the 1$ illustrativeDNA is here

[Geba]

Guess my ethncity.

[djipon]

[AndreiDNA]

Clustering of humans, apes, Neanderthals, and Denisovans based on f2 statistics. FST becomes inflated in lineages with very low heterozygosity, such as Neanderthals, Denisovans, and apes, because it is normalized by within-population diversity, making it poorly suited for phylogenetic clustering in this context. In contrast, f2 directly measures shared genetic drift and is far less sensitive to low heterozygosity, producing more stable and interpretable relationships.

[Peterski]

Clustering of humans, apes, Neanderthals, and Denisovans based on f2 statistics. FST becomes inflated in lineages with very low heterozygosity, such as Neanderthals, Denisovans, and apes, because it is normalized by within-population diversity, making it poorly suited for phylogenetic clustering in this context. In contrast, f2 directly measures shared genetic drift and is far less sensitive to low heterozygosity, producing more stable and interpretable relationships.

This clustering makes no sense because according to it Neanderthals and Denisovans are closer to apes than to humans.

BTW - AV2 is not an Early Medieval Pole, this sample is from Hungary.

[AndreiDNA]

This clustering makes no sense because according to it Neanderthals and Denisovans are closer to apes than to humans.

The thing is according to Fst and F2, they are.
That happens because most ancient-DNA datasets are built from SNP panels, such as the 1240k panel, and the SNPs are chosen because they are polymorphic among modern humans, especially West Eurasians. Sites where modern humans differ from the ancestral state are overrepresented, while Sites where archaic hominins or apes differ from each other are underrepresented. This causes apes, neanderthals, and various other non humans to cluster tightly with one another in F2, because they are sharing many of the same ancestral alleles.
If I added wolf to this it'd also cluster with apes and neanderthals. Same with any other non human.

But i dont really have to explain this to you and everybody else, do I? My post is a bit of an IQ test isnt it.

[Peterski]

Okay, I get it. I wasn't aware that the 1240K panel is useless for archaic hominins as I don't really deal with archaic DNA.

And how do you explain the fact that when you upload apes to GEDmatch, they score like Sub-Saharans in calculators?
(although the number of SNPs used in calculation is really small)
I didn't test it myself but few years ago someone posted the results of apes in Eurogenes K13/K15 calculators and Oracles.
Neanderthals and Denisovans also score SSA in Eurogenes calculators.

[Peterski]

Ok I just checked it.

Chimp_HO from Reich's dataset in Eurogenes K36:

Sample Chimp_HO.txt
Amerindian 0.00
Arabian 0.00
Armenian 0.00
Basque 0.00
Central African 0.00
Central Euro 0.00
East African 0.00
East Asian 0.00
East Balkan 0.00
East Central Asian 0.00
East Central Euro 0.00
East Med 0.00
Eastern Euro 0.00
Fennoscandian 0.00
French 0.00
Iberian 0.06
Indo-Chinese 0.00
Italian 0.00
Malayan 0.00
Near Eastern 0.00
North African 0.00
North Atlantic 0.51
North Caucasian 0.00
North Sea 0.00
Northeast African 0.00
Oceanian 1.49
Omotic 7.81
Pygmy 80.52
Siberian 0.00
South Asian 0.00
South Central Asian 0.00
South Chinese 1.59
Volga-Ural 0.65
West African 7.38
West Caucasian 0.00
West Med 0.00

Is it because the panel is built especially for Eurasians?

[AndreiDNA]

Okay, I get it. I wasn't aware that the 1240K panel is useless for archaic hominis as I don't really deal with archaic DNA.

And how do you explain the fact that when you upload apes to GEDmatch, they score like Sub-Saharans in calculators?
(although the number of SNPs used in calculation is really small)
I didn't test it myself but few years ago someone posted the results of apes in Eurogenes K13/K15 calculators and Oracles.
Neanderthals and Denisovans also score SSA in Eurogenes calculators.

Since most of these calculators use the same SNPlists as the harvard datasets, they fall victim to the same issue. The big difference, however, is that in my analysis with F2 I selected multiple "non human" anchors, hence why you see the variability even among non humans, why you see denisovans and neanderthals clustering together, and separate from the apes, why you see closely related apes (Chimps and Gorillas) clustering together, rather than - for example, chimps clustering with macaques. The phylo tree I posted really is brilliant. With G25 or GEDmatch calcs, all these apes and archaic humans wouldn't even make a separate clusters with themselves - they'd just plot with Sub Saharans. Thats because GEDmatch calcs and G25 don't anchor archaic hominid or Apes. There are actually some GEDmatch calcs that do anchor archeic humans though, I know some of the MDLP ones do. These calculators are goated. With them you can tell apart blacks and neanderthals. But if you ran a monkey through these calcs it would just score like a neanderthal.

[Peterski]

Interesting, I will check which MDLP calcs anchor archaic humans.

[17571imre]