Genetic distance in Europe and South Asia

By Razib Khan 40 Comments


People always ask about genetic distance. Above is a NJ tree of a pairwise Fst.

Here is the raw table.

Please stay chill in the comments.

(used Plink’s Fst with 200,000 markers)

3+

40 Replies to “Genetic distance in Europe and South Asia”

  1. So stupid question time. Is the correct way to read the graph to measure the actual physical distance between the population labels? I.e. confirming (as I read in the reich book) that indeed south asians are far more distant from each other than europeans are from each other?

  2. it’s using the pairwise Fst matrix as input. if you want real pairwise comparisons, then look at that. also, there are different measures and Fst can get inflated by bottlenecks (so i removed pulliyar).

    but any way you measure it there is more within pop strat in s. asia than europe.

  3. Thanks. Have you ever thought about opening a store like eurogenes where you could provide g25-like coordinates but more specific and tailored for south asians? Charge a 20 bucks per request or something. People can find out how much steppe, farmer, and aasi they are?

    I bet you could sell something like that to one of the ancestry companies eventually. My 23andme data told me nothing that I would actually consider relevant.

  4. Why don’t you ever include Bengalis. It’s a freaking disgrace SMH.

    Also, the bigger the Fst number between populations, the closer they are, right? Closer to 0? If two populations have Fst of 0.1, they are closer than two population with 0.01?

    Add BENGALIS!!! You’re supposed to represent us. It kind of irks me. Finally we a prominent pop geneticist Who is Bengali but he rarely speaks on Bengali genetics anymore. You used to before and it was some of the most interesting pieces I’ve ever read, it made me get involved in pop genetics. But now, freaking Jatts? Patel’s? Rors? What’s wrong with you fool.

      1. Bengali 2 is Bengali_SC types found in some BEB samples I suppose. They generally cluster with Indian dalits and might’ve migrated during british era or they could be actual Bengali dalits.

        1. they have no east asian. so that’s why i’m skeptical they’re NOT migrants. but who knows.

          also they were collected/sampled same time (there are consecutive IDs).

        2. Why would Bengali Dalits be different if no bullshit “Aryan migration” happened and the whole real “Caste” thing was a European feudal serf introduction? “Dalit” means “knowledge giver” in Sanskrit btw and Brita made up a lot of the negative connotation around them.

    1. Bengali Brahmins can be used. I don’t think they are off cline. Their e asian is similar and in some calculators lower than Kash Pandits, Dogra Brahmins, Nepali Brahmins, Pashtuns and Dardi groups like Kalash as can be seen in the G25 spreadsheet here: https://docs.google.com/spreadsheets/d/1rodW-JJ63C5ox7gaWnJxhWgyvWJs4WGyd_tB7UMMy50/edit#gid=977819834 and all of these groups were tested as cline-groups in Narasimhan. The basic Bengali type in my opinion would be the mid-low caste Agrarian Hindu groups of southern West Bengal or nearby regions in Bangladesh like the Mahishyas, Baishyas, Sahas, Kolis, Barujbis, etc. Their e Asian would be of similar percentage to that of the “BangladeshiSC” pop which they likely would have received from AA speakers nearby. BEBs are not the basic type in my opinion since their elevated e Asian looks like they recently (past few hundred years) got that kind of ancestry from nearby Boro-Kachari groups, compounded by the fact that Muslims are much more liberal (more foreign admixture can be observed among the Muslim counterparts of other groups of South Asia as well). There are cases of Muslim royalties from Bengal marrying princesses from neighboring Tripuri kingdoms.

      1. Their e Asian would be of similar percentage to that of the “BangladeshiSC” pop which they likely would have received from AA speakers nearby. BEBs are not the basic type in my opinion since their elevated e Asian looks like they recently (past few hundred years) got that kind of ancestry from nearby Boro-Kachari groups, compounded by the fact that Muslims are much more liberal (more foreign admixture can be observed among the Muslim counterparts of other groups of South Asia as well). There are cases of Muslim royalties from Bengal marrying princesses from neighboring Tripuri kingdoms.

        i know more about this than almost anyone, (looking for 5 years at this). so

        1) the bangladesh_SC are 0% e asian. they cluster with dalit type groups from the rest of india. they don’t have residual east asian at all

        2) the bangladesh samples are from dhaka. they should be all over the country. yes, their east asian seems to be more tibeto-burman. i don’t know why are you acting as if they are atypical when ‘bangla’ types are quite numerous

        3) i will look more closely at the west bengal kayasthas to see if the have munda or tibeto-burman ancestry, but over the years i have gottne skeptical of the idea that mundas were a presence in NE india at all, aside from recent migrants. the tell is the mtdna. the mundas have almost no e asian mtdna, but bengals have a fair %

        4) west bengal is undersampled. pretty frustrating

        1. “west bengal is undersampled. pretty frustrating”
          Pretty frustrating indeed. Same with neighboring regions. Would’ve like to know more about all eastern Indo-Aryan communities.

          “i will look more closely at the west bengal kayasthas to see if the have munda or tibeto-burman ancestry”
          Wonder if there would be a distinction between “Ghoti vs Bangal” in east Asian, i.e. Bangal’s east Asian being tibeto-burman like Bangladeshis.

  5. @Razib

    Just a question regarding your samples/datasets

    Are your Jat samples all Punjabi Jatts (Both Muslim and Sikh)? The Pathan and Sindhi samples are all from HGDP and include the outliers in those datasets? Finally, are the Gujjars the Indian Rajasthani ones or the Pakistani ones from KPK?

  6. indian jatts. sikh or hindu, i forget (not Muslim, i left that individual out).

    hgdp. the gujjars are indian i’m assuming. they’re from the pathak et al paper

    1. Thank you for the clarification! Pathak has the Rajasthani Gujjars and Daamgard has the KPK ones so it’s the Indian ones as you noted.

      Do you mind adding the Khatris and Gujarati Brahmin from Daamgard?

      Same for the Yusufzai/Uthmankhel/Tarkalani from Daamgard? They don’t have the same type of outliers as the HGDP Pathan/Pashtun set.

  7. @Razib
    Is Velamas a drifted population too?,as it is showing 0.008 distance even with South Indian Brahmins.
    You should have used some other group to represent Non Brahmin Non Dalit South Indian.

  8. Unless I am reading this wrong, how come Arabs be closer to French Basques than Germans? It’s quite odd when one considers their positions and ancestral components.

    Furthermore Greeks have a lot of Slavic admixture so some of them would not overlap with Sicilians and be closer to Germans.

    I might be mistaken of course, just what I’ve gathered on European genetics so far.

    1. It may depend on who is classified as Arab. IIRC, after EEF left Anatolia, eastern and western (i.e. EEF-related) Middle Easterners mixed up pretty evenly. But maybe parts of Arabia preserve more EEF-like ancestry without so much eastern ME admixture? I think that’s also the case with Egypt.

      1. That is true. Also, the Hindu Jats are right there with the Rors, since they are one and the same group. The “Jat” group here is more representative of the Sikh Jats in this regard, who have more AASI admixture, and a wider range of AASI, owing to being historically more assimilating of different groups. We already talked about this in far more detail, so feel free to go back and read that discussion.

        Finally, Iranians are not the parental group for Jats — Jats have many different sources of ancestry, only some of which is shared with most Iranians in any great magnitude. Jats have a different ethnogenesis, and Iranians have a very different one. Even if Hindu Jats and Rors were 100% West Asian, (as opposed to ~85%) they wouldn’t cluster with modern-day Iranians. Heck, even Parsis have been shown not to cluster with modern-day Iranians when their 10-15% AASI admixture is taken away in a model. This is because Iranians have been subject to more admixture events after the migration of the Parsis to India more than a thousand years ago.

        Therefore, comparing any South Asian population to Iranians in terms of genetic distance or PCA plotting is a moot point. Even Parsis don’t fit the bill. We are most similar to our source population, which wasn’t Iranians to begin with. Also, West Asian doesn’t just mean Iran, far from it. FYI, even the Baloch, who are an Iranian people, plot equidistant to South Indian Brahmins and Iranians. This shows that Iranians have a lot of diversity within their population, and that everything isn’t as obvious as it may seem.

  9. 1) the iranians seem to be sampled from tehran. those from khorasan would be closer to indians

    2) look at the Fst matrix, not the NJ, if you see weird patterns. the NJ forces stuff into the unrooted tree. gives you a general sense, don’t take all details literally

  10. I finally got my g25 co-ordinates and have been playing around with them on vahaduo. Obviously the best fit comes when I run with Shahr-iSokhta BA2, but in an attempt to breakdown the AHG vs Iran_farmer bit I chose Ganj Dareh and simulated AASI. Here are the results, tell me what I did wrong:

    Target: Mohan_scaled
    Distance: 3.1671% / 0.03167145
    44.2 IRN_Ganj_Dareh_N
    36.6 100AHG
    10.8 RUS_Sintashta_MLBA
    8.4 Levant_PPNB

    1. Here it is with Shahr-i-Sokhta ba2 as I mentioned.

      Target: Mohan_scaled
      Distance: 2.5450% / 0.02545049
      86.0 IRN_Shahr_I_Sokhta_BA2
      8.0 Levant_PPNB
      6.0 RUS_Sintashta_MLBA

      Again, just looking for feedback on how i can improve the granularity. Is the approach in the above comment where i breakdown ganj dareh and simulated aasi a fair one?

  11. Razib, can you compare Southeast Asian groups with Bengalis. I want to know how far we are to people of the Philippines etc.

    1. That would be quite far. Bengalis are closer to a Lebanese than a Burmese, with whom we share the border.

  12. Wow, WTF!!!. A Bengali is closer to a Middle Eastern Lebanese than to a Burmese!!! I thought because of my 13-15% SE Asian admixture, I’d be somewhat close to SE Asians. I started feeling some closeness to them. I guess the Iran_HG and Steppe (the CHG part?) is what connects us to Middle Eastern people’s? But what the heck, I considered us being well away from Middle Eastern people’s.

        1. lol sorry, to clarify…

          … I thought because they had east asian admix they would be genetically somewhat more distant from other south asian groups, than this graph indicates.

  13. Razib!!!!

    1)Which population (ethnicity) outside of South Asia are Bengalis most related to?!?

    2) Which population are Bengalis most distant to?!?

    Please, I need to know. I can’t believe Bengalis are more related to Middle Eastern groups than Burmese. All this time I felt a closeness with East Asians. I looked at them as my brothers, my refuge when South Asians annoyed me with their bickering. I started relating to Japanese people and dreamt of moving to Singapore to be with my Chinese and Malay brethren. What is this??? This news came as such a shock. I’m still reeling.

  14. Isn’t almost every mainstream ethnicity at least slightly more Iran_N + Steppe than not?

  15. For future analyses would be beneficial if it is made a distinction between let say ‘basic’ or ‘organic’ (genetic) groups and ‘composite’ groups and not classifying based on passport holdings. In the first group are for example Serbs in the second are French, Italian, Greeks, Germans, etc. Would not mean anything if it is presented the genetics of, for e.g. Americans, Soviets, Frenches, Brits, Indians, Pakistanis or recent inventions such as Ukrainians, Bosnians, Macedonians, etc.

Comments are closed.