Posts: 336
Threads: 11
Joined: May 2024
Gender: Male
Ethnicity: Afro-Caribbean
Nationality: American
Y-DNA (P): E-M132 [E-FTB87402]
mtDNA (M): L1b1a3
Country:
(01-02-2025, 03:39 AM)Kale Wrote: - The vast majority of samples in G25 are Eurasian.
- The Eurasian bottleneck was extensive.
- PCA formulates dimensions which get the most 'bang for the buck', which roughly means # of samples multiplied by amount of drift.
Consequently Eurasian diversity is much more represented on G25.
A similar principle applies to admixture. IIRC at K2 you see an 'African' component and a 'Eurasian' component. The African populations closer to the OOA bottleneck get increasing levels of the 'Eurasian' component.
The direct f4-stats aren't showing a lack of mixing between Africans, quite the contrary, it's showing that there have been multiple mixtures between multiple deeply distinct layers of ancestry.
Look at West African y-hg for example, A00, A0, A1a, A1b1, B, D0, and of course E.
South-Africa seems more like an exception, correlating with the more constrained uniparentals (mt-L0, y-hg A1b1)
Chimp.REF SouthAfrica_2000BP.SG Sardinian.DG JuHoanNorth.DG 0.0106440 0.000234200 45.4486 0 1100820
So, do you believe that there are indeed some Aboriginal African populations that are genetically closer to Non-Africans than they are to other Aboriginal Africans? Do you think that fstats are correct in this regard?
Posts: 647
Threads: 3
Joined: Sep 2023
01-02-2025, 05:36 AM
(This post was last modified: 01-02-2025, 05:37 AM by Kale.)
Yes. Of the African populations I've looked at, Dinka seem to be the closest to Eurasians (as you might expect from geography), they obviously have some deeper ancestry though (A-M13, etc.)
Posts: 52
Threads: 0
Joined: Jun 2024
01-02-2025, 05:50 AM
(This post was last modified: 01-02-2025, 06:20 AM by ModusOperandi.)
(01-01-2025, 11:27 PM)Inquirer Wrote: Vahaduo gives the impression that all Aboriginal Africans are more closely related to one another than they are to Non-Africans. Why do you think that Vahaduo gets this right (in my opinion) while f4stats do not? While I believe formal stats majority of time gives an accurate representation of population structure and demographic history, there occasionally are those results that throws you for a loop and are hard to make sense of, more especially with African population. They seem to have a very strange aversion to each other particularly when running the test with a non African. While its easy to just write it off as deep population divergences, substructure, and whatnot, some stats are just hard to justify. For instance in closely related African populations, a setup like (Chimp, Igbo, Yoruba, non African) produces a Z score that is just barely significant(-3.59), for contrast the same setup but with say French and Sardinian in place of Igbo and Yoruba, produce a highly significant Z score(-33.7).
And some of them are these are just down right indefensible, a setup like (Chimp, African, Somali, Korean), is actually positive for most Africans aside from Nilotic Africans like Dinka, meaning they prefer Korean over Somali, but even then something like (Chimp, Somali, Dinka, Korean) produces a highly significant positive Z score(23.5), implying Somalis share much more drift with Koreans than they do with Dinka
Code: pop1 pop2 pop3 pop4 est se z
<chr> <chr> <chr> <chr> <dbl> <dbl> <dbl>
1 Chimp.REF Kenya_Somali.DG Dinka.DG Korean.DG 0.0110 0.000469 23.4
Like what is going on lol?
Remember that setup from earlier? Well replace Yoruba with Kongo and now you're left with a stat that is indicating that Kongo and Korean form a clade and that Igbo shares little to no drift with one over the other
Code: Chimp.REF Igbo.DG Kongo.DG Korean.DG 0.000210 0.000370 0.568
Theres lots of other examples similar to this. I've even heard of situations where members of some African populations will prefer the non African over other members of that same exact population. It just doesn't make sense and to me I feel as though it has to be some kind of technical issue.
If I remember correctly, G25 is PCA based, it seems to do relatively fine with distances in most cases, provided the sample isn't super old
Posts: 336
Threads: 11
Joined: May 2024
Gender: Male
Ethnicity: Afro-Caribbean
Nationality: American
Y-DNA (P): E-M132 [E-FTB87402]
mtDNA (M): L1b1a3
Country:
(01-02-2025, 05:36 AM)Kale Wrote: Yes. Of the African populations I've looked at, Dinka seem to be the closest to Eurasians (as you might expect from geography), they obviously have some deeper ancestry though (A-M13, etc.)
I cannot imagine Nilotes being more closely related to Eurasians than they are to even Niger-Congo peoples (West Africans and Bantus) or Khoisan peoples. Maybe their relatedness to Eurasians is inflated due to Eurasian admixture; Nilotes carry mtDNA M and I, which stem from Eurasians.
Posts: 336
Threads: 11
Joined: May 2024
Gender: Male
Ethnicity: Afro-Caribbean
Nationality: American
Y-DNA (P): E-M132 [E-FTB87402]
mtDNA (M): L1b1a3
Country:
(01-02-2025, 05:50 AM)ModusOperandi Wrote: (01-01-2025, 11:27 PM)Inquirer Wrote: Vahaduo gives the impression that all Aboriginal Africans are more closely related to one another than they are to Non-Africans. Why do you think that Vahaduo gets this right (in my opinion) while f4stats do not? While I believe formal stats majority of time gives an accurate representation of population structure and demographic history, there occasionally are those results that throws you for a loop and are hard to make sense of, more especially with African population. They seem to have a very strange aversion to each other particularly when running the test with a non African. While its easy to just write it off as deep population divergences, substructure, and blah blah blah, some stats are just hard to justify. For instance in closely related African populations, a setup like (Chimp, Igbo, Yoruba, non African) produces a Z score that is just barely significant(-3.59), for contrast the same setup but with say French and Sardinian in place of Igbo and Yoruba, produce a highly significant Z score(-33.7).
And some of them are these are just down right indefensible, a setup like (Chimp, African, Somali, Korean), is actually positive for most Africans aside from Nilotic Africans like Dinka, meaning they prefer Korean over Somali, but even then something like (Chimp, Somali, Dinka, Korean) produces a highly significant positive Z score(23.5), implying Somalis share much more drift with Koreans than they do with Dinka
Code: pop1 pop2 pop3 pop4 est se z
<chr> <chr> <chr> <chr> <dbl> <dbl> <dbl>
1 Chimp.REF Kenya_Somali.DG Dinka.DG Korean.DG 0.0110 0.000469 23.4
Like what is going on lol?
Remember that setup from earlier? Well replace Yoruba with Kongo and now you're left with a stat that is indicating that Kongo and Korean form a clade and that Igbo shares little to no drift with one over the other
Code: Chimp.REF Igbo.DG Kongo.DG Korean.DG 0.000210 0.000370 0.568
Theres lots of other examples similar to this. I've even heard of situations where members of some African populations will prefer the non African over other members of that same exact population. It just doesn't make sense and to me I feel as though it has to be some kind of technical issue.
If I remember correctly, G25 is PCA based, it seems to do relatively fine with distances in most cases, provided the sample isn't super old
1. Based on the examples that you've mentioned, I'm inclined to believe that fstats produce erroneous results in regard to Aboriginal African populations. Perhaps the software was designed initially with Eurasian populations in mind and therefore prefers Eurasian-oriented results and cannot process the wider degree of genetic diversity among Aboriginal African populations.
2. Would you say that the Vahaduo distances that I posted earlier are actual, accurate representations of relatedness between Aboriginal African populations?
ModusOperandi likes this post
Posts: 52
Threads: 0
Joined: Jun 2024
01-02-2025, 06:16 AM
(This post was last modified: 01-02-2025, 07:50 AM by ModusOperandi.)
(01-02-2025, 03:39 AM)Kale Wrote: - The vast majority of samples in G25 are Eurasian.
- The Eurasian bottleneck was extensive.
- PCA formulates dimensions which get the most 'bang for the buck', which roughly means # of samples multiplied by amount of drift.
Consequently Eurasian diversity is much more represented on G25.
A similar principle applies to admixture. IIRC at K2 you see an 'African' component and a 'Eurasian' component. The African populations closer to the OOA bottleneck get increasing levels of the 'Eurasian' component.
The direct f4-stats aren't showing a lack of mixing between Africans, quite the contrary, it's showing that there have been multiple mixtures between multiple deeply distinct layers of ancestry.
Look at West African y-hg for example, A00, A0, A1a, A1b1, B, D0, and of course E.
South-Africa seems more like an exception, correlating with the more constrained uniparentals (mt-L0, y-hg A1b1)
Chimp.REF SouthAfrica_2000BP.SG Sardinian.DG JuHoanNorth.DG 0.0106440 0.000234200 45.4486 0 1100820 I understand that multiple layers of divergent ancestry can skew stats one way or another. But some of these stats simply make no sense. I don't think Kongo would have so much deeply diverged ancestry relative to Igbo, that a stat like (Chimp, Igbo, Kongo, Korean) would be skewed positive(albiet insignificant, but still)
Posts: 52
Threads: 0
Joined: Jun 2024
01-02-2025, 06:40 AM
(This post was last modified: 01-02-2025, 02:19 PM by ModusOperandi.)
(01-02-2025, 06:15 AM)Inquirer Wrote: Would you say that the Vahaduo distances that I posted earlier are actual, accurate representations of relatedness between Aboriginal African populations? Can't speak on the values used to represent distance, but I'd say the relative dscending order of relatedness seems to be all good. In other words I would expect Yoruba to be related to their neighbors first, then to other West Africans, then to other far flung Niger Congo speakers, then to East African pastoralist, then CAHG and SAHG in that order. Additonally I would expect the distance between the HG and non Africans to be bigger than the distance between non HG Africans and non Africans, which is also demonstrated in your post. I'd say its accurate enough
Posts: 336
Threads: 11
Joined: May 2024
Gender: Male
Ethnicity: Afro-Caribbean
Nationality: American
Y-DNA (P): E-M132 [E-FTB87402]
mtDNA (M): L1b1a3
Country:
01-02-2025, 06:43 AM
(This post was last modified: 01-02-2025, 06:44 AM by Inquirer.)
(01-02-2025, 06:40 AM)ModusOperandi Wrote: Can't speak for on values used to represent distance, but I'd say the relative dscending order of relatedness seems to be all good. In other words I would expect Yoruba to be related to their neighbors first, then to other West Africans, then to other far flung Niger Congo speakers, then to East African pastoralist, then CAHG and SAHG in that order. Additonally I would expect the distance between the HG and non Africans to be bigger than the distance between non HG Africans and non Africans, which is also demonstrated in your post. I'd say its accurate enough
This is exactly what I think. It would be too much of a coincidence for Vahaduo's distance results to match the geographic distances between the targets and the sources, as well as the differences in lifestyle between them. I think that Vahaduo has an advantage relative to fstats in this regard, even though it's the simpler program.
Posts: 1
Threads: 0
Joined: Jan 2025
Gender: Male
Ethnicity: Somali
Nationality: Somaliland
Y-DNA (P): E-V32>E-Y219973
Y-DNA (M): T-M70
mtDNA (M): L0a1d1d
mtDNA (P): L3d1a1
Country:
Most Cushitic & Semitic Horners besides outliers in Southern Somalia don’t carry any Niger-Congo ancestry. This is simply just overfitting.
Posts: 336
Threads: 11
Joined: May 2024
Gender: Male
Ethnicity: Afro-Caribbean
Nationality: American
Y-DNA (P): E-M132 [E-FTB87402]
mtDNA (M): L1b1a3
Country:
01-02-2025, 03:15 PM
(This post was last modified: 01-02-2025, 03:21 PM by Inquirer.)
(01-02-2025, 02:28 PM)SubeyriAwal Wrote: Most Cushitic & Semitic Horners besides outliers in Southern Somalia don’t carry any Niger-Congo ancestry. This is simply just overfitting.
It could be trace amounts from the Bantu expansion that was mediated indirectly.
EDIT: It's actually probably due to the lack of Nilotic sources in the OP's image.
Posts: 58
Threads: 0
Joined: Oct 2023
(01-02-2025, 05:50 AM)ModusOperandi Wrote: (01-01-2025, 11:27 PM)Inquirer Wrote: Vahaduo gives the impression that all Aboriginal Africans are more closely related to one another than they are to Non-Africans. Why do you think that Vahaduo gets this right (in my opinion) while f4stats do not? While I believe formal stats majority of time gives an accurate representation of population structure and demographic history, there occasionally are those results that throws you for a loop and are hard to make sense of, more especially with African population. They seem to have a very strange aversion to each other particularly when running the test with a non African. While its easy to just write it off as deep population divergences, substructure, and whatnot, some stats are just hard to justify. For instance in closely related African populations, a setup like (Chimp, Igbo, Yoruba, non African) produces a Z score that is just barely significant(-3.59), for contrast the same setup but with say French and Sardinian in place of Igbo and Yoruba, produce a highly significant Z score(-33.7).
And some of them are these are just down right indefensible, a setup like (Chimp, African, Somali, Korean), is actually positive for most Africans aside from Nilotic Africans like Dinka, meaning they prefer Korean over Somali, but even then something like (Chimp, Somali, Dinka, Korean) produces a highly significant positive Z score(23.5), implying Somalis share much more drift with Koreans than they do with Dinka
Code: pop1 pop2 pop3 pop4 est se z
<chr> <chr> <chr> <chr> <dbl> <dbl> <dbl>
1 Chimp.REF Kenya_Somali.DG Dinka.DG Korean.DG 0.0110 0.000469 23.4
Like what is going on lol?
Remember that setup from earlier? Well replace Yoruba with Kongo and now you're left with a stat that is indicating that Kongo and Korean form a clade and that Igbo shares little to no drift with one over the other
Code: Chimp.REF Igbo.DG Kongo.DG Korean.DG 0.000210 0.000370 0.568
Theres lots of other examples similar to this. I've even heard of situations where members of some African populations will prefer the non African over other members of that same exact population. It just doesn't make sense and to me I feel as though it has to be some kind of technical issue.
If I remember correctly, G25 is PCA based, it seems to do relatively fine with distances in most cases, provided the sample isn't super old
Regarding your (Chimp, Somali, Dinka, Korean) example, even if we were to take the Eurasian bottleneck explanation at face value for the extremely positive Z score, it still wouldn't absolve fstats of being fundamentally biased or defective in elucidating African phylogeny. It would just mean that fstats are heavily distorted by extreme drift in either one of the test populations or in the target populations, the underlying implication of the stat itself - Somalis being massively more related to Koreans than Dinkas - would still be bogus.
This is something that I noticed a while ago myself when I first tried tentatively playing around with Admixtools, and I've honestly never seen a good explanation or at least work-around solution devised for it. The fact that fstats seem to have trouble with Africans also makes me skeptical in regard to a lot of deeper population modeling and our understanding of concepts like Basal Eurasian, Africa/Eurasian gene-flow, or even some of these extremely deep divergence estimates within Africa (are San really some 300k+ years old diverged branch relative to the rest of humanity?). Honestly, the fact that professional geneticists have never corrected this defect in Admixtools, or even seem to really notice that it's much of an issue, is frankly bizarre, given that resolving it could open up a lot of new interpretive insights regarding overall human population history.
I wonder if someone emailed Nick Patterson or someone else at the Reich lab (maybe Lazaridis, he's still active on Twitter I think), if that might stimulate some additional investigation into this issue.
Posts: 193
Threads: 5
Joined: Oct 2023
(01-02-2025, 06:05 AM)Inquirer Wrote: (01-02-2025, 05:36 AM)Kale Wrote: Yes. Of the African populations I've looked at, Dinka seem to be the closest to Eurasians (as you might expect from geography), they obviously have some deeper ancestry though (A-M13, etc.)
I cannot imagine Nilotes being more closely related to Eurasians than they are to even Niger-Congo peoples (West Africans and Bantus) or Khoisan peoples. Maybe their relatedness to Eurasians is inflated due to Eurasian admixture; Nilotes carry mtDNA M and I, which stem from Eurasians.
They also have Y E-M35 and are cattle pastoralists so surely they should have some Cushitic admixture. I also think they have Niger-Congo ancestry and not the other way around, through Kordofanians.
It's interesting that this possibly admixed profile fits well for Nilo-Saharans when linguists propose a Paleolithic divergence. But also almost all Nilo-Saharan speakers are agriculturalists (sorghum farmers or catte pastoralists) and agriculture would make sense for the spread of the family (hard to imagine anything else with also independent adoption of agriculture in each branch) so must be something else affecting the divergence or linguists are wrong again.
Posts: 4
Threads: 0
Joined: Jan 2025
Country:
(12-31-2024, 03:01 AM)Inquirer Wrote: (12-31-2024, 02:22 AM)Tomenable Wrote: (12-31-2024, 02:18 AM)Inquirer Wrote: Then how do you explain the models produced by Vahaduo in my above post?
Your model for the Dinka has a very bad distance, over 10.
This indicates that the model is not realistic.
The model below has distances from 0.07 to 0.10. The distances aren't very large when considering that what they reflect are divergences in ancestry that date back to at least the Mesolithic (Green Sahara Period).
![[Image: attachment.php?aid=3095]](https://genarchivist.net/attachment.php?aid=3095)
In other words, what matters is that ~50% of the ancestry in Nilotes pulls towards Mota and ~50% pulls towards Niger-Congo peoples. This indicates that the Dinka descend from a lineage that was closely related to Mota, as well as a lineage that was closely related to Niger-Congo peoples.
This is reflected in their haplogroups: they carry yDNA B-M60, as do the Hadza (in whom Mota-like ancestry peaks); and they carry mtDNA L1 and L2, as do Niger-Congo peoples. although I dont deny some Dinkas have some west african like admix, it doesn't really make sense to me to use mota as the measure for how East African ancestrally the dinka are, because they are the more genetically East African shifted ones, this also doesn't really take into account the sahg like ancestry mota carries or even little bit of eurasian. Even more interesting is that East African nilotic (what folks online call "proto nilotic" even tho proto nilotic is really more of a linguistic term) like ancestry and west african ancestry act like polar opposites. Which is another thing that makes me skeptical abt the idea that nilotes like the Dinka or Nuer are "significantly mixed w niger Congos". I made a run with the proto mota cord and I had some interesting results comparing my own g25 cords to measure east African ancestry in mota vs the proto mota cord which was supposed to be the simulated version of mota.
Here were the cords I used: ETH_Mota,-0.515619,0.041637,-0.003394,-0.001292,-0.00277,-0.011713,0.054992,-0.052382,0.090604,-0.094581,-0.013478,0.003297,-0.031962,-0.001927,0.023615,-0.031291,0.031944,0.043454,-0.002388,-0.00963,-0.001248,0.00643,-0.005176,-0.003133,0.004071
Yoruba:HGDP00944,-0.628304,0.061947,0.021496,0.01615,-0.001539,0.003347,-0.042067,0.049152,-0.050108,0.030616,-0.000974,0.003897,0.023191,-0.006468,0.00475,-0.004773,0.005737,-0.007601,0.00729,-0.002876,0.004991,0.000495,-0.006409,0.002771,-0.005029
(Mines) I am dinka btw, lol
BiorMalual_scaled,-0.574807,0.05687,-0.000754,-0.015827,-0.007386,0.001116,-0.018331,0.017307,0.087127,-0.119182,-0.020299,0.026377,-0.049207,0.001514,0.014115,-0.008088,0.02386,-0.012162,0.023128,-0.015883,0.000499,0.000247,-0.003697,-0.003494,0.004071
Posts: 4
Threads: 0
Joined: Jan 2025
Country:
(01-02-2025, 06:05 AM)Inquirer Wrote: (01-02-2025, 05:36 AM)Kale Wrote: Yes. Of the African populations I've looked at, Dinka seem to be the closest to Eurasians (as you might expect from geography), they obviously have some deeper ancestry though (A-M13, etc.)
I cannot imagine Nilotes being more closely related to Eurasians than they are to even Niger-Congo peoples (West Africans and Bantus) or Khoisan peoples. Maybe their relatedness to Eurasians is inflated due to Eurasian admixture; Nilotes carry mtDNA M and I, which stem from Eurasians.
its only for Khoisan but its because the Khoisan are very divergent , and thats pretty much the case for even most ssa, but nilotes are way closer related to west africans then Eurasian
Posts: 336
Threads: 11
Joined: May 2024
Gender: Male
Ethnicity: Afro-Caribbean
Nationality: American
Y-DNA (P): E-M132 [E-FTB87402]
mtDNA (M): L1b1a3
Country:
(01-07-2025, 08:00 AM)Biormalu Wrote: (12-31-2024, 03:01 AM)Inquirer Wrote: (12-31-2024, 02:22 AM)Tomenable Wrote: Your model for the Dinka has a very bad distance, over 10.
This indicates that the model is not realistic.
The model below has distances from 0.07 to 0.10. The distances aren't very large when considering that what they reflect are divergences in ancestry that date back to at least the Mesolithic (Green Sahara Period).
![[Image: attachment.php?aid=3095]](https://genarchivist.net/attachment.php?aid=3095)
In other words, what matters is that ~50% of the ancestry in Nilotes pulls towards Mota and ~50% pulls towards Niger-Congo peoples. This indicates that the Dinka descend from a lineage that was closely related to Mota, as well as a lineage that was closely related to Niger-Congo peoples.
This is reflected in their haplogroups: they carry yDNA B-M60, as do the Hadza (in whom Mota-like ancestry peaks); and they carry mtDNA L1 and L2, as do Niger-Congo peoples. although I dont deny some Dinkas have some west african like admix, it doesn't really make sense to me to use mota as the measure for how East African ancestrally the dinka are, because they are the more genetically East African shifted ones, this also doesn't really take into account the sahg like ancestry mota carries or even little bit of eurasian. Even more interesting is that East African nilotic (what folks online call "proto nilotic" even tho proto nilotic is really more of a linguistic term) like ancestry and west african ancestry act like polar opposites. Which is another thing that makes me skeptical abt the idea that nilotes like the Dinka or Nuer are "significantly mixed w niger Congos". I made a run with the proto mota cord and I had some interesting results comparing my own g25 cords to measure east African ancestry in mota vs the proto mota cord which was supposed to be the simulated version of mota.
Here were the cords I used: ETH_Mota,-0.515619,0.041637,-0.003394,-0.001292,-0.00277,-0.011713,0.054992,-0.052382,0.090604,-0.094581,-0.013478,0.003297,-0.031962,-0.001927,0.023615,-0.031291,0.031944,0.043454,-0.002388,-0.00963,-0.001248,0.00643,-0.005176,-0.003133,0.004071
Yoruba:HGDP00944,-0.628304,0.061947,0.021496,0.01615,-0.001539,0.003347,-0.042067,0.049152,-0.050108,0.030616,-0.000974,0.003897,0.023191,-0.006468,0.00475,-0.004773,0.005737,-0.007601,0.00729,-0.002876,0.004991,0.000495,-0.006409,0.002771,-0.005029
(Mines) I am dinka btw, lol
BiorMalual_scaled,-0.574807,0.05687,-0.000754,-0.015827,-0.007386,0.001116,-0.018331,0.017307,0.087127,-0.119182,-0.020299,0.026377,-0.049207,0.001514,0.014115,-0.008088,0.02386,-0.012162,0.023128,-0.015883,0.000499,0.000247,-0.003697,-0.003494,0.004071
I've always considered Mota-like ancestry to be more East African shifted than Nilotic ancestry, because it's farther from Niger-Congo peoples (West Africans & Bantus) than Nilotic samples are according to G25 and PCA charts that I've seen. For example, in Figure 1a of Llorente et al (2015), you can see that Mota is farther from Niger-Congo peoples than Nilotic peoples are. Additionally, Mota's homeland and those of the peoples with the most Mota-like ancestry are geographically farther from the homeland of Niger-Congo peoples (West Africa) than the homelands of Nilotic peoples.
I think that during the Green Sahara period, before the ancestors of Niger-Congo peoples migrated into West Africa, that there occurred an introgression of Proto-Niger-Congo ancestry into Proto-Nilotic peoples via Proto-Niger-Congo females; this is consistent with Nilotic peoples carrying Maternal Haplogroups L1 & L2, which are carried predominantly by Niger-Congo peoples. Alternatively, perhaps Proto-Nilotic peoples took Bantu females when Bantus arrived in East Africa during the Bantu Expansion.
A good way to determine which scenario likely occurred would be to determine the divergence dates of the L1 and L2 branches that Nilotic peoples carry relative to those that Niger-Congo peoples carry.
|