okarinaofsteiner's East Eurasian GEDmatch megathread

[Tsakhur]

This individual wasn't in my original n=18 Nusantara dataset, and they score 69.02% East Eurasian (the lowest scoring person is in my original n=18 dataset is 67.32% and they have a lot of Papuan-like ancestry). This person seems ~12% Euro.



Two of them have Burmese-sounding names and two of them seem to be Thai. The likely Burmese samples are at (6.85, 1.98) and (7.46, 2.65)- they score ~62% and ~72% East Eurasian respectively. The other 2 are probably not Burmese because they score higher on my N-S East Eurasian cline- those two are around 77% and 66% East Eurasian.

On an unrelated note, one of the "Filipino" samples in my original n=80 dataset is Guamanian/Chamorro, but has partial Filipino ancestry through a grandparent. This person scored 54% AN, 20% S_EA, 6% T_A, 3-4% Amerind, and ~81% East_Eurasian.



That may be true, but it isn't obvious from where the "Amerind" ancestry component plots on the Global PCA. It is directly below "Austronesian" and "South_East_Asian" on the PC2 axis, at somewhere around (9.60, 3.80) or so.

Do you have the Gedmatch kit for this Nusantara individual? Actually he/she could have a lot of Indian ancestry as well besides 12% Euro admixture; South Asian gene flow could also make this person shifts significantly west. Furthermore. what's the ethnicities of the other three Nusantara samples who seem to be the most western-shifted after that particular Indonesian individual?

Right. Can you share there Gedmatch kits please? They definitely possess a lot of South Asian ancestry which is half West Eurasian in composition. This makes them a lot more West Eurasian-shifted than other East Asian individuals and make them plot faraway from the East Asian cluster.

Are you basing your Global PCA on MDLP K23b? That could be because Amerind component is initially assumed to be 100% East Eurasian by the calculator creator just like Austronesian and Southeast Asian components, when actually Amerind are a mix between East Eurasian+ANE which seems to be much more closer to West Eurasians than East Asians and is closely related to Upper Paleolithic Europeans such as Sunghir, Kostenki from what I have read.

Can you post the Global PCA showing these components again?

[okarinaofsteiner]

Do you have the Gedmatch kit for this Nusantara individual? Actually he/she could have a lot of Indian ancestry as well besides 12% Euro admixture; South Asian gene flow could also make this person shifts significantly west. Furthermore. what's the ethnicities of the other three Nusantara samples who seem to be the most western-shifted after that particular Indonesian individual?

Right. Can you share there Gedmatch kits please? They definitely possess a lot of South Asian ancestry which is half West Eurasian in composition. This makes them a lot more West Eurasian-shifted than other East Asian individuals and make them plot faraway from the East Asian cluster.

Are you basing your Global PCA on MDLP K23b? That could be because Amerind component is initially assumed to be 100% East Eurasian by the calculator creator just like Austronesian and Southeast Asian components, when actually Amerind are a mix between East Eurasian+ANE which seems to be much more closer to West Eurasians than East Asians and is closely related to Upper Paleolithic Europeans such as Sunghir, Kostenki from what I have read.

Can you post the Global PCA showing these components again?

Check your PMs. Idk if any of the kits are still available to the GEDmatch public, I last looked at those (on GEDmatch) in 2018-2019.

The global PCA is based on MDLP K23b, more specifically this chart I found on Anthrogenica 3-4 years ago.




It's possible that the Amerind ancestry component is actually a 70:30 split between something East Eurasian-like and something ANE-like, but the East Eurasian-like component is much more divergent from non-East Eurasian than modern-day East Eurasians (even E_Siberian and Paleo_Siberian) that would be even further to the right on the x-axis and further up on the y-axis. The Amerind component is already somewhat shifted towards "West Eurasian" relative to the East Eurasian components as it is.

[Max_H]

La368 is ~5% more West Eurasian (EHG + EEF + Caucasian + N_African + Near_East + S_Central_Asian) and ~5% more African than Onge. I'm guessing that's more of an issue with the calculator working with ancient samples.

Sounds more like ancient DNA damage. Also Onge is very drifted so that could also affect these estimates.

But Global25 is not very accurate when using populations older than 15kya from what I've gathered.

[Max_H]

Thanks! I think the "Demographic History of Southern Chinese" PCA plot shows how different populations score on MDLP K23b. Ami is further away from Visayans and Ilocanos, which also score higher on "South_East_Asian" and lower on "Austronesian" than Ami in MDLP K23b (and therefore are closer to Mainland SE Asians and Han Chinese).

---

GEDmatch One-to-Many Sample count:

• Cambodia (n = 14)
• Nusantara (n = 22)
• Japanese (n = 31)
• Korean (n = 52)
• Northern Chinese (n=94) (assigned based on MDLP K23b results, not known ancestry)
• Southern Chinese (n = 160) (assigned based on MDLP K23b results, not known ancestry)
• Vietnamese (n = 80)
• Filipino (n = 68)

Means and Medians for my East Asian population samples | N-S cline vs % East Eurasian. I think the grey dots are "normalized" medians, calculated from the aggregate results (sum of all the ancestry components), whereas the black dots are calculated from finding the median of each ancestry component.


Means and Medians for my East Asian population samples | Global PCA cline. Means are shifted away from the East Eurasian ancestry components because of outliers with more "non-East Eurasian" noise and/or more Hoabinhian/Papuan/South Asian-like ancestry.

---

Japanese mean- 42.12% T_A, 35.13% S_EA, 19.51% AN, 0.84% Paleo_Sib, 0.60% E_Sib, 0.46% Amerind, 0.42% S_Indian. 98.20% East Eurasian, 0.3702
Japanese median- 42.07% T_A, 35.09% S_EA, 19.47% AN, 0.72% Paleo_Sib, 0.14% E_Sib, 0.12% Amerind, 0% S_Indian. 98.24% East Eurasian, 0.3687

Korean mean- 40.73% S_EA, 38.24% T_A, 18.49% AN, 1.68% E_Sib, 0.21% Paleo_Sib, 0.19% Amerind. 99.36% East Eurasian, 0.3816
Korean median- 40.87% S_EA, 38.21% T_A, 18.35% AN, 1.67% E_Sib, 0% Paleo_Sib, 0% Amerind. 99.80% East Eurasian, 0.3798

"N"_Chinese mean- 46.39% S_EA, 29.94% T_A, 21.22% AN, 0.91% E_Sib, 0.18% Paleo_Sib, 0.21% Amerind. 98.62% East Eurasian, 0.4448
"N"_Chinese median- 46.25% S_EA, 29.95% T_A, 21.06% AN, 0.36% E_Sib, 0% Paleo_Sib, 0% Amerind. 99.31% East Eurasian, 0.4456

"S"_Chinese mean- 46.99% S_EA, 33.15% AN, 18.85% T_A, 0.91% E_Sib, 0.10% Paleo_Sib, 0.21% Amerind. 99.25% East Eurasian, 0.5694
"S"_Chinese median- 46.93% S_EA, 32.59% AN, 19.51% T_A, 0% E_Sib, 0% Paleo_Sib, 0.21% Amerind. 99.39% East Eurasian, 0.5607

TW [ROC] mean- 46.71% S_EA, 32.17% AN, 21.18% T_A, 0.08% S_Ind, 0.11% Aus, 0.17% E_Sib, 0.14% Paleo_Sib, 0.07% Amerind. 99.34% East Eurasian, 0.5485
TW [ROC] median- 47.06% S_EA, 31.13% AN, 20.37% T_A, 0% S_Ind, 0% Aus, 0% E_Sib, 0% Paleo_Sib, 0% Amerind. 99.69% East Eurasian, 0.5583

HK/Canto mean- 46.50% S_EA, 33.62% AN, 18.47% T_A, 0.14% S_Ind, 0.11% Aus, 0.23% E_Sib, 0.09% Paleo_Sib, 0.17% Amerind. 98.92% East Eurasian, 0.5733
HK/Canto median- 46.45% S_EA, 35.62% AN, 16.74% T_A, 0% S_Ind, 0% Aus, 0% E_Sib, 0% Paleo_Sib, 0% Amerind. 99.15% East Eurasian, 0.5983

Viet mean- 45.04% S_EA, 40.61% AN, 10.14% T_A, 2.05% S_Ind, 0.45% Aus, 0.17% Mel-Poly, 0.13% E_Sib, 0.22% Paleo_Sib, 0.17% Amerind. 96.14% East Eurasian, 0.6549
Viet median- 45.06% S_EA, 41.07% AN, 9.69% T_A, 1.93% S_Ind, 0.24% Aus, 0% Mel-Poly, 0% E_Sib, 0% Paleo_Sib, 0% Amerind. 96.45% East Eurasian, 0.6601

Filipino mean- 52.51% AN, 29.88% S_EA, 7.46% T_A, 2.96% S_Ind, 1.25% Aus, 2.44% Mel-Poly, 0.27% Amerind. 90.11% East Eurasian, 0.7471
Filipino median- 52.67% AN, 29.21% S_EA, 7.15% T_A, 2.78% S_Ind, 1.08% Aus, 2.28% Mel-Poly, 0% Amerind. 90.06% East Eurasian, 0.7544

Not posting the "Cambodian" and "Nusantara" means/medians because the sample sizes were pretty small (< 20).

That was very interesting but a bit confused , what's the non-East Eurasian component of Japanese? Papuan/ASE-like?

I would think in most East Asians it's probably something deeply Onge-related except for those with West Eurasian ancestry like NW Han etc. On that topic would you have an estimate of West Eurasian ancestry for Shandong_Han?

[okarinaofsteiner]

Sounds more like ancient DNA damage. Also Onge is very drifted so that could also affect these estimates.

But Global25 is not very accurate when using populations older than 15kya from what I've gathered.

La368 is from only 8000 years ago, but that's still long ago enough for weird noise to mess things up. I remember the "Han_Xiongnu" ancient GEDmatch samples also had some weird noise in the MDLP K23b calculator

---

Expanded N-S cline vs % East Eurasian plot of almost all of my East Eurasian samples that I classified as not ethnically mixed + several additional GEDmatch kits that @Tomenable asked me to look at in early 2019. This one includes the "Chinese" samples that I specified as "Taiwan [ROC]" and "HK/Canto" in clear red and yellow boxes, with additional "SEA Chinese" samples as clear maroon boxes. It also includes 3 Hmong samples and several Tibetan/Himalayan samples from my original dataset.

Tomenable's requested samples include:

• 32 or 33 Cambodians
• 1 Cambodian/Laotian
• 11 Nusantarans (mostly Indonesian)
• 10 Thai nationals
• 2 confirmed Laotians
• 6 additional Island SE Asians
• 1 Thai sample from HGDP

I labeled the country of origin clusters to give a better idea of where Laotians, Cambodians, native Indonesians, etc.- basically non-Vietnamese, non-Filipino SE Asians score relative to my Chinese, Japanese, Korean, Vietnamese, and Filipino samples.

We can see that:

1. the Nusantara samples are still mostly in between the Filipino and Cambodian samples. Which checks out with our current knowledge of Indonesians and native Malaysians having both Austronesian and Austroasiatic ancestry.
2. Burmese are more West Eurasian/Indian-shifted AND more NE Asian-shifted than other SE Asians
3. Laotians are more distant from Han Chinese AND more Indian + Hoabinhian shifted than Vietnamese, but still more "pure East Eurasian" than most Thais who do not have significant Chinese ancestry.
4. Thais seem to have a lot of variance, but generally fall in between Burmese, Khmers, Malays/Indonesians, Lao, Kinh (Vietnamese), and SEA Chinese (who have varying levels of admixture with native SE Asians)
5. Hmong are similarly "southern/northern" and similarly "pure East Eurasian" as Guangdong Han- but have lower levels of "Austronesian" or "Tungus_Altaic".
6. Tibetans and Himalayans are somewhat more NE Asian-shifted than Northern Han Chinese, but also have significantly more Onge/Hoabinhian-like ancestry. Although with Nepalese much of this "non-East Eurasian" ancestry is probably actual South Asian (steppe + Indus Valley + AASI) ancestry.
7. Mainland SE Asians almost never score >0.75 on my N-S cline. 0.75 is where Dai falls on my cline (50% S_EA 50% AN in MDLP K23b), which is probably a good proxy for how "southern" the "pure East Eurasian" component in Austroasiatic is.
8. Island SE Asians usually score >0.75 on my cline, which makes sense because they almost always score higher on "AN" than "S_EA" + "T_A". We can think of modern-day Island SE Asians as being on a cline between "Dai" (continental/Austroasiatic) and "Igorot" (insular/'pure' Austronesian).

[tgndjs]

This might be a bit random, but how did you get all these GEDmatch kit numbers?
and is there a reason why you picked MDLP K23b over other calculators such as Harappaworld or Eurogenes k36?
I am still fairly new to whole genetic admixture thing and I'm just curious.

[okarinaofsteiner]

This might be a bit random, but how did you get all these GEDmatch kit numbers?
and is there a reason why you picked MDLP K23b over other calculators such as Harappaworld or Eurogenes k36?
I am still fairly new to whole genetic admixture thing and I'm just curious.

I played around on GEDmatch Genesis throughout 2018 after uploading my 23andMe raw data file in late 2017, and compiled a list of (mostly East Asian) samples I found using the One-to-Many DNA Relatives tool.

I chose MDLP K23b because it not only has the most reference populations, but also has the greatest variety of East Asian reference populations that correspond to specific ethnolinguistic groups and subpopulations. No other calculator has more than 1 reference population for Korean or Vietnamese, or for native Cantonese speakers.

[tgndjs]

I played around on GEDmatch Genesis throughout 2018 after uploading my 23andMe raw data file in late 2017, and compiled a list of (mostly East Asian) samples I found using the One-to-Many DNA Relatives tool.

I chose MDLP K23b because it not only has the most reference populations, but also has the greatest variety of East Asian reference populations that correspond to specific ethnolinguistic groups and subpopulations. No other calculator has more than 1 reference population for Korean or Vietnamese, or for native Cantonese speakers.

Thank you for such as detailed answer

[jortita]

Thai are quite heterogeneous, as evident from this latest article linked to the Thai genomic project with Northern Thai being closer to Chinese Dai, Isaan or NE Thai being mix of austroasiatic khmuic and Dai while central thai and southern thai non chinese admixed are closer to Mon and hence Monic with recent Indian admixture, https://www.biorxiv.org/content/10.1....424294v1.full.

As i am separated in my real life, I went on to Thai tinder for long term relationship and I got 25 matches from only Thais as they thought I was northern Thai or thai chinese admixed, given my Daiic features. No single western or Indian match or even likes which I found interesting inspite of having an Indian passport

[cmd06]

"Island SE Asians usually score >0.75 on my cline, which makes sense because they almost always score higher on "AN" than "S_EA" + "T_A". We can think of modern-day Island SE Asians as being on a cline between "Dai" (continental/Austroasiatic) and "Igorot" (insular/'pure' Austronesian)."

As you've mentioned, island southeast asians are on a cline from dai-like to igorot like. Which among malays/indonesians and filipinos would cluster closer to dai or to igorot. It was not indicated on the graph where the dai would be, i'm assuming close to the viet or lao samples? Also, can this cline be applied to cambodians? They seem rather genetically close to malays/indonesians. Based on the graph, filipinos are exactly on the 0.75 mark? With nusantara samples also really close the 0.75 mark.

Also I agree with one comment above. Northern/central/southern thais are probably genetically closer to their immediate neighbours such as dai (for those in the north), mon-khmer (central - south probably with additional austronesian input from the malays particularly in the south).