Pashtun genetic deep ancestry

[Dilawerkhan]

I have always been curious as to why the Sorani Kurdi language sounds the most similar to Pashto from the other languages in the area. Here is a clip from a Kurdi film https://www.youtube.com/watch?v=gZRE2tcaGJI

Aside from this, I have also wondered why there are common Kurdi-Pashto words, but are missing in Farsi.

This motivated me to take the time to look at Pashtun deep ancestry using formal methods of genetic analysis.Since these tests reveal information on deep ancestry and population mixing events thousands of years ago, I will explain the difference between these tests, and tests such ADMIXTURE (available on gedmatch.com) and 23andMe, which convey information on mixing events only 100-1000 years ago.

ADMIXTURE, 23andMe, Ancestry DNA, etc.:

These types of tests are based on allele frequencies or haplotype segment matching, and as such cluster individuals into groups based on shared recent genetic
drift, or recent mixing events. In the case of 23andMe for example, their algorithm basically eliminates minor components, or components that have become diluted
due to age. As such, it is only informative to mixing events and drift within the last 200 years or so.

For example, assume there is a group of people in the middle east called A. Lets say that A splits into 2 groups, group B, and group C. B stays in the middle
east, and C migrates to SC Asia. If you test B and C right after they split, 23andMe will show similar results for both. However, if you test B and C 1000 years after they split using 23andMe, the results will show B as about 100% Middle Eastern, and C as 100% South Central Asian, even if B and C remain isolated and don't mix with other groups.

In other words, let us say that A is common ancestor of some Kurds and some Brahui or Baloch, and B represents Kurds, and C Baloch. Right after A splits into B and C, 23andMe will show both B and C (Kurd and Baloch) as 100% Middle eastern, but 1000 years after the split, 23andMe will show B as 100% Middle Eastern, and C as 100% SC Asian, even if B & C (Kurd and Baloch) remain isolated for 1000 years and don't mix with any other groups.

The reason for the different results for B and C after 1000 years is allele frequency change due to drift, mutation, and natural selection.

Keep in mind that the number of SNPs that define a population cluster, such as E Asian or Sub-Saharan African, in ADMIXTURE can be as few as 1000 SNPs. In other words, if one tests 100,000 SNPs, theoretically, 2 individuals can show up as 100% E Asian, even if 99,000 of their SNPs are different from each other, as long as those
specific 1000 SNPs match the E Asian reference.


FORMAL METHODS SUCH AS DSTATS, QPADM, ETC.:

Formal methods, by contrast to the above, are based on shared total drift, and whereas in the above example, the results may be driven by as few as 1000 SNPs, with
Dstats for example, the results are based on the 99,000 SNPs, and not the 1000 SNPS as was the
case in the above example with ADMIXTURE.

Therefore, in the above example, if B & C were one group (A) for 100,000 years, then split 1000 years ago, Dstat results will be reflect a very high similarity between B and C, because they were one group (A) for 100,000 years, and the last 1000 years that they have been separate will not affect the results significantly.

23andMe on the other hand will ignore for the most part that B & C were one and had shared drift for 100,000 years, and show them very different from each other, since the results
are based on mixing events, drift, mutations, etc in the last couple of hundred of years only.

This is why 23andMe is totally useless for determining shared ancestry in 2 groups, if they split from each other 1000 or 2000 years ago. For that we have to use formal methods, such as Dstats.

[Dilawerkhan]

The Pashtun samples I will be using are HGDP Pakistani Pashtun samples, as well as some Afghan Pashtun samples. I will run tests time permitting and post results over the next couple of weeks.

To start, I compare Pakistani Pashtun DNA sample HGDP00244 with various European, W Asian, SC Asian, S Asian, and African samples.

Each line of the following table compares 2 samples, Iraqi Kurd sample C3 with another sample, to see, which one shares the most genetic drift (similarity) with Pashtun12 (HGDP00244). If the D score is +ve, that indicates that Kurd C3 is more similar to Pashtun12 than sample 2 . If -ve, that indicates that sample 2 is closer to Pashtun 12. Since there are no -ve results, that means that Kurd C3 is more similar to Pashtun 12 than all the other samples in the study.

For example, in the Kurd C3 - Kalash comparison the D score is small, indicating that Kurd C3 is only slightly more similar to Pashtun 12 than Kalash, whereas with Kurd C3 - Somali, the D score is large, indicating that Kurd C3 is much more similar to Pashtun 12 than Somali is to Pashtun 12.

The table is sorted with the most similar to Pashtun 12 to least similar

[Dilawerkhan]

To visualize the results in the above table, I plot them using (1 - D) * 1000.

A couple of observations on the results:

1- Shared drift with Pashtun 12 is reduced for any sample that has significant African or E Asian ancestry, such as Makrani or Tajik

2- Pashtuns in the past (> 500 years ago) appear to have been majority W Asian (Kurd type) plus NE European type. This is evident by the scores favoring Kurd and NE Europeans (Finnish, Russian, etc). IMO, ancestral Pashtuns split off from proto-Kurds, and subsequently received a good dose of Andronovo related steppe admixture, a couple of thousand years ago. Later, 0-1000 years ago, they received small doses of admixture component modal to S Indians via admixture with their neighbors. This is reflected in tests that convey information on recent drift, such as ADMIXTURE or 23andMe.

IMO, had Pashtuns not received significant geneflow from Andronovo types subsequent to their split from proto-Kurds, the tests would have shown them even closer to Kurd C3, as close to Kurd C3, as perhaps Brahui and Balochi are.

3- Tajiks and Pashtuns have alightly divergent drift histories as shown by Tajiks rank in the table. This is likely due to E Asian admixture decreasing total shared drift. However, tests such as ADMIXTURE and 23andMe will show them close to Pashtuns because they are based on recent admixture events.

[Dilawerkhan]

This comparison involves the average of 4 Afghan Pashtuns

[Aryana]

So your point is that Pashtuns are very close to Kurdish people? But then, why are Pashtuns and Kurds much more similar than Pashto and Kurds are to Persian, despite the geographic separation? Any sources on this?

[Dilawerkhan]

So your point is that Pashtuns are very close to Kurdish people? But then, why are Pashtuns and Kurds much more similar than Pashto and Kurds are to Persian, despite the geographic separation? Any sources on this?

Kurds and Pashtuns are very tribal, and lack scholars from 2000 or 3000 years ago, when proto-Pashtuns split off proto-Kurds (ancestors of Kurds and some other caucuses groups such as Georgians and Chechens), to document anything like this.

So although there is variation among Pashtuns, with some sharing more drift with Chechens or Georgians than Kurds, for me the tie-breaker is the closer language similarity between Kurdi and Pashto.

A plausible scenario to explain this maybe proto-Pashtuns (pre-Andronovo event) split from proto-Kurds in the caucuses area and migrated to SC Asia. Subsequently, a Kurd group, some Kurds of the Malik Kurd Dynasty, as detailed in the book I posted in the thread "Kurds in the history of Khorasan), left the area known os Ghor when Tamerlane captured it in 1380AD and back-migrated to the Zagros mountains to seek shelter from the Mongols. They would later be known as Sorani Kurds (Kurds from Soran). Perhaps they are related to Soori Pashtuns as someone posted here. That would explain why Sorani sounds like Pashtun and shares vocabulary and grammer with Pashto.

This may also explain elevated recent SC Asian admixture component in Kurds in general, and especially Sorani Kurds (ADMIXTURE), and the relatively high frequency of haplogroup R1a1 among Kurds.

Again due to the tribal nature of Pashtuns and Kurds, and lack of scholars from around 1380 AD, I doubt we will see anything written. We will just have to depend on genetics and other circumstantial evidence such as language, grammer, pronounciation.

It appears that the elevated SW Asian input into Iranians is causing them to have less total shared drift with Pashtuns, and placing them further down in the table. Something else that can be causing Iranians to be further down is that they received less Eurasian steppe Andronovo related geneflow than Kurds and Pashtuns.

[Iblis_Efreetzai]

[SIZE=3]I have always been curious as to why the Sorani Kurdi language sounds the most similar to Pashto from the other languages in the area. Here is a clip from a Kurdi film https://www.youtube.com/watch?v=gZRE2tcaGJI

I'm going to be honest with you, it sounds NOTHING like Pashto, it sounds like Farsi....Iranian Farsi.

[Sarbani]

[MENTION=18350]Dilawerkhan[/MENTION]

How come Pashtuns and Kurds seem so close on your graphs yet they never plot close to each other on PCA plots and have completely different autosomal DNAs?

[Iblis_Efreetzai]

[MENTION=18350]Dilawerkhan[/MENTION]

How come Pashtuns and Kurds seem so close on your graphs yet they never plot close to each other on PCA plots and have completely different autosomal DNAs?

He does some data manipulation to skew the plots.

[Dilawerkhan]

I'm going to be honest with you, it sounds NOTHING like Pashto, it sounds like Farsi....Iranian Farsi.

I am not sure how much Pashto and Kurdi you speak and understand, but according to my Pashtun friends, and myself (I can speak and understand Kurdi and Pashto and am familiar the grammer structure), and according to posters such as

Originally Posted by Shapur View Post
That clip is hilarious, they sound like they are speaking Pashto but its not.

My college physics teacher was a Kurd from Northern Iraq and he said one day he was just going through different radio stations and tuned into a station he thought was Kurdish, he listened for a few minutes but couldnt undrestand what was being said. Later when Afghanistan was mentioned he realised it was in pashto. So Kurds also hear this similarity.

the similarity in pronounciation is evident. Also Farsi lacks letters such as Zh, Dz, Tz. I have also jokingly said that a sentence that takes 30 secs to say in Farsi can be said in 15 secs in Pashto or Kurdi.

[Dilawerkhan]

@Dilawerkhan

How come Pashtuns and Kurds seem so close on your graphs yet they never plot close to each other on PCA plots and have completely different autosomal DNAs?

Please see my 1st post in this thread, I have explained the reasons in detail.

[Iblis_Efreetzai]

I am not sure how much Pashto and Kurdi you speak and understand, but according to my Pashtun friends, and myself (I can speak and understand Kurdi and Pashto and am familiar the grammer structure), and according to posters such as



the similarity in pronounciation is evident. Also Farsi lacks letters such as Zh, Dz, Tz. I have also jokingly said that a sentence that takes 30 secs to say in Farsi can be said in 15 secs in Pashto or Kurdi.

I'm conversant in Pashto, and it doesn't sound like Pashto, the accent is way different, the way the vowels and the "zha" consonants are different.

As for the grammar, all Indo-Iranian languages share more or less the same grammatical structure.

[Dilawerkhan]

I'm conversant in Pashto, and it doesn't sound like Pashto, the accent is way different, the way the vowels and the "zha" consonants are different.

As for the grammar, all Indo-Iranian languages share more or less the same grammatical structure.

I did not say it was 100% Pashto, but almost all Pashtun speakers that have listened to Sorani Kurdi have said that it is SOUNDS much more similar to Pashto than Farsi or other area languages.

Add to that pronouns such as la, pa, etc, and letters such as zh, dz, tz, and the the similarity becomes greater than with Farsi-Pashto.

Here is some similar Kurdi-Pashto words. I am sure I can find more if I spend the time. So the greater similarity Kurdi-Pashto vs Farsi-Pashto is what I am referring to.

[Iblis_Efreetzai]

[SIZE=2]

I did not say it was 100% Pashto, but almost all Pashtun speakers that have listened to Sorani Kurdi have said that it is SOUNDS much more similar to Pashto than Farsi or other area languages.

]

Can you provide and concrete data analysis of studies done when Pashto speakers were asked to comment on Kurdish?

Sorry, but to me it sounds like Farsi

[Dilawerkhan]

This Dstat comparison was done using a different Iraqi Kurd sample than before. The table and graph show total shared drift with Pakistani Pashtun DNA sample HGDP00262. The results are consistent with those of Iraqi Kurd sample C3.

Kurd C2 shares more total genetic drift (genetic similarity) with Pashtun18 than all the other W Asian, European, SC Asian, and S Asian samples in the study.

Kurd C2 is followed by Baloch, then Kalash, then Russian. The NE European affinity of Pashtun 18 is clear, and very likely a result of Eurasian steppe, Andronovo related geneflow from about 2000 years ago.

To get a better understanding of the deep ancestry formal methods used here. please carefully read my initial posts.