral origins and labelled sets of individuals (populations). Components emerge from unsupervised learning over a set of individual genotypes, and it is only a posteriori that labels are assigned to the inferred components, by observation. Indeed, one could forego the assignment of labels altogether!

My amicable advice to 23andMe is to drop supervised learning altogether. It will only get worse as new customers (aka new test data) join in.


Couldn't agree more. We should let the data speak for themselves. All other approaches are bound to be faulty one way or another.

"Middle Eastern" bases on:
123 Iranians
96 Turks
47 HGDP Druze

That explains why Assyrians end up part-European and Turks from the Balkans end up 100% Middle Eastern.

A classic case of overfitting to the training data.


How can a very large scale genetic testing company such as 23andMe make such grave mistakes?

Westgoth said...

I am Romanian, I fit perfectly to myself :) 99% Balkan... Nice "estaminate" :)

MMaddi said...

I'm DOD021 in the Dodecad Project, with 50% Sicilian and 50% Southern Italian ancestry. Since I filled out the ancestry survey at 23andMe and answered that all 4 of my grandparents were Italian, I think it's probable that I'm included among 23andMe's use of 426 of its customers as an addition to their Italian reference population.

Here are my Ancestry Comparison Standard Estimate results:

99.6% Italian
0.2% Non-specific Southern European
0.2% Non-specific European
0.1% Unassigned

My sister has not filled out the ancestry survey, so probably is not among their customer-based Italian reference sample. Here are her Standard Estimate results:

86.3% Italian
8.6% Non-specific Southern European
0.2% Non-specific Northern European
<0.1% Ashkenazi
4.1% Non-specific European
0.1% Sub-Saharan African
0.6% Unassigned

It looks like, since she's probably not part of the reference sample, her results are a bit more nuanced.

Anonymous said...

I would assume to be around 25 % in the German-French cluster, but is instead nearly 100 % in the Scandinavian cluster.

I wonder if that is because the test is self-referring to the initial survey answers where I answered that all of my grandparents birthplaces were in Scandinavia?

In the Dodecad runs I frequently look Dutch.

MMaddi said...

Regarding the use of Balkan customers at 23andMe for their reference sample, I have an estimated 4th cousin (sharing one 20.5 cM segment) in Relative Finder who tells me that all his ancestors that he knows about were Greek, from mainland Greece (including the town of Metamorphosis in southern Greece) and the Greek colonies along the Black Sea coast. I don't think he's used in their Balkan customer reference sample, since he's not in my Ancestry Finder list, meaning he didn't fill out the ancestry survey.

Here are his Ancestry Composition Standard Estimate results:

21.5% Italian
20.5% Balkan
36.8% Non-specific Southern European
0.8% Eastern European
0.7% Finnish
<0.1% Non-specific Northern European
19.0% Non-specific European
0.1% East Asian
0.6% Unassigned.

I had previously thought that our common ancestor might be one of my ancestors from Bari province, on the Adriatic, who was originally Greek. But looking at his 21.5% Italian and the chromosome view that shows the area of our shared segment on chromosome 9 as colored Italian for him, it seems that our common ancestor was an Italian who married into his Greek ancestry. In fact, both of the pair of his #9 chromosomes are almost entirely painted Italian.

Unknown said...

Will African Americans still get those single digit false "Asian" percentages?

Dienekes said...

It looks like, since she's probably not part of the reference sample, her results are a bit more nuanced.


That is actually a very good example. Remember that your sister shares about half of your genome. So, half her genome is covered by you -who belongs to the training set- and makes up for ~50% Italian. So, in the remaining ~50%, she is 36/50 = 72% Italian.

I am fairly sure that new Italian customers will not get 99% Italian like you did, simply because they are not part of the training set, unless 23andMe manages to get ancestry surveys from everyone and retrains their model continuously (unlikely).

Dienekes said...

I wonder if that is because the test is self-referring to the initial survey answers where I answered that all of my grandparents birthplaces were in Scandinavia?

That is likely the case; the "test" has simply reported your self-reported ancestry because you have been included as a Scandinavian, so the test's definition of "Scandinavian" includes you. If a person had the exact same genotype as you but did NOT fill the ancestry survey, they would not get ~100% Scandinavian.

Dienekes said...

I had previously thought that our common ancestor might be one of my ancestors from Bari province, on the Adriatic, who was originally Greek. But looking at his 21.5% Italian and the chromosome view that shows the area of our shared segment on chromosome 9 as colored Italian for him, it seems that our common ancestor was an Italian who married into his Greek ancestry. In fact, both of the pair of his #9 chromosomes are almost entirely painted Italian.

I wouldn't bet on that.

To see why, consider that if this guy didn't fill up his ancestry survey, or he filled it with 2 grandparents born in Greece and 2 in Turkey, then his sample would NOT have been used as a reference sample.

So, a common segment shared by you and him is used to train the model, and is identified as "Italian" because _you_ filled the survey and he did not.

I am not denying the possibility of Italian gene flow into Greece. But, I can't really see at all how a Cretan Greek would turn out 100% Balkan, and the Greek you mentioned would turn out like a mix of so many influences. The difference is that a Cretan Greek would have 4 grandparents from greece -and so he'd be included in the "Balkan" population-, while the other one did not, so his ancestry is broken up every which way that fits.

MMaddi said...

Dienekes wrote:

"So, a common segment shared by you and him is used to train the model, and is identified as "Italian" because _you_ filled the survey and he did not."

I'm not sure you're correct about that. Looking at the chromosome view for this Greek RF match, besides both pair of #9 being painted almost entirely Italian, both pair of #1 are painted almost entirely Italian, half or more of one of the #2, #3, #8, #14 and #15 chromosomes are painted Italian and about 1/3 of one of the #10 chromsome is painted Italian. That's an awful lot of Italian showing on different chromosomes for there not to be some actual Italian ancestry in his tree.

At the very least, by your explanation, it means that he must be matching with multiple members of the Italian training set. But if he's matching with multiple people with all 4 grandparents born in Italy, that still may indicate that he has actual Italian ancestry in his tree.

mooreisbetter said...

Dienekes, this is one of the most important posts you have ever written. You and I don't always agree on various hypotheses, but here, the facts are incontrovertible. If 23andMe has garbage in, it will spew garbage out. If the baseline data is false, it will throw off everything.

How many people that you know have ridiculous family legends? (I know so many.) Like, grandpa was from Tunisia, but we are pretty sure he was Italian.

How many people were conceived out of wedlock?

How many people simply choose to wrongly self identify? Like in your example...

I am very very cautious of labs claiming to predict with this certainty one's ethnic background.

There used to be another lab who did such tests. They told someone I know that he was part Native American. His parents were born in Switzerland and Sardinia, respectively. When pressed, they basically admitted that their test was bogus, and that most people who paid for it were trying to prove Indian ancestry for gaming purposes. Now 23andme is not that way, but since this I have been very wary of these overhyped tests with their way exaggerated claims of accuracy.

Anonymous said...

Italy is the most genetically diverse European population. Italians have separated into isolated breeding groups for hundreds of years. 23andMe uses its Italian customers as the basis for the Italian reference, and most of those are American of mixed Southern Italian ancestry: Sicilian, Sicilian and various Southern Italian groups and the rest a mix of Southern and Central Italians. Northern Italians are like finding hen's teeth at 23andMe. So the Italian category is not representative of Italians but American Italians. I am not Italian and don't have Italian born ancestry practically to the start of church and official records 500-600 years, yet Italian is my highest ancestry at 23andMe.

This is how the three ancestry estimates breakup for Italian, non specific Southern European and Unassigned. Standard: Italian 56.4%; non specific Southern Euro 14.4%; and unassigned 17.9%. Conservative: Italian 30.8%; n/s S Euro 14.4%; and unassigned 36.7%. Speculative: Italian 73.3%; n/s S Euro 7.5%; and unassigned 1.4%.

Now, those results are rather crude and totally non specific to my ethnic group and true ancestry in the last 500 years. It is useless.

CC Bilgin said...

That explains why ... Turks from the Balkans end up 100% Middle Eastern.

It is true that most non-Balkan Turks that I share with on 23andme turn out as almost full Middle Eastern. However, at the Standard Estimate level, one full and two half-Balkan Turks are shown to be 41%, 74% and 95% European, respectively.

Gui S said...

I now regret filling out my ancestry questions, it is very obvious I was includes in 23andme's training sample.
I had hints from Dodecad and other projects of genealogically unknown ancestry, but now any hope of finding them out has been destroyed.

Sanshou said...

The French/German side of my family is absent on 23andme AC. I come in @ 0.7%... Klass, Graulou, La Salle, and Viruet... All French/German surnames, grand parents and great grandparents. I can't imagine all of their DNA adding up to 0.7%... It just doesn't make sense. Absent from my Puerto Rican side are any traces of North African, 0.0%! I was puzzled so I looked at numerous PR AC breakdowns... The average North African seemed to be 0.1%, even though a significant amount of PR ancestry went thru the Canary Islands.

Dienekes said...

That's an awful lot of Italian showing on different chromosomes for there not to be some actual Italian ancestry in his tree.

What I'm trying to tell you is that this person's segments are a "free-for-all" because he didn't fill in his ancestry survey. If he had put "4 grandparents Greece" in his ancestry survey, the non-Balkan portion of his ancestry would diminish or disappear.

The same thing would happen with you and your sister; 23andMe keeps one sample in a group of relatives, so they kept you (because you filled the survey), so you ended up ~100% Italian and your sister ~86% Italian.

It is true that most non-Balkan Turks that I share with on 23andme turn out as almost full Middle Eastern. However, at the Standard Estimate level, one full and two half-Balkan Turks are shown to be 41%, 74% and 95% European, respectively.

It would be interesting to know what these people filled in their survey.

A good sanity check is with the "Mediterranean" and "West_Asian" components of globe13. I have people (both Greek and Turkish) who have both and fall on the European-Anatolian continuum, and some of them become ~100% European and others ~100% Middle Eastern.

Onur Dincer said...

However, at the Standard Estimate level, one full and two half-Balkan Turks are shown to be 41%, 74% and 95% European, respectively.

Is the full Balkan Turkish sample the one 41% "European" in 23andMe's test? Logic says that he/she is the one 95% "European". Anyway, those results may be more related with whether any of those samples or genetically very similar samples were included in the training set rather than the actual genetics of those samples.

Judy said...

I am an adoptee but found my new 23andMe ancestry results when I use the "speculative" setting to be very close to what I was given for information by NY State regarding the ethnic background of my birth parents, and what I have seen for matches that I have with 23andMe, FTDNA and Ancestry.com (I have tested with all three). I think it works better when your parents have different ethnic ancestry regions such as mine (mother Dutch and Irish, father Italian). My results show a relatively even split between Northern and Southern Europe.

Spartacus said...

I am a Turk from Bulgaria, in my Ancestry Composition my results are shown as 100% European, in the sub-Regional Composion my results are 99,3% Balkan, 0,6% non-specific Europena, < 1% non-specific South-European and < 0,1% Unassigned European. In the speculative mode the results are 100% Balkan. In the old Ancestry Painting my results were 99% European and <1% Asian, of which equally North and South European composition. In the Ancestry Finder I have indicated that my both parents are born in Bulgaria, in fact for the last 4 known generation all my ancestors have been Turks from Bulgaria. I find the results strange as I have "relatvies" in 23andMe from the Balkans (Bulgaria, Romania, Greece and fromer Yugoslavia) but only 2 of my connections (one being another Turk from Bulgaria) are shown as 99% Balkan, the rest are shown as 30 - 60% Balkan. Not sure why the others are with much less Balkan % composition.

MOCKBA said...

MMaddi - are there any non-Italian segments on your sister's chromosomes among the segments which you share / IBD? On a Jewish mailing list, I heard a story of a person with extensive IBD with a Finnish guy who has this shared segment painted "Finnish" (and who is 99.8% Finnish overall). But the problem is, the Jewish person has the same segment painted "unspec European". ... so I wondered if 23andMe's algorithm uses prior probabilities assigned based on your profile. Like if you have all 4 Italian grandparents, then it might bias every score 4-fold in favor of Italianness, but if you don't specify any ancestry then you get an unbiased comparison of scores?

For some of the more distinct and more homogenous countries, like Finland, where they used external public data, the results tend to be fairly interpretable. Like in my case, my great-grandmother was of Pomor extraction, as I may have mentioned before. It's a somewhat heterogeneous ethnic group from the White Sea shores with strong Finnish cultural influences. The Pomor oral tradition maintains that the Southern Pomors are of ethnic Russian stock, while Northern Pomors are heavily Finnish-admixed. My ancestors were from the South shore. Well, almost my entire N European component is Finnish according to the new 23andMe tool, and it totals 1.5% to 3.2% (conservative to speculative) which is consistent with my great-grandma being just over a quarter Finnish.

But there follows an immediate disappointment: the admixture within the Pomors would have to be fairly old, and therefore highly fractured by recombinations. But my Finnish blocks quite long, even in conservative assignment including one comprising 2/3rds of chromosome 11. So the algorithm may have preference for long blocks, and decisions on small margins even for "conservative" approach? If it isn't just a statistical fluke like if this chromosome 11 was relatively homozygously Finnish in my mggm, and was spared recombinations more recently?

On a different topic, now that we learned that Gypsy genes have been strongly influenced by drift, just as Ashkenazi genes were ... when can we get a Roma component in chromosome painting? And since both Ashkenazi and Roma components won't appear prominently as PC components, but are relatively homogenous, just how one can beat supervised training algorithms to assign AJ / Roma segments? Any better ideas?

eurologist said...

I wouldn't say 23andMe's analysis is completely useless or wrong. I think it makes some sense for their US customers or others of mixed origin who'd like to know some rough reference regions. But it is indeed pretty useless for e.g. Europeans. For example, combining France, Belgium and Germany and counting Hungary as Eastern European makes no sense whatsoever from the autosomal analyses published over the years.
Hungary always clusters very closely with Slovenia, Austria, Czech Republic and Slovakia, and even with Germany. I'm certain most people from France/Belgium (outside those east of the arc Dunkirk/ Brussels/ Liege/ Luxemburg/ Colmar/ Mulhouse/ Bern) won't cluster with Germans.

Onur Dincer said...

I find the results strange as I have "relatvies" in 23andMe from the Balkans (Bulgaria, Romania, Greece and fromer Yugoslavia) but only 2 of my connections (one being another Turk from Bulgaria) are shown as 99% Balkan, the rest are shown as 30 - 60% Balkan. Not sure why the others are with much less Balkan % composition.

Those rest of the "relatives" may all be indeed from the Balkans and shown as less "Balkan" just because they were not included in the training set rather than their genetics.

aspromavro said...

I saw a half-Syrian half-Russian with the result of 26% Italian. Makes no sense whatsoever.

MMaddi said...

MMaddi - are there any non-Italian segments on your sister's chromosomes among the segments which you share / IBD? On a Jewish mailing list, I heard a story of a person with extensive IBD with a Finnish guy who has this shared segment painted "Finnish" (and who is 99.8% Finnish overall). But the problem is, the Jewish person has the same segment painted "unspec European". ... so I wondered if 23andMe's algorithm uses prior probabilities assigned based on your profile. Like if you have all 4 Italian grandparents, then it might bias every score 4-fold in favor of Italianness, but if you don't specify any ancestry then you get an unbiased comparison of scores?

I think you've probably identified how Ancestry Composition works in a sibling case like mine, where one sibling is part of the training set and the other is not. The algorithm doesn't know my sister's ancestry, but knows mine since I'm in the training set. If there is a question about what ancestry a segment represents, it defaults to "Italian" for me. But in the case of my sister, it chooses "Ashkenazi."

My sister is assigned one Ashkenazi segment (<0.1%) near the beginning of chr 2. Comparing chr 2 for each of us on gedmatch, using Dodecad k7b, our admixture there is almost identical. It's mostly Atlantic-Baltic, with some significant Southern and West Asian. But since Ancestry Composition considers me 99.6% Italian, that same segment is labelled Italian for me.

Anonymous said...

I'm missing something, I am mostly European with some known native and african ancestry within the last 6 generations. Our adopted daughters are mixed race. The eldest has three grandparents of african ancestry, 1 of European. The youngest has 1 grandparent of native ancestry, 1 of African and 2 European.. yet our percentages are a bit strange:

mine is 104% European
eldest is 140% sub-saharan African
youngest is 100% undetermined

Questions: how does one get over 100% of an ancestry?
Why does the chromosome view show more granular ancestry (which looks roughly accurate based on genealogy) than the % which seems.

zack k said...

Apparently there is no Croatians from Dalmaia in their Balkan reference because this is where my grandfather was from and I get zero Balkan (and 10% undetermined Eastern European). The rest I assume might be some undetermined Southern European.

Anonymous said...

Balkan = Romania, Greece and Bulgaria; all from customers of 23andMe.

Middle Eastern = Iran, Turkey and the Druze from Israel; the Druze are from HGDP and the others are 23andMe customers.

Are those reference groups appropriate or apt for all people from Balkan Europe or the Middle East?

The only people who have benefited from the Ancestry Composition are those who derive from the specific reference groups and the lack of genetic complexity of those reference groups. Northern Europeans lack genetic complexity and what differentiation exists is easy to separate with minor usage of SNPs. The downside is Finns and other Northerners lose their East Asian ancestry and come out 100% European. The same thing happens with Southern Europeans with their minor North African and sub Saharan ancestry provided they are 100% of the reference group used for Southern Europeans.

Valikhan said...

My 23 new Ancestry composition is pretty close to Doug McDonald's interpretation.

Valikhan said...

I'm not surprised of my Ancestry composition results, giving that fact I have RF matches from all those populations I have in Ancestry painting.
http://i49.tinypic.com/153ob9s.png

Davidski said...

It looks like all the Europeans from the AC reference populations who filled out their ancestry info are screwed. 23andMe will mostly paint their chromosomes whatever they scribbled on the form.

But all those too lazy to fill out the forms are getting pretty good results that line up with a lot of the other local ancestry analyses.

Fanty said...

That kind of reminds me of "Physical Anthropologist" who ask for the persons nationality to put their finger to the right phenotype.

Kurti said...

This may mean anything ranging from a Mesopotamian Kurd within the boundaries of Turkey.

A question is in my head.

I cant agree with this designation.

How can someone come to the conclusion to label Kurds from Anatolia as "Mesopotamian Kurds withing the boundaries of Turkey" while labeling Greeks, Armenians and Turks which are recorded to be more recent new comers to Anatolia (as if Anatolia is a united landmass to begin with) compared to the Iranic and Hurrian elements (Mitanni, Medes, Scythians, Alans, Cimmerians).

While the proto-Armenian (Phrygian) are relatively new comers from Balkans via Central Anatolia. Eastern Anatolia is and was always naturally connected via different Empires (the Hurrian, Mitanni, Urartaen all of them stretched from Caucasus and Eastern Anatolia to Mesopotamia and Iran).

Also Eastern Anatolia is geographically obviously more connected to Mesopotamia the Caucasus and Iran as to Western Anatolia or the Balkans.

So using the Greeks and Armenians as a proxy for Anatolian genetic make up is in my opinion more than just historically wrong.

My two cents ;-)

Unknown said...

My mom's parents were from about 100km of each other in the mountains of southern Poland. My dad's grandpa was from somewhere in Poland, and the rest of his ancestors in North America for over 200 years, of Irish, German and English background.

My 23andMe numbers are:
48.1% Eastern European
11.8% French and German
2.5% British and Irish
0.8% Finnish
0.2% Scandinavian
18.6% Nonspecific Northern European
3.9% Balkan
4.8% Nonspecific Southern European
9.0% Nonspecific European
0.3% Sub-Saharan African

Onur Dincer said...

Kurti,

I don't know the situation with Anatolian Greeks (as I have yet to see any full Anatolian Greek genetic sample), but Armenians are surely among the most native Anatolian/Armenian Highlander modern populations. This is clear from their genetic similarity with Anatolian Turks and especially Assyrians. The Assyrian similarity is important here, as Assyrians have very little, if any, known Indo-European or general European potential origin.

Kurti said...

@Onur The genetically Armenians are clearly and visibly distinct to Turks. Armenians lack North European influence which you can find in Turks, on the other hand Armenians are higher on Southwest Asian and West Asian components.

The Armenians and Armenian language are relatively new in Eastern Anatolia compared to Caucasian and Iranic languages. Turks are genetically considerable Balkan influenced, many Turks are in fact either from Western-Central Anatolia or the Balkans, with significant Central Asian input. you are a good example. Genetically you have absolutely little to do with Anatolians but in one generation, even though your children might be only half Anatolian, they will clearly identify as fully Anatolian Turks. And as I already clearly mentioned. Anatolia is and never was one one unified territory. Eastern Anatolia throughout history always interacted with Western Iran, Caucasus and Mesopotamia and borders these areas.

Expecting Eastern Anatolians to be genetically the same as Western Anatolians is as smart as expecting Germans being genetically the same as French. Even close neighbors like Greeks and Albanians are gebetically significant different from each other.
And now consider the geographic distance between East and West Anatolia being quintuple.


Assyrians are as long as Iranics and clearly longer present on Southern and Eastern Anatolia than other elements, They are surely native component of Anatolia yet they have also measurable Levant ancestry. The same goes actually also for Armenians who appear distinct to all their Caucasian, Iranic neighbors in having more Mediterranean component.

If you look it that way, though Iranics being geographically closer to Armenians, Georgians appear genetically closer to them. This lets only one possibility open for me. It almost seems like a group from further West has pushed a gab between Iranics and Georgians. And this group logically appears to be the Armenians.

I would place the Armenian language in the Balkans but most of their ancestry somewhere in Central-South Anatolia, around Cilicia.

Kurti said...

@Onur

I have no idea if you really don't understand the definition of what is native and what isn't. I suspect the reason for this is rather your odium towards Kurds which you haven't hide in several Platform/Forums.


I will try to explain it another time. It doesn't require more than the ability to read, to understand it.

Dienekes has several times proven that Anatolian Turks trace at least 30% of their ancestry to Central Asia, but this doesn't mean that 30% of their ancestry has to be East Asian since Central Asian Turkic groups share at least 50% West Eurasian genes from which most is West Asian.


Western and South-Central Asia was always connected not only by empires and people moved back and forth. This connection is also called Central Asian corridor. The Iranic presents on Anatolia is older as the Armenian, Greek and especially Turkic. If the Iranians brought South Asian component with them this only makes it a native component of Anatolia. Of course it depends on your own definition of ancient. If 2000 bc is not ancient enough for an individual from an ethnicity which arrived 3000 years later, well than I have to "accept" it.

There is a fluent connection between the South Asian and West Asian components just like there is one with the Southwest Asian component too, and I don't believe that Southwest Asian is more native to Anatolia as the South Asian, since South_Central Asia was always connected to Anatolia but the same cant be said about the Southern Levant, Arabian Peninsula. Also someone seems to exaggerate the South Asian influence on Western Iranics like Kurds. Most calculators show around ~4% This is only ~3% more than other Anatolian groups own. This is less as the additional Southwest Asian Assyrians or East Eurasian Turks posses which ranges from 6-10%.

I never argued about Georgians being genetically closer to Iranics as to Armenians. What I said is that considering the geography than Armenians should be closer to Iranians as Georgians or Lezgians for example. But this is not the case. And since we both agree on that Georgians and Lezgians are native, this only brings me to the conclusion that Armenians genetically do not belong to where they live today.


Someone can accept or deny it but history confirms my arguments.

Eastern Anatolia, the Caucasus, Mesopotamia and North- and Western Iran was always the geographic area where Caucasian and Iranic groups met and merged.


1.Hurrians and tribes of Hurrian origin (like the Kassites) populated a territory stretching from Eastern Anatolia, the Levant, Mesopotamia and Western Iran. Later they merged with earlier proto Indo-Iranic tribe and formed the Mitanni kingdom

2. Urartu connected Eastern Anatolia, Mesopotamia, Caucasus and Northwestern Iran. And as we know from Assyrian sources and archaelogic findings they merged with Iranic Cimmerians and Alans.

3.Mannaeans, pred. Caucasian people which populated Southern Caucasus and North-Northwest Iran merged with Medes and Scythians.

4. The Carduchoi living in Southeastern Anatolia were a fusion of Hurrians, Scythians, Medes and Gutians.

5. Median empire connected all the way from Central Asia to Central- and Northeast Anatolia, the same goes for the Achaemenid Empire.



There is not a single empire you can list me here which connected earlier exclusively Western Anatolia or the Balkans with Eastern Anatolia. Beside Actually there is no other beside the Byzantine Empire.


Fact is Eastern Anatolia has historic and geographic connection to Western Iran and Mesopotamia. It was always populated by the same people. So there is no sense if someone comes and tells us Central- Western Anatolians or Balkanians should be more representative for native Eastern Anatolians as Mesopotamians or Western Iranians.


Onur Dincer said...
This comment has been removed by the author.
Onur Dincer said...

Kurti,

I don't write in any forum and I have never written in any anthropology or genetics-related forum (I only write in blogs). Also, no one until this day has noted any negative feeling towards Kurds on my part. On the contrary, I have several times been regarded as a Kurdophile by Kurds and non-Kurds alike (I myself don't claim to be any group X-phile, I am just reporting other people's impressions of me). So I suspect that you are confusing me with someone else.

Dienekes has not proven anything about the extent of the Central Asian ancestry in Anatolian Turks. He several times estimated that Anatolian Turks have on average between 10% and 15% Central Asian ancestry:

http://dienekes.blogspot.com/2005/02/how-turkish-are-anatolians.html
http://dodecad.blogspot.com/2010/11/how-turkish-are-anatolians-revisiting.html
http://dienekes.blogspot.com/2011/05/central-asian-element-in-turks-part-3.html

BTW, Kazakhs, the Kyrgyz and the Altai all have less than 50% West Eurasian ancestry according to ADMIXTURE analyses (Uyghurs sometimes show about 50% West Eurasian and sometimes less West Eurasian ancestry in ADMIXTURE analyses). Also the Caucasoid part of the ancestry of the Central Asian Turkic groups is not uniform, being more stereotypically Iranic-like (not non-Iranic West Asian-like) in the south and Europeanizing towards the north (probably northern Scytho-Sakas were genetically something between Europeans and stereotypic Iranics plus some Mongoloid admixture that increases towards the east).

As for the East Eurasian or Mongoloid ancestry in Turks, it ranges from 0% to about 10% according to ADMIXTURE analyses, the average being somewhere in the middle.

Kurds and Iranic peoples in general seem to have arrived the lands that would later be the traditional lands of the western Iranic languages (e.g., Kurdish, Median, Persian) only during the Iron Age, during the course of the early 1st millennium BCE, and this is the same time as the commonly accepted time of the arrival of Armenians to their present location (the Urartian Highland, which would later be called the Armenian Highland). More importantly, northwestern Iranic peoples (including Kurds/Proto-Kurds) lived predominantly within the borders of the Iranian Plateau during most of their history in West Asia and arrived more western lands of West Asia in more recent times.

The South Asian genetic element in Kurds is too much for Anatolian, the Armenian Highlander, Transcaucasian and Mesopotamian standards and Kurds usually group with Iranians in detailed West Asian genetic analyses. You can ask Palisto, the owner of the KurdishDNA blog, if you don't believe me.

Lastly, Armenians show very little, if any, genetic influence from the Balkans. They are genetically typical northwestern West Asians.

Kurti said...

@Onur sure that you aren't the same Onur which is active on several Forums and has similar views as you?

If so, than I am sorry. It seems I have confused you for a long time with a racist person who traces his ancestry also to Balkans (Bulgaria) and his nickname is Onur too.

I will answer you later, when I got more time.

Onur Dincer said...

Kurds and Iranic peoples in general seem to have arrived the lands that would later be the traditional lands of the western Iranic languages (e.g., Kurdish, Median, Persian) only during the Iron Age, during the course of the early 1st millennium BCE

By "Kurds" I meant Proto-Kurds here, as Kurds probably did not exist as a distinct group that early.

Kurti said...

@Onur

Dienekes has not proven anything about the extent of the Central Asian ancestry in Anatolian Turks. He several times estimated that Anatolian Turks have on average between 10% and 15% Central Asian ancestry:.

Thats still 7-12 % more foreign element as among Kurds (if you consider the South Asian as foreign which I don't and have detailed explained why read my former posts). One of the reasons why most Turkish samples end up in Central and West Anatolia is exactly this East Eurasian influence which pushes them more towards East. If it wasn't for this most Turks would end up somewhere between West Anatolia and the Balkans.



BTW, Kazakhs, the Kyrgyz and the Altai all have less than 50% West Eurasian ancestry according to ADMIXTURE analyses (Uyghurs sometimes show about 50% West Eurasian and sometimes less West Eurasian ancestry in ADMIXTURE analyses).

I couldn't find any Kazakh samples but the Uzbek ones seem to be around 40-50% West Eurasian. This is very close to my estimated 50%. And while moving westward its very obvious that they took more West Eurasian genes.

As for the East Eurasian or Mongoloid ancestry in Turks, it ranges from 0% to about 10% according to ADMIXTURE analyses, the average being somewhere in the middle.

I haven't come across a Turkish individual who scores less than 5% East Eurasian, but of course what haven't can happen. When I look at the Turkish results in Dodecad I see the average East Eurasian around ~6% (Turks and Turkish_D) and ~13% (Turkish_Aydin).

Kurds and Iranic peoples in general seem to have arrived the lands that would later be the traditional lands of the western Iranic languages (e.g., Kurdish, Median, Persian) only during the Iron Age, during the course of the early 1st millennium BCE, and this is the same time as the commonly accepted time of the arrival of Armenians to their present location (the Urartian Highland, which would later be called the Armenian Highland). More importantly, northwestern Iranic peoples (including Kurds/Proto-Kurds) lived predominantly within the borders of the Iranian Plateau during most of their history in West Asia and arrived more western lands of West Asia in more recent times.

This is not true, the Armenian element appeared for the first time 600 BC not earlier. Of course it depends on if you consider Urartu and Mitanni as Armenian though both had linguistically (Caucasian and Iranic languages present) nothing in common with them.

And we werent talking about the Western Iranian element per se or when the ancestors of the Kurds arrived on Anatolia. This is not very important in this case since we were discussing about which element (the Iranic or Armenian and Greek) was earlier present on Anatolia. In this is obviously the Iranic one. About the ancestors of Kurds. linguistically they might be a West Iranic group, but ethnically Northeast Iranic and Hurrian elements are also documented, like the Alan tribal confederation among Kurds or the ergativity in Kurdish dialects which is of Caucasian (Hurrian?) origin and differs it from Persian that lacks it.

Kurti said...

.....
The South Asian genetic element in Kurds is too much for Anatolian, the Armenian Highlander, Transcaucasian and Mesopotamian standards and Kurds usually group with Iranians in detailed West Asian genetic analyses. You can ask Palisto, the owner of the KurdishDNA blog, if you don't believe me.
Kurds are closest to Iranians but they cluster westward from them.
"Grouping" with Iranians which is just located East, geographically closer and borders to Anatolia, does not say much.
Your answer does not refute my arguments in any way. Let me quote my earlier post to make clear why.

Eastern Anatolia, the Caucasus, Mesopotamia and North- and Western Iran was always the geographic area where Caucasian and Iranic groups met and merged.

What is native Anatolian and what isn't is not defined by Armenians, Greeks or Turks since these components are relatively new arrivals compared to the Iranic element. if anything than the lack of South Asian component would make them less Anatolian.

Also I clearly illustrated why the "South Asian genes" argument does not work. If we start to determine the Anatolianess of an ethnicity by the lack or addition of specific components, than Assyrians are not Anatolian because they have considerable higher Southwest Asian or Armenians because of their lower North European, higher Southwest Asian or Turks because of their higher East Eurasian and North European components. this is a very self destructive argument.

And if we consider that the South Asian among Kurds (~4%) is only ~3% more than among other Anatolian groups, while Assyrians have considerably more Southwest Asian (~7%) and less North European(0%!), Turks East Eurasian (~8%), Armenians less North European (~5% less) and more Southwest Asian, than the argument of the "more South Asian genes" is not an argument at all and I consider it more as a joke or trollish (do not take it personal).


I am not denying any other group their Anatolian nativeness, I just feel sometimes that I have to defend the Kurdish nativeness on the lands which is somehow ironic and not questionable.

There is no doubt that Kurds are a native people to Eastern Anatolia, Mesopotamia, Western Iran.

Onur Dincer said...
This comment has been removed by the author.
Onur Dincer said...

Kurti,

As I explained above, I am not that Onur (whoever he is) or any other forumer and I am half Balkan Turk (and not only from Bulgaria in that, half of it being from Greece).

As for the Turkish Mongoloid ancestry range according to ADMIXTURE analyses, see the population portraits of the Turkish samples used in Dienekes' ADMIXTURE runs (Dienekes posted the population portaits of his ADMIXTURE runs in several occasions). You will see that some of them have no or almost no Mongoloid component. BTW, the Dodecad and Eurogenes participant Turkish samples are more representative of the Turkish genetic variation than the Turkish samples of the academic studies, as the latter are from pretty resticted regions (often single provinces) of Turkey while the former are from all over the ethnic Turkish parts of Turkey and the other former Ottoman lands inhabited by ethnic Turks.

BTW, which genetic maps are you referring to specifically?

I have seen ADMIXTURE/STRUCTURE results of all the Central Asian Turkic ethnic groups. Most of them have less than 50% West Eurasian ancestry according to those analyses. Uzbeks and especially Turkmens have very significant recent (the last 1000 years) southern Iranic admixture, so they are not representative of the Central Asian Turkics of 1000 years ago, who were then restricted to more sparsely populated northern regions of Central Asia and overall had less Iranic admixture thus less Caucasoid ancestry than today.

The Mitanni (or more correctly the Mitanni ruling class) seem to be Indo-Aryan rather than Iranic based on their personal and deity names and horse terminology. In any case, they were probably nothing more than a small ruling elite in a land with a Hurrian majority.

Anatolia proper consists of the land west of the Euphrates, so the regions that are today called Eastern Anatolia and Southeastern Anatolia (obviously for political reasons) consist of mostly the Armenian Highland and northern Mesopotamia respectively rather than Anatolia. Kurds/Proto-Kurds or any other Iranic group surely did not inhabit Anatolia in ancient times (in fact, not until the Ottoman times). They may have inhabited parts of the Armenian Highland and Mesopotamia beginning from the Iron Age at the earliest but were at most in the minority in those regions.

Armenians with that name first appear around 600 BCE as a political entity (in the Armenian Highland). But that does not mean that they had not already existed in the Armenian Highland as a people during the preceding several centuries. Most historians today date the beginning of the Armenian existence in the Armenian Highland to several centuries before the 6th century BCE. Some even date it to the Bronze Age, as the Armenian self-designation Hay resembles the name of the Late Bronze Age kingdom of Hayasa.

The South Asian element of Kurds is lower than that of Persians, obviously because Kurds are geographically more western and have more non-Iranic admixture, but still, Kurds' South Asian element is too high compared to all of their western neighbors (Turks, Armenians, Assyrians, Georgians and Arabs) and sharply genetically distinguishes Kurds from them and Kurds are genetically closest to the other Iranic groups from Iran except Balochis. So Kurds' levels of the South Asian element are surely too high for Anatolia, the Armenian Highland, the Transcaucasus and Mesopotamia.

Lastly, can you see anything frivolous or trollish in what I write? I am making a serious discussion here. I don't take what you say about me as personal but as name-calling (unfortunately many people resort to name-calling when they are hard pressed in discussions).

Kurti said...

@Onur

The Mitanni (or more correctly the Mitanni ruling class) seem to be Indo-Aryan rather than Iranic based on their personal and deity names and horse terminology. In any case, they were probably nothing more than a small ruling elite in a land with a Hurrian majority.


Its not clear if the Mitanni were Indo_Aryan or Iranic. Most scientist agree to call them simply Indo-Iranic since their language has characteristics of the non spliten Indo-Iranian. And it actually doesn't matter either. Both Indo-Aryan and Iranic have a relatively recent common root.

And the Mitanni are not the only, or first Indo-Iranic group in this area.

Hubushkia was a kingdom in present day Hakkari which has archeologic connections to Proto-Iranic tribes (Kurgan stelae).

Anatolia proper consists of the land west of the Euphrates,

Possible, but today if someone mentions Anatolia all people automatically think of Eastern Anatolia too.

so the regions that are today called Eastern Anatolia and Southeastern Anatolia (obviously for political reasons) consist of mostly the Armenian Highland and northern Mesopotamia respectively rather than Anatolia.

"Armenian Highlands" is a term introduced by Armenian guides to Europeans who traveled across Asia minor. Historically this area all the way up to Ararat was called by Assyrians "Beth Qardu". Ask some Arameans they will confirm this.


Kurds/Proto-Kurds or any other Iranic group surely did not inhabit Anatolia in ancient times (in fact, not until the Ottoman times). They may have inhabited parts of the Armenian Highland and Mesopotamia beginning from the Iron Age at the earliest but were at most in the minority in those regions.
No you are wrong, see above. Iranic elemets are also recorded during the Urartaen period.


That Kurds came after Ottomans is the general urban believe among Turks, which is not right at all.
Badinian a Kurdish principality in todays Northern Iraq was found by a Kurd who settled 1300 CE from Hakkari. Saladin who was born 1137 in Tigrit, originated from a tribe which comes from present day Armenian (Dwin) which borders Ararat.

In fact Eastern Anatolia was for a long times settled by Parthian people, which after a inner conflict split into a Western and Eastern Parthian empire. The large part of the Western Parthians in fact converted to Christianity and become Armenians. This is very well documented and the main reason for the huge Parthian vocabulary among modern Armenians.


Also as I noted earlier, there is a huge Alan tribal confederation among Kurds. And many of them live in modern day Van and Hakkari area. The arrival of Alans exactly into the same area is historically recorded too.



Armenians with that name first appear around 600 BCE as a political entity (in the Armenian Highland). But that does not mean that they had not already existed in the Armenian Highland as a people during the preceding several centuries.

Let me get that straight, you speculate of a possible presence of Armenians in Eastern Anatolia before 600 BCE even though there is no historical evidence, while at the same time you deny any Kurdish presence before Ottomans because you don't see (doesn't mean there isn't) any evidences for their presence on Anatolia?

Doesn't sound very fair, almost hypocritical.



Kurti said...

Most historians today date the beginning of the Armenian existence in the Armenian Highland to several centuries before the 6th century BCE. Some even date it to the Bronze Age, as the Armenian self-designation Hay resembles the name of the Late Bronze Age kingdom of Hayasa.

Just unimportant hypothesis based on similarities of the two names nothing more. I know many historians who connect the term Karda, used during Sumerian period to describe a mountain people living in Eastern Anatolia, with the modern term Kurds which are also often connected with the mountains.



The South Asian element of Kurds is lower than that of Persians, obviously because Kurds are geographically more western and have more non-Iranic admixture, but still, Kurds' South Asian element is too high compared to all of their western neighbors (Turks, Armenians, Assyrians, Georgians and Arabs) and sharply genetically distinguishes Kurds from them and Kurds are genetically closest to the other Iranic groups from Iran except Balochis. So Kurds' levels of the South Asian element are surely too high for Anatolia, the Armenian Highland, the Transcaucasus and Mesopotamia.

I have already answered this.

Sorry, I am not trying to be rude or too direct but I can't keep a conversation with you if you permanently ignore my arguments just to repeat the same over and over again.

This is why I will quote my earlier post.

And if we consider that the South Asian among Kurds (~4%) is only ~3% more than among other Anatolian groups, while Assyrians have considerably more Southwest Asian (~7%) and less North European(0%!), Turks East Eurasian (~8%), Armenians less North European (~5% less) and more Southwest Asian, than the argument of the "more South Asian genes" is not an argument at all and I consider it more as a joke or trollish (do not take it personal).

the "foreign" elements among Kurds is smaller than the "foreign" elements among Turks, Armenians, Assyrians.

Lastly, can you see anything frivolous or trollish in what I write? I am making a serious discussion here. I don't take what you say about me as personal but as name-calling (unfortunately many people resort to name-calling when they are hard pressed in discussions).

I wished you would more often take up my arguments instead of ignoring them often, since they often are the answer to the question/statements you do thereupon

Onur Dincer said...
This comment has been removed by the author.
Onur Dincer said...

Kurti,

We are using two very different and incompatible definitions of Anatolia (Asia Minor, to be more correct). It should have been clear by now that I do not regard the regions east of the Euphates as parts of Anatolia (so read what I wrote carefully before calling me unfair and hypocrite). That is a Kemalist/Turkish nationalist invention; territories of Turkey east of the Euphates were incorporated into Anatolia by Kemalists/Turkish nationalists during the early decades of the Turkish Republic. Thus by calling those territories as parts of Anatolia you are using Kemalist/Turkish nationalist terminology.

The rest of what you write on history is mostly speculation based on poor historical data.

I should point out that I do not put much stock in the Hayasa-Armenian theory. But Armenian presence in the Armenian Highland during the early Iron Age is certainly not a stretch. For instance, Hittites and Luwians, too, are thought to have arrived in their historical locations centuries before their first appearance in the historical records as a political entity.

Armenia is an at least 2500-year-old historical region. Armenia (as a region name) and the Armenian Highland are almost synonymous terms. Beth Qardu (=Corduene) was not even in the Armenian Highland, it was in northern Mesopotamia. Corduene may have been populated by Proto-Kurds/Kurds, even if not exclusively, already during the Antiquity.

As for your indigenous vs. foreign elements argument, I have already more than adequately replied to it, so won't waste my time on it further.

Onur Dincer said...

Euphates would be Euphrates. Sorry for the typo, which is due to fast writing on my part.

Alexandros said...

I need some help solving a DNA ancestry mystery (at least for me it is..). We are a group of (unrelated) Greek Cypriots originating from all over the island and we have all recently tested with 23andme. The mystery comes with Ancestry composition, where each one of us is getting something in the range of 50-60% Southern European and 20-40% Middle Eastern. This is understandable as migrations from the Levant and Anatolia to Cyprus have been extensive through the centuries. The puzzle comes with the European ancestry though, where we all appear as Italian! In other words, practically all of our European ancestry is assigned to Italy! This applies to each one of us who has tested. Given that the Greek sample is assigned to 'Balkan' in 23andme's reference populations, I would expect that Greek Cypriots would cluster genetically with Greeks more than with Italians. It makes absolutely no historical sense for us to cluster with Italians rather than Greeks. A small number of Italians (Venetians) was present in the island only for one century or so, while Greeks have been continuously present in the island since the 14th century BC. If anything, we speak Greek today and not Italian.. Something that I can think is that there is a specific sub-population of Greeks with which the Greek Cypriots cluster genetically, but these are not included in 23andme's reference samples. On the other hand, some Italians (i.e. Calabrians and/or Sicilians) may cluster well with Greek Cypriots as they may share a common Greek genetic heritage. Could this be the answer here? Any other thoughts?

Dienekes said...

@Anaxagoras, I think the new Ancestry Composition is not very good for non-West Europeans because of (i) the limited training samples, and (ii) systematic effects when one's samples was included in the training set vs. when it was not (not sure if the latter has been fixed).

With respect to your specific question, of "Italian vs. Balkan", I am not very surprised since "Balkan" includes populations with quite dissimilar, but non-trivial levels of "Slavic" or North European admixture from the Balkans. Italians line up better with Cypriots not because they are particularly close, but because they mostly lack the Slavic component found in the Balkans. Anyway, that's my opinion.

Alexandros said...

Thanks Dienekes for your comments on my query. Your explanation sounds logical. It is somewhat dissapointing that 23andme's Ancestry composition is more suited for North Western Europeans, as I really like their interface. I hope they will feed more info into their system and make it more usefull for Southern Europeans as well. I am currently waiting for my Geno 2.0 results, so I am curius to see what my admixture looks there. I will also try FTDNA's Family Finder, given the grat price reduction in their summer sales! I will definately upload these data in your DODECAD interface. If you are interested about Cypriot population genetics, please don't hesitate to contact me at any point..

Onur Dincer said...

I hope they will feed more info into their system and make it more usefull for Southern Europeans as well.

As you are a Cypriot, they should make their genetic analyses more useful for West Asians as well. Because Cypriots are genetically clearly more West Asian (of the northern variety) than Southern European despite the obvious Aegean genetic admixture in Cypriots (Cyprus is not far from the southern Aegean islands and has had close relations with them during the last few millennia).

Alexandros said...

Hi Onur,

On what grounds do you base your claim that Cypriots are more West Asian than Southern European?? I think you 've missed my previous post. I was actually saying that many of us (the Cypriots..) have actually TESTED our genetic admixture with 23andme and FTDNA. Not a single one of us got a larger proportion of West Asian than Southern European ancestry. Actually most of us turned out to be around 60-70% South European, 20-30% West Asian and 0-10% unassigned. It is ok to speculate but the reality is another thing. I can share some of the results here if you like..

Alexandros said...

Following up from my previous comment and just to get things right regarding my island.. Onur mentions that "Cyprus is not far from the southern Aegean islands and has had close relations with them during the last few millennia". The misconception here is that the European admixture in Cyprus is via close contact with the Aegean islands!? Cyprus has been SETTLED NOT VISITED by Greeks from the region of Arcadia (have you ever heard of the ancient Arcado-Cypriot dialect?) since 3 thousand years ago. From the 8th century BC onwards there has been a complete Hellenization of the island and the local population has been culturally (and to a large extent genetically) Greek until today. Since then of course several migrations occurred into and out of the island, but the Greek heritage remained alive and vivid until today. That is among Greek Cypriots. For our compatriots, the Turkish Cypriots, the story is different and much more recent but I will not go into that now..

Onur Dincer said...
This comment has been removed by the author.
Onur Dincer said...

On what grounds do you base your claim that Cypriots are more West Asian than Southern European?? I think you 've missed my previous post. I was actually saying that many of us (the Cypriots..) have actually TESTED our genetic admixture with 23andme and FTDNA. Not a single one of us got a larger proportion of West Asian than Southern European ancestry. Actually most of us turned out to be around 60-70% South European, 20-30% West Asian and 0-10% unassigned. It is ok to speculate but the reality is another thing. I can share some of the results here if you like..

Because 23andMe has a very different definition of West Asian/Middle Eastern and Southern European than usual due to their quite arbitrary grouping of the reference populations (Dienekes dealt with this issue at a separate thread). It is ultimately a matter of semantics. Part of the problem is that West Asia is a genetically quite heterogeneous region (e.g., Asia Minor is genetically closer to most of Europe than to Arabia), that is why we see a northern West Asian and a southern West Asian modal component in ADMIXTURE and STRUCTURE analyses.

Onur mentions that "Cyprus is not far from the southern Aegean islands and has had close relations with them during the last few millennia". The misconception here is that the European admixture in Cyprus is via close contact with the Aegean islands!? Cyprus has been SETTLED NOT VISITED by Greeks from the region of Arcadia (have you ever heard of the ancient Arcado-Cypriot dialect?) since 3 thousand years ago.

By "last few millennia" I was referring to the Hellenization of Cyprus and its later contacts with the western Greek world. The closest Greek dialects to Modern Cypriot Greek are those spoken in the southern Aegean islands; that is why I explicitly referred to the southern Aegean islands.

From the 8th century BC onwards there has been a complete Hellenization of the island and the local population has been culturally (and to a large extent genetically) Greek until today.

You are genetically much more indigenous Cypriot than original Greek.

Alexandros said...

Onur I agree with the majority of your comments but I still do not understand why you try to persuade me and the rest of Greek Cypriots that they are more West Asian than Southern European (Greek)?? Maybe you have studied the Cypriot genome in detail? If yes, I would be very happy to discuss further with you, not in this post though, as our conversation is becoming irrelevant to the initial topic. Your claim of Cypriots being genetically much more 'indigenous Cypriot' than Greek is contradicting the actual facts of DNA testing which show 60-70% European ancestry among Cypriots. Indigenous Cypriots were clearly non-Europeans. They were Levantine people who arrived in Cyprus mainly from modern day Lebanon and Syria around 10,000 years ago. These people had clearly a West Asian genetic profile. Where does our 70% Southern European come from then??

Onur Dincer said...

Anaxagoras,

The answer to your question "why you try to persuade me and the rest of Greek Cypriots that they are more West Asian than Southern European (Greek)" is simple: the genetically closest populations to Cypriots are in West Asia, more specifally in Asia Minor and the eastern lands adjacent to it - namely, Anatolian Greeks, Anatolian Turks, Armenians, Assyrians, Azeris, Kurds and Iranians. Cypriots are genetically closer to these populations than to Balkan Greeks (also than to Balkan Turks) and form genetic clusters with these West Asian populations that exclude Balkan Greeks (and also Balkan Turks).

If you don't believe me, then look at Dienekes' cluster analyses involving Cypriots and the relevant populations such as:

http://dodecad.blogspot.com/2012/01/fastibd-analysis-of-balkanswest-asia.html

http://dodecad.blogspot.com/2012/02/chromopainterfinestructure-analysis-of.html

http://dodecad.blogspot.com/2012/03/fastibd-analysis-of-italybalkansanatoli.html

Also look at Dienekes' and other researchers' ADMIXTURE or STRUCTURE analyses involving Cypriots and the relevant populations (there are so many on the Internet and they can easily be found with a Google search).

Fst distances of Cypriots to other populations also confirm my conclusions.

As for Cypriots' 23andMe ancestry painting results, as Dienekes covered in detail at this thread and also at another thread, the new ancesrty composition feature of 23andMe is highly problematic. For one thing, they apply a supervised analysis approach that groups the reference populations based on the greater region of the world they reside rather than the genetic affinities of the reference populations. So, for instance, they include Turks and the other populations of Asia Minor in the same "Middle Eastern" reference population group as Arabians, which is silly as populations of Asia Minor are genetically closer to most Europeans than to Arabians. Such methodological flaws give rise to implausible results such as those we see in Cypriots' ancestry painting results. A better approach is to put the northern and southern West Asian populations in different reference population groups.

In short, it is clear from genetics that Cypriots have much more pre-Greek Cypriot ancestry than Greek settler ancestry.

Alexandros said...

Onur,

I am glad to see that you support your statements with evidence and it is not just a result of the unfair propaganda going on, trying to persuade the world that Greek Cypriots are not Greek, etc., etc. I am sure both of us don't want to go into that. I consider myself a scientist and not a politician. Your arguments make total sense, however the evidence that you present is extremely weak. Two of the three studies that you are referring to have no Greek Cypriots in their sample (they have Turkish Cypriots and Behar Cypriots). The other study has a single Greek Cypriot. I am not sure if we can derive any valid conclusions from this. Also, in the DODECAD sample of 12 Cypriots indeed the West Asian admixture is higher than the South European one. However, I am not sure of the ethnic origin of these 12 individuals. Are they Greek Cypriots, Turkish Cypriots, Maronite Cypriots, Jews?? In general,I was referring to Greek Cypriots as being primarily Southern European and to a smaller extent West Asian. Also, let me remind you that Greek Cypriots comprise 82% of the total Cypriot population and they are in fact the indigenous inhabitants of the island before the extensive settlement of Cyprus by Turks post 1571. Up to now I have seen the admixture results of 8 unrelated Greek Cypriots and all of them have a larger European than West Asian admixture.

Now going to the point that 23andme's ancestry composition admixture results are flawed, I cannot argue about this. I agree with your point that it is silly to put Anatolian Turks (especially western Anatolians) in the same reference population as Arabs, due to the high European admixture among such Anatolians. This however goes against the argument that due to this, Cypriots appear with a higher European admixture than they should. If anything, Cypriots should have had an even higher European admixture, given the constant exchange of populations between Cyprus and the west Asia Minor during the Byzantine and Ottoman years. In other words, Cypriots with a West Anatolian ethnic background may turn out to be primarily West Asian, where in fact they may have a larger European ancestry admixture. Anyhow, me and a few friends are waiting for our GENO 2.0 and FTDNA results and I will post the admixture here when I have it. Maybe you will be right and these tests show a higher West Asian admixture than South Asian. Maybe not. We will see..

Onur Dincer said...

I am glad to see that you support your statements with evidence and it is not just a result of the unfair propaganda going on, trying to persuade the world that Greek Cypriots are not Greek, etc., etc.

I am not a part of such a propaganda, whatever it is. Evidently, Greekness is not something in the exclussive possession of a certain gene pool, and mainland Greeks are not the sole representatives of Greekness. Already during the Antiquity the Greek identity spread to such a large territory that it became impossible to speak of a Greek racial type.

Your arguments make total sense, however the evidence that you present is extremely weak. Two of the three studies that you are referring to have no Greek Cypriots in their sample (they have Turkish Cypriots and Behar Cypriots). The other study has a single Greek Cypriot. I am not sure if we can derive any valid conclusions from this. Also, in the DODECAD sample of 12 Cypriots indeed the West Asian admixture is higher than the South European one. However, I am not sure of the ethnic origin of these 12 individuals. Are they Greek Cypriots, Turkish Cypriots, Maronite Cypriots, Jews??

The Behar Cypriots are exclussively ethnically Greek Cypriots. I know this because I contacted with Mait Metspalu, one of the lead authors of the Behar et al. study, and talked with him this issue. He told me that the Behar Cypriots are exclusively ethnically Greek Cypriots (he even contacted with the sample collectors and re-verified it). The fact that the Dodecad Greek individual genetically clusters with all of the other Cypriots to the exclusion of all of the non-Cypriot Greeks in Dienekes' cluster analyses confirms him. BTW, the Turkish Cypriots are genetically almost indistinguishable from the Greek Cypriots; their ADMIXTURE results are almost the same and in all the cluster analyses I have seen they cluster together to the exclusion of the non-Cypriots (the one non-Cypriot Dodecad Turkish individual who clusters with the Cypriots has recent Turkish Cypriot ancestry FYI), and this despite the fact that one of the Dodecad Turkish Cypriots has recent Sub-Saharan African ancestry.

In general,I was referring to Greek Cypriots as being primarily Southern European and to a smaller extent West Asian. Also, let me remind you that Greek Cypriots comprise 82% of the total Cypriot population and they are in fact the indigenous inhabitants of the island before the extensive settlement of Cyprus by Turks post 1571.

The vast majority of those settled Turks returned to Asia Minor within a generation (see pg. 283 in Suraiya Faroqhi's book "Towns and Townsmen in Ottoman Anatolia: Trade, Crafts, and Food Production in an Urban Setting, 1520-1650", where on the Turkish colonization of Cyprus it is stated, "a rescript dated several years after the initiation of the project remarks that of 12,000 families or households deported to Cyprus, only a few hundred remained"), so I suggest you not to base your view of the Turkish colonization of Cyprus on the number of people deported to Cyprus. I already told you how extremely genetically similar Greek and Turkish Cypriots are.

As for more on the "Southern Europeanness vs. West Asianness" issue, see my replies below.

Up to now I have seen the admixture results of 8 unrelated Greek Cypriots and all of them have a larger European than West Asian admixture.

Those are 23andMe results. I already told you how methodologically flawed 23andMe's genetic analysis is.

[continued below]

Onur Dincer said...

Now going to the point that 23andme's ancestry composition admixture results are flawed, I cannot argue about this. I agree with your point that it is silly to put Anatolian Turks (especially western Anatolians) in the same reference population as Arabs, due to the high European admixture among such Anatolians. This however goes against the argument that due to this, Cypriots appear with a higher European admixture than they should. If anything, Cypriots should have had an even higher European admixture, given the constant exchange of populations between Cyprus and the west Asia Minor during the Byzantine and Ottoman years. In other words, Cypriots with a West Anatolian ethnic background may turn out to be primarily West Asian, where in fact they may have a larger European ancestry admixture. Anyhow, me and a few friends are waiting for our GENO 2.0 and FTDNA results and I will post the admixture here when I have it. Maybe you will be right and these tests show a higher West Asian admixture than South Asian. Maybe not. We will see..

It is silly to put all of Anatolian Turks in the same reference population group as Arabians (not West Asian Arabs as a whole), and not just Anatolian Turks, but also it is silly to put populations such as Armenians, Assyrians, Azeris, Kurds, Iranians, non-Arabian West Asian Jews, Druze and to some extent even northern Arabs in the same reference population group as Arabians. 23andMe's reference populations are horribly grouped, so I do not take 23andMe's new ancestry painting feature seriously. A much better ancestry painting is that of DNA Tribes' new SNP analysis (I know that DNA Tribes' STR analysis is problematic, but I have not seen any clear methodological flaw in DNA Tribes' new SNP analysis). Here are the results of DNA Tribes' new SNP analysis:

http://www.dnatribes.com/dnatribes-snp-admixture-2013-05-14.pdf

See how there are two components modal in West Asia (one centered in the north and one centered in the south) in the first round of DNA Tribes' new SNP analysis and three (two centered in the north and one centered in the south) in its second round.

More importantly, Dienekes' ADMIXTURE and cluster analyses, too, provide pretty accurate and informative genetic results for Cypriots. I always recommend them.

Finally, I have never stated that the Greek colonization did not have a significant genetic impact in Cyprus. Cyprus was most probably more genetically impacted by the Greek colonization than most of Asia Minor during the history of Greek colonizations (I emphasized this fact in various occasions in the past). This is quite understandable since Cyprus is an island with a relatively small population and geographically close to the southern Aegean islands, also Cyprus was colonized by Greeks pretty early. But, still, modern-day Cypriots (whether Greek or Turkish) have much more pre-Greek Cypriot ancestry than Greek settler ancestry.

Onur Dincer said...

This is quite understandable since Cyprus is an island with a relatively small population and geographically close to the southern Aegean islands, also Cyprus was colonized by Greeks pretty early.


I should also add to these the fact that Greek spoken by Cypriot Greeks groups with (though not the same as) the Greek dialects of the southern Aegean islands rather than the Anatolian Greek dialects. The Anatolian Greek dialects are more eastern type of Greek dialects with one exception: Greek of the western Anatolian coasts, which is essentialy the same as the Greek dialects of the Aegean islands due to the fact that the western Anatolian coasts experienced a significant Greek recolonization from the Aegean islands beginning from the 17th century because of the then current Ottoman policy of cranking up the maritime trade and production in the western Anatolian coasts by stimulating Greek immigration to there from the Aegean islands.

Onur Dincer said...

To summarize, according to genetic analyses, Greek Cypriots and Turkish Cypriots are genetically almost indistinguishable from each other; the two populations have almost the same results in ADMIXTURE analyses and cluster together to the exclusion of non-Cypriots in cluster analyses. So we can quite confidently talk of a Cypriot genetic cluster that is exclusive to Greek and Turkish Cypriots. What are the characteristics of this genetic cluster, i.e., Cypriots? They are primarily genetically a West Asian population (more from the northern West Asian variety than from the southern West Asian variety) but have a Southern European minority genetic element that is elevated compared to the West Asian standards, which is a strong sign of some elevated level of mainland Greek ancestry. Nonetheless, the first populations Cypriots genetically cluster with are all West Asian populations and the populations most similar to them in ADMIXTURE analyses are all, again, West Asian populations, not to mention the fact that the closest populations to them Fst-wise are all West Asian populations. But certainly not all West Asian populations are genetically closer to Cypriots than European populations are, because West Asia has a strong north-south dichotomy (the north strongly leans towards Europe) and thus is not a genetically unified region. So if a genetic analysis has a single West Asian reference population group (that is what we see in 23andMe), it is garbage, at least for genetically primarily West Asian populations such as Cypriots.

Alexandros said...

The propaganda I was referring to started during the Ottoman occupation of Cyprus and continued during the British occupation and the aim was to persuade the Greek Cypriots that they are not really Greeks so they should stop demanding unification with mainland Greece as the Cretans and all other islanders did. I really do not believe that you are a part of this and I can accept that you have a natural interest in determining the genetic admixture of Greek Cypriots, however initially your arguments sounded like that. Regarding ‘Greekness’, what do you mean exactly by saying ‘the mainland Greeks are not the sole representatives of Greekness’?? Again, you are along the lines of another propaganda this time trying to prove that modern day Greeks have little in common, genetically, with ancient Greeks. Also, I do not understand your argument that the ‘Greek identity spread to such a large territory that it became impossible to speak of a Greek racial type’? There were Empires that spread much much more than the Greeks (Persians, Romans, Mongols, Ottomans, British, etc.). The Greeks always lived where modern day Greece is, as well as in the shores of Asia Minor, southern parts of Italy (Magna Grecia) and in the shores of the Black sea (hence a substantial Greek admixture in those populations). This was the historical homeland of the Greeks. I do not see a great spread there. The only great spread was achieved with the expedition of Alexander the Great, which lasted only for a few decades before it collapsed and then things returned back to ‘normal’ (i.e. Greeks concentrated in the limited geographical region I mentioned above, as well as north Egypt and Cyprus). The genetic identity of ancient Greeks is pretty much known. The indigenous population was most likely mainly of Y-haplogroup J2b, EV-13, G2a and probably some I and T. The first settlers, who actually initiated today’s Greek civilization and in fact the concept of the ‘Greek’ ethnic identity, where the Mycenaeans, who based on archaeological and linguistic evidence came from the Ukrainian steppes (ca. 1600 BC) and were most likely primarily R1a people. The next great migration into the area of modern day Greece was from the Dorians who again based on archaeological and linguistic evidence came from northern Europe (somewhat western this time) and were most likely primarily R1b people. After the Dorian invasion and especially after the era of Alexander the Great, when the Greek tribes united for the first time, we have the formation of the Greek ethnic identity (both cultural and genetic). If you notice, the paternal haplogroup composition of ancient Greeks I am describing above is remarkably similar to the haplogroup composition of modern day Greeks. And you know why? Because there was no major spread of Greeks as you suggest. Instead, there was always a nation living in the geographical area of Greece through the centuries with very limited recent admixture from outside (the major one being that of the Slavs). Greeks are as heterogenous genetically as any other European nation (if we exclude maybe the Iberians and the British).

Alexandros said...

Regarding the 'Behar Cypriots', what needs to be clarified here is that you speak of ‘Behar Cypriots’ as if they are a specific ethnic group (i.e. Behar..). Maybe I misunderstood it like that initially. So what you are saying is that the ‘Behar Cypriots’ are just the Cypriots included in the study by Behar et al. Ok… If you have contacted the author and these participants are all Greek Cypriots, I cannot argue about that. The main issue here though is that we are talking about very small sample sizes. Even the n=12 sample in DODECAD is extremely small to derive any valid conclusions about the genetic admixture of such a genetically heterogenous population, as the Cypriots. For example, based on the very few studies that were done among Cypriots looking at Y-haplogroup composition, the prevalence of haplogroup R1a was found to be around 3%. From the results that I saw from 11 Greek Cypriots who have tested their Y-haplogroup with FTDNA, 3 of 11 were haplogroup R1a (including me). This indicates how a small sample can give rise to erroneous/biased estimates. The point here is that unless we have a big study with a representative sample from all over Cyprus, we cannot conclude the genetic admixture of this population for sure. So instead of making claims that Greek Cypriots are more West Asian than European maybe you could help in determining this for sure? You seem so passionate about Cypriot genetics that I am starting to think that you are as well a Cypriot. Unless you are so keen to demonstrate that a population has this and that admixture for other populations and not just the Cypriots. As I told you, I have no problem accepting that Greek Cypriots are more West Asian than they are European, but I need to see some hard evidence (large representative samples). All I was saying in my initial post was that from 8 tested Greek Cypriot samples in 23andme, all had a larger European admixture than West Asian. If you believe that these are spurious results generated by the flaws of 23andme, then I cannot convince you about the opposite. However, in contrast to you, I do not go round announcing that Cypriots are more European than West Asian or whatever. What I do is try to discuss it with people who are really into this field (like Dienekes) and maybe try to have some more people tested to clarify the situation.

Alexandros said...

Now going to the Turkish Cypriots. Do you really claim that the 120,000 or so Turkish Cypriots (I am using the 1974 estimate before the Turkish invasion of the island) are descendents from a few hundreds?? This is mathematically and biologically implausible. Also, during the Ottoman rule of Cyprus there has been an unprecedented architectural, linguistic and generally cultural identity shift in the island. For example, almost all major Christian churches were turned into Mosques. Was all of that fuss just to please a few hundred individuals? Indeed many 16th century settlers returned back to Turkey, but many more arrived after that. In addition, many western European travelers who visited Cyprus throughout the Ottoman rule were speaking of a large Greek community and an almost as large Turkish community. I do not think that this was just their imagination. I know the history of Cyprus from inside out and of course I know that genetically Greek Cypriots and Turkish Cypriots are very similar, mainly for 2 reasons: (1) indigenous Cypriots and Anatolians are genetically quite similar, (2) some Greek Cypriots converted to Islam due to the reduced taxation imposed on Muslims in comparison to the Christians and these people gradually assimilated to the Turkish Cypriot community. On the other hand, marriage between individuals from the 2 communities was extremely rare due to strict religious barriers. In addition, there are Turkish Cypriots that have ancestors from deep inside Eastern Turkey, some even from Turkmenistan (settlers were broad from there as well). I have Turkish Cypriot friends who have ancestry from such regions. Clearly, these individuals would not cluster with Greek Cypriots, as they would have a much higher Central Asian admixture. The point again here is that with such small sample sizes, you risk recruiting by chance individuals from the 2 communities who may be extremely related or on the other hand completely unrelated. Therefore, you cannot derive any sound conclusions unless you have larger sample sizes.

Finally, It is interesting what you say about the ‘DNA tribes’ ancestry analysis. I was reading some time ago that it was quite flawed, but maybe they have improved it since then. Maybe I should give it a go..

Onur Dincer said...

Anaxagoras,

I am speaking based on hard data. You, on the other hand, are making speculations about the haplogroup composition of historical peoples without recouse to any solid evidence and making inferences about peoples' autosomal genetics solely based on "family lore". Besides, neither modern-day Greeks nor Greek Cypriots are hign in R1a or R1b, this is especially true for Greek Cypriots, whose frquency of the totality of the Y-DNA R haplogroups does not even exceed 15% (see Eupedia).

Autosomal analyses does not require high sample sizes, and most of the time a sample size of 10 is enough to give detailed information about the autosomal genetic variation of a population, especially of small and genetically homogeneous populations such as Cypriots (the Greek Cypriot sample of the Behar et al. paper consists of 12 individuals, which is more than enough for Greek Cypriots). The fact that all the Behar Greek Cypriot individuals, the Dodecad Greek Cypriot individual and all the Dodecad Turkish Cypriot individuals have all very similar ADMIXTURE results and all cluster together to the exclusion of the non-Cypriot populations in cluster analyses proves conclusively that all those Cypriot samples (whether Greek or Turkish) are representative of Greek and Turkish Cypriots (this is especially true for Greek Cypriots since there are more publically available individuals for genetic analysis for Greek Cypriots than those for Turkish Cypriots).

I did not say Turkish Cypriots are exclusively descended from the few hundred remaining households. It should have been clear by now that what I am saying is that the bulk of the ancestry of Turkish Cypriots comes from Greek Cypriot converts to Islam. As I have been demonstrating in my posts, genetics and history confirm this. The Dodecad Turkish Cypriots have very small levels of Mongoloid genetic components and are very similar to the Greek Cypriots in all other genetic components (the relatively elevated average Negroid component level of the Dodecad Turkish Cypriot sample is all because of a single Dodecad Turkish Cypriot individual who has recent Sub-Saharan ancestry). And, as I said many times, the Greek and Turkish Cypriots cluster together to the exclusion of the non-Cypriot populations in cluster analyses. So we can quite confidently say that the genetically closest population to Greek Cypriots are Turkish Cypriots and vice versa.

Your view of Greekness and the Greek history is nationalist, biased and romantic rather than factual, objective and rational. You do not even mention the fact that most of Asia Minor was eventually Grecified (it took the early Middle Ages for its many regions). So when the Turkmens came, most of the population of Asia Minor was Greek (the rest being mainly Armenian). I am not even mentioning the historically vibrant Greek communities of Syria. I already explained to you that Cyprus was most probably more genetically impacted from Greek colonization than most of Asia Minor was (genetics seems to confirm this). But don't expect me to accept your unfounded claim that Cypriots are more Southern European in origin than West Asian or more Greek settler in origin than pre-Greek Cypriot in origin. As Dienekes and I explained, 23andMe is weak when it comes to the populations of West Asia and environs.

I have said enough. If you have no new information to add, then it is best to not continue this discussion.

Onur Dincer said...

Lastly, as for my ethnic identity, I am a Turk from Turkey (Anatolian Turk on the paternal side and Balkan Turk on the maternal side) and have no familial connection to Cyprus. I am making this explanation just to satisfy your curiosity, as I do not allow my ethnic identity to influence my views.

Alexandros said...

Onur,

Thanks for sharing your ethnic identity. This makes things clearer (i.e. why you are so interested on determining the true genetic admixture of Cypriots).

For me the issue is closed. You believe that a sample of n=10 is enough to derive the genetic admixture of a nation of over 1 million inhabitants. I do not believe that, but I also do not have the time and energy to persuade you about the opposite. Just to say that your claim is like saying that an election poll of 10 individuals can give you a clear picture of the outcome of an election.

I never said that R1a and R1b are frequent in Greece and Cyprus. In Cyprus especially these are quite rare as you rightly say. The point I was making is that when you take a small sample of people you are risking deriving biased conclusions (this is called sampling error). For example, in the 11 unrelated Greek Cypriot samples that I have access to, the prevalence of R1a is 27%. In the few Haplogroup studies done in Cyprus, using larger sample sizes, the prevalence is around 3%. Such misleading results can be derived also for admixture analysis using 10 or 12 individuals.

Alexandros said...

Something else that I cannot convince you about is that you are 100% certain that Cypriots are primarily a West Asian population and that you consider 23andme’s analysis as completely useless. Fair enough. Time will show. I personally will be working on clarifying the admixture composition of Greek Cypriots and as I said before, I will be posting the results on this blog. I do not claim that Greek Cypriots are more South European than West Asian. I was just saying that 8 out of 8 Greek Cypriot samples in 23andme turn out to have a larger Southern European ancestry than West Asian. Obviously, this fact (not my opinion) annoyed you so much that you are willing to write day after day pages and pages about how much West Asian and how little European the Cypriots are. If more tests are done and we turn out to be more West Asian than European I have absolutely no problem accepting it, because I am not a nationalist as you accuse me. I do not understand how claiming that modern day Greeks are in fact ancestors of ancient Greeks makes me a nationalist. If an Italian says that Italians are the ancestors of ancient Romans will you also call him a nationalist? You claim that modern day Cypriots are descendants of the indigenous population that lived in the island 10,000 years ago (which I agree at some extent), but you find it extremely difficult to accept that modern day Greeks are the direct descendants of ancient Greeks, who lived there 2,500 years ago. If this is not anti-Greek nationalism from your site, I do not know what it is. And to finish with this, the fact that there is Greek admixture in Anatolia and Syria, does not make modern day Greeks less Greek. There is Roman admixture in England. Does this make the Italians less Roman??

Grandal said...

Sorry to pop in, but this Onur-Anaxagoras discussion is getting a bit out of hand..

This post is dedicated to discussing 23andme’s ancestry data and not on determining the ancestry composition of a specific nation.

@Onur
23andme’s Ancestry Composition may be somewhat problematic in determining sub-regional admixture, but it can definitely distinguish European from non-European (eg. West Asian) ancestry! It is somewhat obscured to me why you are being so stubborn trying to convince another person about his genetic identity. If he believes that he and his compatriots are more European than Asian, then let it be. The way I see it, as an outsider, is like trying to persuade a Sicilian Italian that genetically he is more indigenous Sicilian than Italian. Even if this is true, you have to acknowledge and respect that ethnic identity is a sensitive issue to many people and insisting in proving your point on someone else’s genetic identity could be offending to some and will definitely not lead anywhere.

@Anaxagoras
From my experience, 23andme’s subregional admixture is indeed somewhat problematic so do not rely too much on it. If you have access to 11 FTDNA accounts as you say, try to see how many of those Cypriots have Y-DNA and/or mt-DNA matches with mainland Greeks. Even a -1 or -2 marker match can give an indication of very ancient common ancestry. This may help you in your quest for determining the genetic link between mainland Greeks and Greek Cypriots.

Dienekes said...

23andme’s Ancestry Composition may be somewhat problematic in determining sub-regional admixture, but it can definitely distinguish European from non-European (eg. West Asian) ancestry!

I haven't seen recent reports on this, but it did not distinguish these correctly at the time I wrote this post.

In general, I would strongly advise against resolving issues of "European-ness" or "Asian-ness" using 23andMe's ancestry composition.

Onur Dincer said...

Thanks for sharing your ethnic identity. This makes things clearer (i.e. why you are so interested on determining the true genetic admixture of Cypriots).

My ethnic identity is irrelevant as long as I stick to the facts, and that is what I have been doing so far.

For me the issue is closed. You believe that a sample of n=10 is enough to derive the genetic admixture of a nation of over 1 million inhabitants. I do not believe that, but I also do not have the time and energy to persuade you about the opposite. Just to say that your claim is like saying that an election poll of 10 individuals can give you a clear picture of the outcome of an election.

You obviously do not understand genetics and simple maths. If you analyze more than 10 unrelated individuals from a homogeneous and relatively small population and all of them show very similar ADMIXTURE results and tightly cluster together in cluster analyses, then chances of those individuals being genetically unrepresentative of that population is extremely low.

I never said that R1a and R1b are frequent in Greece and Cyprus. In Cyprus especially these are quite rare as you rightly say.

Then I misunderstood you. Sorry for that.

The point I was making is that when you take a small sample of people you are risking deriving biased conclusions (this is called sampling error). For example, in the 11 unrelated Greek Cypriot samples that I have access to, the prevalence of R1a is 27%. In the few Haplogroup studies done in Cyprus, using larger sample sizes, the prevalence is around 3%. Such misleading results can be derived also for admixture analysis using 10 or 12 individuals.

Autosomal studies require much less sample sizes than haplogroup studies to be representative of populations.

Something else that I cannot convince you about is that you are 100% certain that Cypriots are primarily a West Asian population and that you consider 23andme’s analysis as completely useless. Fair enough. Time will show.

As Dienekes and I have been telling you, 23andMe is very inadequate in distinguishing between "South Europeanness" and "West Asianness" and estimating their levels. But you are free to believe whatever you wish, of course.

I was just saying that 8 out of 8 Greek Cypriot samples in 23andme turn out to have a larger Southern European ancestry than West Asian. Obviously, this fact (not my opinion) annoyed you so much that you are willing to write day after day pages and pages about how much West Asian and how little European the Cypriots are.

If you made a clarification by saying "according to 23andMe" rather than treating it as a fact, I would not need to debate with you so long. I did not say anything different from Dienekes regarding the shortcomings of 23andMe.

If more tests are done and we turn out to be more West Asian than European I have absolutely no problem accepting it, because I am not a nationalist as you accuse me.

Nationalism is not a defect or crime. I have some nationalist friends. They are not so bad people. Nationalism becomes really a problem when it begins to distort one's sense of reality.

Onur Dincer said...

I do not understand how claiming that modern day Greeks are in fact ancestors of ancient Greeks makes me a nationalist. If an Italian says that Italians are the ancestors of ancient Romans will you also call him a nationalist?

Do you actually read what you write? What does it mean "modern Greeks are ancestors of ancient Greeks" and "modern Italians are ancestors of Romans"?! Did you mean "descendant" rather than "ancestor"? If so, read below.

You claim that modern day Cypriots are descendants of the indigenous population that lived in the island 10,000 years ago (which I agree at some extent)

Genetics of modern-day Cypriots strongly implies that they are largely descended from pre-Greek Cypriots. But I did not go that far as 10,000 years. I am writing about the historical times.

but you find it extremely difficult to accept that modern day Greeks are the direct descendants of ancient Greeks, who lived there 2,500 years ago. If this is not anti-Greek nationalism from your site, I do not know what it is.

Do not put words in my mouth. I have never claimed such a thing. Mainland Greeks probably have not genetically changed much since ancient times and Cypriots probably have not genetically changed much since ancient times either. It is you who claims that Cypriots genetically changed much during the Greek colonization.

And to finish with this, the fact that there is Greek admixture in Anatolia and Syria, does not make modern day Greeks less Greek. There is Roman admixture in England. Does this make the Italians less Roman??

As I have made it clear, what I have been writing has nothing to do with Greekness, as, unlike you, I do not define Greekness based on blood. So you are no less Greek than a Mainland Greek irrespective of how much pre-Greek Cypriot blood you have.

Onur Dincer said...

I think Dienekes has given the adequate answer to Grandal, so I see no need to respond to him.

Alexandros said...

Onur,

It doesn't seem that your ethnic identity is irrelevant, otherwise you wouldn't bother spending a whole week trying to prove the ethnic identity of a nation, other than yours. You do stick to the facts, but only to the facts that are along the lines of your preconceptions.

Thanks for disputing my knowledge. It is so kind of you. I do understand genetics and complex (not just simple) maths, because it is part of my job. I am a genetic epidemiologist. What about you? Does your advance training in statistics and genetics tell you that s sample of 10 from a nation comprising of 5 distinct ethnic groups (Greeks, Turks, Maronites, Armenians, Latins - the latter are the ancestors of Franks and Venetians) is representative of the whole population?? And to answer further your next ironic comment, yes I do know that autosomal analyses require smaller sample sizes but still 10 individuals are too few and cannot be representative. Even in the 8 23andme Greek Cypriot samples that I saw, South European ancestry ranges from 50% to 75%. Even if we accept that the ancestry composition of 23andme is completely useless in distinguishing south Europeans from west Asians, the fact that there is a considerable variation in the admixture composition of these 8 Greek Cypriots, proves that there is genetic variation among Cypriots. Unless you believe that 23andme's ancestry composition is entirely a random guess. You put your comments in line with Dienekes’. Obviously these have nothing in common, as he always gives polite and constructive comments in contrast to you who try to impose your opinion to the other person, usually by being ironic and arrogant.

If you go back to my initial post you will see clearly that I was ASKING about the specific ancestry composition results and I never took them as a fact regarding the ancestry composition of my country. It is you who claim that you know FOR SURE that Cypriots are a west Asian nation with little European admixture. I never claimed that the Greeks have replaced the indigenous population of Cyprus as you suggested. On the contrary, I always believed that we are a mixture of west Asian and European ancestry. Our main disagreement that caused all this debate is that you are entirely certain, based on the existing data, that the west Asian ancestry is much larger than the European, while I prefer to wait a bit more for more evidence to accumulate in order to derive a conclusion.

Alexandros said...

Do you actually read what you write?

Yes I do read what I write. I wrote 'ancestor' instead of 'descendant' by mistake. Of course, being as ironic as you are, you have highlighted this and published it. Thanks! Much appreciated.. At least we agree on the fact that modern day Greeks are genetically linked to ancient Greeks and on the fact that ethnic identity has nothing to do with genetics.

Genetics of modern-day Cypriots strongly implies that they are largely descended from pre-Greek Cypriots. But I did not go that far as 10,000 years. I am writing about the historical times

Indigenous Cypriots arrived in the island mainly during prehistoric times. The main extensive migrations during historic times were from Greeks, Phoenicians and Jews.

Onur Dincer said...
This comment has been removed by the author.
Onur Dincer said...

Anaxagoras,

The only relevancy of my ethnic identity in this discussion is the fact that I am, quite naturally, more interested in knowing about the genetics of my ethnic group and its neighbors. I tried to end this discussion as soon as I felt that it began to heat up, and I have been trying to end this discussion since then. As I made it clear, the 12 Cypriots of the Behar et al. study are exclusively Greek Cypriots and from no other ethnic group. I have seen enough of their and other Greek Cypriots' genetic results to make reliable inferences about Greek Cypriots' genetic affiliations. I have never tried to impose you anything. You are free to believe whatever you wish. But that should never make you immune to criticism. Despite that, I have directed most of my criticism at 23andMe rather than you.

I did not say you claimed a complete or almost complete replacement of the indigenous population of Cyprus by the Greek colonizers. Again, you are putting words in my mouth.

I said the West Asian ancestry is larger than Southern European ancestry in modern-day Cypriots, but did not clarify how much larger. I also said that modern-day Cypriots have much more pre-Greek ancestry than Greek settler ancestry. Note that pre-Greek Cypriot does not mean exclusively West Asian, as pre-Greek Cypriots already must have had some level of Southern European minority genetic element, however smaller than that of post-Greek-colonization Cypriots.

Lastly, when I mentioned pre-Greek Cypriots in the context of historical times I was referring just to their existence in Cyprus, not to their arrival there. Surely, in early historical times Greeks and Phoenicians had not arrived in Cyprus yet.

Alexandros said...

Onur,

Thank for your much more constructive comment this time. I agree that as the discussion was becoming somewhat too personal, it has to end at some point.

In the end, I believe we agree on the majority of things. I went again through the Behar et al paper and indeed in the supplements it says that these individuals were Greek Cypriots. So apologies if I initially disputed that. There is also another paper by Haber et al (2013) which is using (from what I understand) the same sample of individuals. Of course I acknowledge that both of these articles are of high quality and that both show a higher west Asian than European admixture among this sample of Cypriots. What I had in mind and I kept questioning the representativeness of these 12 individuals, is that there is historical evidence of some clustering, within Cyprus, of people with a higher European admixture than the Cypriot average. Two of such regions are some villages in the Famagusta region (major port during Frankish rule) and some isolated villages up in the Troodos mountains. Initially, I was thinking that these were just urban (or maybe rural..) legends, but after testing 3 individuals from the Troodos region (one of them is me) all were found to be of Y-DNA haplogroup R1a-M458, which is extremely surprising!

Anyway, overall, if we disregard the 23andme data, which I accept that are somewhat flawed, indeed the evidence so far points to a very strong Levantine admixture among Cypriots.

Onur Dincer said...
This comment has been removed by the author.
Onur Dincer said...

Anaxagoras,

Thank you for your kind remarks. I am glad to see that there are so many things we can agree on when we look at all the available data. I admit that some of my statements to you were a bit harsh. Also I think I misjudged some of your views. I apologize for these.

Did all of those 3 Troodos Cypriots get their autosomes tested? If so, I suggest you to send their genotype files to Dienekes for analysis in the Dodecad Project. Unless they are relatives, I guess he will admit all of them despite the closed status of submissions to the Dodecad Project, as they are all from a Greek population, i.e., Greek Cypriots.

Levitylab said...

I was surprised how much non-specific showed up in my three modes. For Conservative, Standard, and Speculative, respectively, it was 98.5%, 91.5%, and and 36.8%. My background is 3/4 British Isles (all 4 countries), and 1/4 German.

Conservative:

0.9% British & Irish
<0.1% Scandinavian
51.8% Non-spec. N. Euro.
46.7% Non-spec. Euro.
0.7% Unassigned

Standard:

6.8% British & Irish
1.0% Scandinavian
0.4% French & German
76.4% Non-spec. N. Euro.
15.1% Non-spec. Euro.
0.4% Unassigned

Speculative:

47.3% British & Irish
11.4% French & German
2.7% Scandinavian
33.6% Non-spec. N. Euro.
1.7% E. Euro.
1.1% Non-spec. S. Euro.
2.1% Non-spec. Euro.
<0.1% Sub-Saharan African
0.1% Unassigned

Anai Rhoads said...

I did the 23andme and received my results in less than two weeks after processing. I am a product of 100% Greek parents, grandparents - going back many, many generations (according to our extensive family tree). Yet, 23andme lists me as Italian. 44.4% to be exact. I am a woman, so this is this figure solely reflects my maternal side. I, too, and unclear how they came to this conclusion and how they map regions like this.

Yalchin said...

Onur, I read through the thread above, and felt like saying thank you for your excellent grasp of genetics as well as politics, sociology and nationalist ideologies in general. Not to mention commendable tact. :-)

Best regards

Yalchin

Gypsy Woman said...

I recently joined 23andme and my results are quite surprising. The one thing that I cannot reconcile is the difference in ancestry assignment between the conservative and standard view.
Conservative view: (90% confidence Interval)
61% European (of that 59% unassigned)
1.6% subsaharan african
.9% east asian/Native American
36.5% UNASSIGNED
HMMM…That's a lot of DNA not assigned to a specific group.
Now for the Standard..this is where it get interesting:
70.8% European Consisting of : 16.9% northern(.7% French/German and 16.2 unassigned northern), 13.4% southern European (3.1 Italian, 1.3 Iberian, 8.9% nonspecific southern), 40.7% non specific European.
19.3% SubSaharan African consisting of: 14% West African, 5.3% non specific.
1.7% East Asian/Native American consisting of: .8% Native and 1% non specific asian.
1.1% Middle Eastern consisting of: .9% Middle Eastern, .2 north african.
7.1% UNassigned to any group.

I know that the Confidence interval is 75% for standard and 90% for conservative. I took a stats class before and the one thing that definitely learned is that anything under a 90 confidence interval is not reliable. Yet, some people feel that the standard view is pretty accurate.

What I do know about my ancestry for sure (I have the genealogy to prove it) is that my father's mother is 100% Southern Italian and his father is a mix of English and German ancestry.
My mother is from Quebec (French speaking) Canada, but her ethnic background is a mystery (although her birth records report Native American and Indian). That's why I took the DNA test, to confirm what her birth records report.

I would like an opinion on whether or not I can conclude that one of her parents was probably black or at least mixed with black. Also, can I rule out that her mother was not Native (my mtdna is X2b4).

If the conservative view is correct, my mother is probably mixed Black and White (French considering she was born in a French speaking Quebec area). So wouldn't I show more French?
Also, my mother does not look French or Black; she is often mistaken for Latina or Mexican.

Ciao,
Ashley

Unknown said...

@Gypsy Woman, Speculative does not = not accurate. Once you go "sub-regional" I would take it with a grain of salt. The rule of thumb, if it does show in conservative and standard, the assignment is correct if showing more in speculative, at the basic level.

For most people of predominately Sub-Saharan African ancestry, the estimate is much smaller in "conservative" and "standard." There are people 90% SSA, while in "conservative" are 20% SSA, appear in "standard" as 60% and then "Speculative" as 90% SSA.
Sub-Saharan African alleles are very unique and easily detectable unlike non-SSA alleles.

Also, ancestry estimation is not telling you how you look. These areas detected for ancestry are "non-exom" regions of the DNA that are not "genes." These areas examined for ancestry are in non-coding regions of the DNA. Something most newbies don't understand. Only 1% of ones DNA accounts for phenotype.

As far as looking "Latina" or "Mexican" those populations are mixed European, Native American and African, in varying amounts. People from the Dominican Republic and Puerto Rico will have more African ancestry than your average Mexican. These areas had a lot of slaves.

There are people "half" Black with one Black parent that don't "look" Black. Look at Derek Jeter and his family. Jeter is part Black and looks Latino, as his sister does. He is probably I would take a wild guess if he tested 40%/45%. His father looks European influenced. Latino is not a specific "look." Many phenotypes in Latin America from Euro to African to Native.

http://img.theepochtimes.com/n3/eet-content/uploads/2014/07/15/93535425-676x450.jpg

Your mtDNA is only a fraction of your DNA and does not represent someone's ancestry but a tiny fraction. It's one maternal ancestor, could be from 20 generations ago. If you click on the info on your mtDNA, it will tell you geographical location(s) of that halpogroup.

Your autosomal (Ancestry Composition) shows recent ancestry, prime at a 3rd great grandparent but can go back to a 5th great grandparent. DNA is randomly shuffled. Can be lost about 4th/5th great grandparent. This gives you a ballpark of what is in your recent ancestry.

If you are 19% SSA, your mother was at least 45%/50%-ish. A parent could have been Black or a multigenerational mulatto. Most New World Blacks have some sort of admixture. None are 100% SSA.

Being Indian by enrollment or status is different than that of being biologically Amerindian. It appears your Native/Asian is 1.7? This would indicate your mother's mother, had more roots that are European/African.