File talk:Phylogenetic structure of Eastern Eurasians.png

Deep ancestries

@Wikiuser1314: I am a bit unhappy about this one, and may you guess what it is about :)

You have Hoabinhians, AASI and Onge nested as closer to East Asian, with Tianyuan basal to the rest. The position of G1/AASI/Onge is still a matter of debate, and alternative topologies that have Onge closer to the Australopapuan lineage are still proposed, e.g. in this paper from Guanglin He's research group:[1] (repository link[2]), see Fig. 8. Guanglin He is a close associate of Chuan-Chao Wang, and they have published heaps of papers in the last 1.5 years where they basically graft various modern-day populations in China and Taiwan to the qpGraph in C. Wang et al. 2021[3] (yes, it's the Reich group paper that has Jomon admixed from a deep East Eurasian Onge-related source and a Fujian-related source – not your favorite model, I guess).

I am totally agnostic in this respect, but to be on the safe side, you should at least flip the position of G1/AASI/Onge and Tianyuan, if not doing a trifurcation. –Austronesier (留言) 14:50, 11 May 2023 (UTC)

Add: Jomon and Longling do not form a clade, but each are equidistant from Southern East Asian + Northern East Asians in f4 stats. Here, a trifurcation is defintely better:

• Jomon
• Northern East Asians + Southern East Asian
• Longlin

–Austronesier (留言) 15:03, 11 May 2023 (UTC)

@Wikiuser1314: You still have a wee branch conjoining Jomon and Longlin? –Austronesier (留言) 17:23, 11 May 2023 (UTC)

@Austronesier: I have adjusted the phylogenetic graph according to Yang's summary. Hoabinhians are described as most basal ESEA branch (as Andamanese are frequently associated with the Hoabinhians, I grouped them within the same clade as Hoabinhians, although deeply diverged; although their position differs significantly depending on different models, I think it is probably the least controversial one). This also should fit their affinity with Tianyuan, who is indeed shown as basal to Andamanese and Hoabinhians in other models as well (such as the Tianyi Wang paper, or the paper on Longlin). Do you see any further points?Wikiuser1314 (留言) 17:44, 11 May 2023 (UTC)

@Wikiuser1314: Have a closer look at Tianyi Wang et al. (2021) (which btw is the Longlin paper), especially Fig 2B and Fig S4. Tianyuan is a basal sister lineage to Hoabinhian and Longlin in the skeletal graph Fig S4B, but then all East Asians (in the narrower sense, i.e. NEA + SEA) derive most of their ancestry from a ghost that is a sister lineage to Tianyuan only, with various alternative models that tap additional deep ancestry from Longlin and/or G1. Which makes Tianyuan basal to extant East Asian groups, but not to other primary lineages. In many models, Onge branches off before Tianyuan (e.g. in C. Wang et al. (2021) mentioned above).

In the latest paper from the group around Guanglin He and Chuan-Chao Wang[4], they iterate their "standard model" from the 2021 Reich team paper, which has Onge as a representative of "South East Eurasians" ("Coastal EE" in the 2021 paper) and Tianyuan as basal to the "North East Eurasians" ("Interior EE" in the 2021 paper). All extant East Asians are then modeled as North East Eurasians with additional South East Eurasian geneflow, except for "Mongolian East_N", which is modeled as unadmixed North East Eurasian (Fig 3).

They also have a "modernized" model in Fig 6, which integrates Longlin. Here, Onge branches off first, then Tianyuan as basal to Longlin and extant East Asians, who receive minor geneflow (17%) from Longlin. –Austronesier (留言) 21:03, 11 May 2023 (UTC)

I'm still waiting for a study that boldly tries to cover all East Eurasians, and not just a part of them. It would interesting to see where Hoabinhian, Onge, Semang and Philippine Negritos would be placed in a wholistic picture. And of course, we need more aDNA samples (e.g. from the newly discovered pre-Neolithic Taiwan Negrito) Until then, I prefer to stay agnostic, given all these conflicting models. –Austronesier (留言) 21:03, 11 May 2023 (UTC)

@Austronesier: Absolutely, a study covering all available East-Eurasian genomes would significantly improve our understanding of their substructure. I have re-read some high quality papers (including the Wang 2021; although I am not very convinced by their approach [and their "mass-publishing of nearly identical papers], however on the other hand, they have (as far as I know) excluded Australasians and AASI (as well as the newly found Deep Tibetan lineage), so their model may actually not differ from the other models). I am personally preferring the models presented by Lipson and Reich, who argue for a trifurication of Australasians, Andamanese/Onge, and Eastern Asians. This would also align with the summary of Yang 2022, which however (:/) places the Hoabinhians on the ESEA lineage as deepest branch, and argues for an Australasian/AASI/ESEA trifurication. The deep ESEA (Onge or Longlin [more likely], alternatively "southern East-Eurasian") contributions to various EA groups seems logical, but we don't know yet if these contribution have their source on a basal ESEA lineage or on another East-Eurasian lineage. It is highly unlikely to simplify East-Eurasian diversity into a Northern and Southern East-Eurasian clade, and this is also contradicted by several other models (eg. the deep Tibetan lineage was found to be neither AASI, Onge or Australasian, but it's own deep branch: https://www.nature.com/articles/s41467-022-28827-2 and https://www.science.org/doi/10.1126/sciadv.add5582). My suggestion would be:

• including the deep Tibetan lineage (and it's relevant content to the article as well)
  
• improving the deep ancestries eg. Bacho Kiro and Ust'Ishim clades (and adding Oase)
  
• displaying Australasians, AASI and ESEA as own branches, with the deep Tibetan branch forming a forth one (a "quadrification"?)
  
• putting the Andamanese/Onge and Hoabinhians as very deeply diverged from the AASI, but on the same ("South Asian clade" in accordance with Lipson and Reich's trifurication model 2017); alternatively as deepest branch of the ESEA clade, basal to Tianyuan and all other ESEA (what do you think is the least controversial solution?).

What may be further relevant, the approach of Chuan-Chao Wang's team (2021 etc.) has been criticized by a recent review (broadly about admixture models): Maier et al. 2023(https://elifesciences.org/articles/85492), who could not reproduce their claims (eg. they got multiple better fitting models). Could you please elaborate on my suggestions above. I want to make an "as reliable as possible" phylogenetic graph without too much personal interpretation, so I am giving strong weight on your opinions. (I know you will stay agnostic as long as there is no perfect model, but I would appreciate some thoughts on the suggestions).Wikiuser1314 (留言) 06:42, 12 May 2023 (UTC)

@Wikiuser1314: It's been a busy day that has left little time for doing WP. I'll get back to you tommorow (hopefully)! –Austronesier (留言) 21:44, 12 May 2023 (UTC)

Hi @Wikiuser1314, I think you are on the right path in regards to your diagram, in the sense that you are trying to map out the structure. However, as Austronesier has explained and you have explained, it can never truly be accurate because we are still learning about all of this. I agree with your suggestion of making Ancient Tibetan as a separate lineage (which I can see you have already done). I also agree that "putting the Andamanese/Onge and Hoabinhians as very deeply diverged from the AASI" is the way to go because as you said it's in accordance to Lipson and Reich, supported by Yang and also correlates with common language that is used in terms of understanding the East Eurasian population structure (terms like "Onge-like", "Onge-related" and "Andamanese-like" are used extensively to describe those East Eurasian lineages that are genetically close to the Onge people or what some have described as "southern East Eurasian lineages"). For example, Guanglin He et al. 2021 uses the term "Onge-related" to describe the Onge-related and Hoabinhian-related ancestry, in contrast to "East Asian" ancestry , that was found among certain Tibetan populations, Chokhopani/"Ancient Nepal" and Jomon. It is generally understood as a more "southern" lineage that is on the "South Asian clade" (deeply diverged but still closest to other East Eurasians), though Yang supports the theory that Hoabinhian is a deep ESEA lineage while Andamanese is not. I understand views are different and some have different theories but all of the scientific papers I have read suggest that there was a clearly defined arc of lineages across Asia that could be described as "Onge-related" and distinguishable from the "core" ESEA lineages. So I am inclined to believe the scientific papers because that is the data we currently have. Just thought I'd provide my 2 cents. Saouirse (留言) 01:51, 14 May 2023 (UTC)

Well, I guess going with Yang's review conclusion may be the least disputed way, eg. putting the Andamanese as deeply diverged from AASI but on a 'South Asian' clade, but Hoabinhian as deepest ESEA clade. This would also (as you said) fit other papers (as well as not official arguments by some based on Vahaduo G25 datasets, which modeled Onge as AASI+Tianyuan-like). The relationship of these deep split lineages (Onge, Hoabinhian, Tianyuan) is less clear, particularly through their relative less drift compared to most modern populations, indicating that the diverging patterns must have been early. I however want to not that Mondal et al. 2016 grouped both Indigenous South Asians and Andamanese as closer to ESEA, with Papuans and Australasians as outgroup of the Eastern clade, but that may also be caused by the relatively higher archaic admixture as well as possible xOoA contribution to Oceanians. Either way, I think this paragraph form Yang 2022 should solve the unclear points (for now, we probably always can update the graphic if new data is published...):

Rapid diversification of an ancestral Asian population led to at least three Asian lineages, associated with Australasians and [Philippines] Negritos (AA), South Asians and Andamanese Islanders (AASI), and East and Southeast Asians (ESEA). Sampling across time and space in eastern Asia indicated that the ESEA lineage was highly sub-structured in the Upper Paleolithic, and populations across Siberia, East Asia, and Southeast Asia are associated with many distinct ancestries. Today, most populations from Siberia, East Asia, and Southeast Asia are associated with a mix of ancestries sampled predominantly in ancient individuals of East Asia.

In this regard, I can agree to put the Andamanese as deep sister branch of the proper AASI, while Hoabinhian is the deepest branch of ESEA. Although I personally think that Andamanese are better explained as basal to other ESEA rather than AASI, this follows the Yang review, and should at least deal with objections regarding accuracy. It probably also is somewhat consistent with some other recent papers, such as Changmai et al. 2022...Wikiuser1314 (留言) 06:25, 14 May 2023 (UTC)

I have a point to discuss here. Though Yang et al, reiterating Narasimhan, talked about deep divergence with the AASI, they didn't demonstrate the exact relationship between the two. Göllner 2022 explicitly mentioned the closeness of Andamanese and Hoabinhian instead. It can also be corroborated by these models using Vahaduo G25:

The one you suggested above (AASI + Tianyuan). Hmm.. OK kinda fit, but...

Target Distance AASI CHN_Tianyuan

Onge, 0.04835513, 19.8, '80.2'

Using Hoabinhian drastically improves the fit, best of the lot

Target, Distance, AASI, CHN_Tianyuan, LAO_Hoabinhian

Onge, 0.03240463, 7.8, 23.4, '68.8'

The fit more or less stays the same without AASI

Target, Distance , CHN_Tianyuan, LAO_Hoabinhian

Onge, 0.03695154, 7.0, '93.0'

Heck! using Hoabinhian alone the fit stays almost the same, far better than (AASI + Tianyuan)

Target, Distance, LAO_Hoabinhian

Onge, 0.03733828, '100.0'

Distance Distance to: Onge

0.03733828 LAO_Hoabinhian

0.07948807 CHN_Tianyuan

0.25900788 AASI:Sim

0.34269546 AASI:NW_SiSBA3

0.54096150 AASI:NW_SiSBA2

Onge (as well as Jarawa) is almost Hoabinhian like and is significantly closer Tianyuan pull then the AASI. I believe what you did in this version for Andamanese, Hoabinhian, Tianyuan and AASI is the best representation, supported by Göllner 2022, Yang and demonstrated above.
It is also geographically obvious considering the close proximity of the ancient Hoabinhians and the Andamanese as compared to the far away AASI of the mainland South Asia. Pinging @Austronesier: as well. - 117.201.116.232 20:36, 14 May 2023 (UTC)

Thank you for your comment. Indeed, Andamanese show highest affinity to Hoabinhians, followed by Tianyuan. That aligns with Göllner et al. (eg. figure 6.) So you would support putting the Andamanese on a clade with Hoabinhians, as sister clade to Tianyuan (as in one of the previous versions)? I can uploaded such adapted version, and additional cite Göllner here. Does anyone has objections?Wikiuser1314 (留言) 21:11, 14 May 2023 (UTC)

I think we are mostly in the same ball park. Here's my take on these issues:

1. Ancient Tibetan: This is a pretty stable research result from various studies, and further strengthened by a recent paper by H. Wang et el. (2023). Wang et al. (2023) is a high-quality source that serves as a secondary source for this particular key point in Tibetan genomic history. Further, this deep ancestry is found in numerous ancient individuals with a wide spatiotemporal distribution, so its inclusion is IMO fine by WP standards. (For comparison, I would hesitate to include the deep East (Eur)asian non-Papuan ancestry component found in the Toalean indivdual from Leang Painnge in Sulawesi (cf. Carlhoff et al. (2021)), as much as I'd like to, because this is still a very preliminary result from based on a very fragmentarily attested genome.)
   The only issue I have here is terminology. All studies use "Ancient Tibetan" for the extant ancient individuals, and not for the deeply diverged ancestry component. In analogy to the AASI ghost population, it might be called "Aboriginal Ancient Tibetan", but that's alas OR. Maybe "Ancient Tibetan ghost population"?
2. Bacho Kiro and Ust' Ishim are important pieces in the puzzle of the spread of OoA populations all across Eurasia. I'm not sure if the current working model is the final word (cf. the new preprint by Maier et al. which also has a short piece about alternative models for Bacho Kiro and Ust' Ishim vis-á-vis Tianyuan)
3. As for the AASI/Andamanese/Hoabinhian question, the problem is, how many studies are there that include Onge, Hoabinhian and AASI-related data at the same time? As far as I know there is none. McColl et al. (2018) compared Onge and Hoabinhian directly and found them to form a clade against all East Asians (incl. Tianyuan) and Papuans (btw their graph Fig 3C has the Onge/Hoabinhian clade closer to Papuans than to East Asians). F4-statistics in Göllner et al. (2022) also support this close relation of Onge and Hoabinhian (+ their Semang decendants) against East Asians. So it has always come as a surprise to me that in M. Yang's review, Hoabinhians are assigned to the ESEA lineage, thus separated from Onge (who are linked to AASI on p. 23). While it is not ours to "criticize" research results of our sources, we should always be able to corrborate statements in secondary sources based on their primary sources. In the case of Yang's treatment of Andamanese and Hoabinhians, I fail to see any support of it in her sources. Which means, I don't mind if we go our own way here based on a couple of other good sources like McColl et al. or Göllner et al.

I still have my reservations about putting Onge/Hoabinhian close to East Asians, especially closer than Tianyuan is to East Asians. My suggestion is to have the Ancient Tibetan ghost, AASI, Papuans and East Asians (in the wider sense) as "quatrifurcation". In the latter branch, I would split Onge/Hoabinhian first, then Tianyuan, then the rest as is. But the current trifurcation is also ok for me. –Austronesier (留言) 21:54, 14 May 2023 (UTC)

Add: just for fun, some f4-statistics from Göllner et al. (Figure 4)

• f4(Mbuti,Hoabinhian;Liangdao/Balangao,Onge) > 0 (which you'd hardly expect if Hoabinhians were ESEA but Onge not)
• f4(Mbuti,Liangdao;Hoabinian,Onge) ~ 0 (so Liangdao is an outgroup to Hoabinhian/Onge, again not compatible with Yang's model)
• f4(Mbuti,Papuan;Balangao,Onge) > 0 (Papuan shares more drift with Onge than with Balangao; if Papuan were an outgroup to Onge and Balangao, we'd expect a zero value)

–Austronesier (留言) 22:14, 14 May 2023 (UTC)

I agree with Austronesier, they bring up a very good point about the number of studies that include Onge, Hoabinhian and AASI data. What is currently being done here is that we are looking at all of the studies and trying to piece them together to create a bigger picture but as we already know, that's not a straight forward thing to do. We actually don't know the real population structure of East Eurasians, we are simply trying to make an educated guess based on the data we have so it will never be 100% accurate. The idea that Onge and Hoabinhian are part of ESEA as a deep lineage is not accepted anywhere, except for Yang to an extent (who claims Hoabinhian is ESEA and Onge is AASI). The realty is that Onge and Hoabinhian form a clade against all ESEA lineages. So placing them close to East Asians doesn't make sense when the data we have today does not support this conclusion. Onge and Hoabinhian are a lot closer to AASI than it is to ESEA lineages for example. Also, this brings me to my next point about AASI and the use of simulated data. We don't have a fully constructed image of who the AASI exactly were because there are no samples that are 100% AASI, if that makes sense. For all we know, AASI could be a hybrid population that formed from a Basal East Asian lineage and a deep East Eurasian lineage or had significant ESEA admixture (it would explain why simulated AASI is closer to ESEA than it is Australasians).

What we know is that Onge and Hoabinhian form a clade against all East Asians, putting them close to East Asians doesn't make scientific sense. We do know that Onge and Hoabinhian are close to one another as is AASI, with Onge and Hoabinhian being the closest to one another. Yang's review paper places Hoabinhian on the ESEA lineage but as Austronesier said, this is the first time I have seen anybody associate Hoabinhian with ESEA (it actually makes me wonder if "Hoabinhian" should be mentioned in the ESEA description over on the Ancient East Eurasians article but that's another discussion). So I'm going to agree with Austronesier's points on the "AASI/Andamanese/Hoabinhian question". Tagging @Wikiuser1314

Slightly off topic but I think "Ayta" should be changed to "Aeta", since that's how it's usually spelled (including on the main Wikipedia article) and so will be easy for others to gather information on the subject if they went searching for it. Saouirse (留言) 02:59, 15 May 2023 (UTC)

Thank you all for the comments and suggestions. I again uploaded an adapted version taking the above points into account, placing the Hoabinhian/Andamanese lineage as basal to Tianyuan and other ESEA, renaming Ayta into Aeta, and placing a small "ghost" tag to "Ancient Tibetans" (alternatively "Ancient Tibetan (Paleolithic)"). I think the now uploaded version should be "quite accurate" (eg. a "quadrification") and a compromise between the different models so far. - Btw. Yang 2022 is not the only one who puts Hoabinhians closer to ESEA (with Tianyuan being basal): Another phylogenetic tree from Cooke et al. 2021 shows a split between Papuan-Tianyuan, followed by Tianyuan-Hoabinhian, followed by Hoabinhian and the remaining ESEA. There is however no clear cut between the models. Going solely by G25 data, Hoabinhians and Andamanese are closer to Tianyuan and ESEA followed by AASI, and than to Papuan or Aboriginal Australians. Lastly we have six possible models for Papuans from Vallini et al. 2022 regarding their formation, pointing to the possibility of admixture among Papuans from a source related to East Asians and Andamanese, and a source more basal (see the Supplementary information of the Vallini paper). - Either way, I think the now uploaded version taking the points of Austronesier and of the IP into account, is the most corresponding one to the various models presented so far.Wikiuser1314 (留言) 06:06, 15 May 2023 (UTC)

@Wikiuser1314: Thank you for the clarification. What I meant was that only Yang (2022) assigns Hoabinhians and Andamanese to different deeply diverged lineages. There are of course a couple of sources that place Hoabinhians or Andamanese (or both, when both are covered) as close to ESEA. Even the TreeMix models in McColl et al. (2018) do so. So from my part, good job! The next stept would be to reflect this in the prose of the "Ancient East Eurasian" article. After all, images always should only serve to illustrate/visualize the text of Wikipedia articles, and not convey information by themselves. This is why I am the last one to shout "OR" if a good user-generated map or diagram matches one-to-one the statememts of our sourced text. And this is also why I am not a big fan of reproducing diagrams like PCAs, TreeMix and qpGraph trees out of context. –Austronesier (留言) 20:25, 15 May 2023 (UTC)

Thank you again for your reply and comment. Yep, the article still needs some further work and information. I will try to do so in the next few days.Wikiuser1314 (留言) 07:12, 16 May 2023 (UTC)