Paleogenomics of Upper Paleolithic to Neolithic European hunter-gatherers

**Kunig** · 02-07-2023, 04:08 AM

Originally Posted by Ryukendo

Its really striking that both analyses show basically the same proportions (+-4%) and almost no preference between the two even in a rotating outgroup setup. This can show that the incoming Battle Axe had EEF related to both GAC and local megalithics almost evenly, but really what it shows is that there is almost no drift that differentiates Swedish megalithics and GAC, nothing specific to each population that is shared with Battle Axe anyway that is detectable in qpAdm, so that the two models can be clearly distinguished in p-value (e.g. differ in p-value by more than 10^1-2).

I don't want to get into 'my set-up is better' contest, because there are no Laws governing. However, I noted Angle's set-up lacks more recent European individuals, and was not using a rotating method, hence the two alternatives were not explicitly examined.

We may also ask - what happens if we use both ? GAC is not needed

left pops:
SWE_BAC.SG
POL_Globular_Amphora
SWE_Megalithic
EST_Corded_Ware

best coefficients: 0.028 0.198 0.774
Jackknife mean: 0.038621794 0.188676160 0.772702046
std. errors: 0.140 0.129 0.03
Tail prob 0.187945

Moreover, there was very little GAC in the Estonia or Finland route, hence the debate is somewhat moot. This highlights the importance of contextual awareness.

BAx.png

**Kale** · 02-07-2023, 04:10 AM

Originally Posted by Ryukendo

Anglesqueville, could you run this f4 stat for us as a favor?
F4 (Mbuti.SG, Swedish_Battle_Axe.SG; Sweden_Megalithic.SG, Poland_Globular_Amphora.SG)
(I'm assuming that there are .SG versions for these, since the middle two are Scandinavian and presumably come from Eske Willerslev's group? And Globular Amphora should have at least a few .SG individuals from Reich?)

Thanks if you can! In the rotating outgroup setup Kunig ran both qpAdms passed, so it seems like whatever preference existed it was not that one-sided, but this is just another way to confirm.

Here's a bunch of such F4. GAC <> GAC barely reaches significance, so there's no way it would be detectable with formal stats in a quantity of 25% or thereabouts.

**Aleph** · 02-07-2023, 05:12 AM

Originally Posted by Billyh

So it is possible that ZK's neanderthal comes from a similar area as Crown's Neanderthal to the point where we couldn't distinguish them in small amounts.
I say this because IUP Euros (BK, Oase1) show a greater affinity to European Neanderthals than contemporary Eurasian groups do, and have more Neanderthal than East Eurasians.

There seems to have been a selection against Neanderthal ancestry among Eurasians who had admixture, even among those without Basal Eurasian DNA, like East Eurasians. I think this is also linked to relatively short Neanderthal segments today. This alone needn't mean there wasn't an additional admixture. For that test, you'd have to look at the details of the Neanderthal ancestry itself. Regarding your first point, if it got admixture from something very similar to the Crown Eurasian Neanderthal, it would be pulling the line really thin regarding its classification as Crown or non-Crown. The paper that introduced it put the Zlaty Kun presented it's phylogeny such that it was in a clade with other Eurasians after the Neanderthal admixture. I don't know if this leaves enough scope for it being an outgroup to Crown Eurasians. I don't know how closely they looked at the differences in Neanderthal profiles of the ancestry, but given what we have, it will be a tall order to show that Zlaty Kun isn't a Crown Eurasian. Link to the paper.

**Ryukendo** · 02-07-2023, 05:19 AM

Originally Posted by Kale

Here's a bunch of such F4. GAC <> GAC barely reaches significance, so there's no way it would be detectable with formal stats in a quantity of 25% or thereabouts.

I think you've found evidence that strongly undermines Allentoft's conclusions... If this is true, GAC is not really a very coherent population, and things like "how many GAC individuals are in the IBD graph and are they actually all GAC" will strongly affect their clustering results. And even after that their qpAdm over the clusters they defined will differ from the qpAm here because they included more recent, non-GAC individuals in their "GAC cluster", which will strongly affect the stats in the absence of GAC themselves having so much GAC-specific drift....

Originally Posted by CopperAxe

You mean Angles' results where all Scandinavia genomes are mixture of Battle Axe + Megalithic Ansarve? I dont think rotation strategies or the usage of larger sample pools were used by Angles, not sure though.

It isn't just with Scandinavians where Swe_TRB doesn't show up, it is basically all their samples. Even the Rathlinn samples were mostly Polish GAC with a tiny amount of SW European/Italian EEF ancestry.

Not related to this topic in particular but in Allentoft there is also an issue in that Scandinavia 4000_3000bp pool has a bunch of iron age samples and from memory even a medieval one, which I think hampers their results with historic implications. Also some of the pools include samples with wildly different genetic profiles but had recent common ancestry.

**Ryukendo** · 02-07-2023, 05:27 AM

Originally Posted by Kunig

I don't want to get into 'my set-up is better' contest, because there are no Laws governing. However, I noted Angle's set-up lacks more recent European individuals, and was not using a rotating method, hence the two alternatives were not explicitly examined.

We may also ask - what happens if we use both ? GAC is not needed

left pops:
SWE_BAC.SG
POL_Globular_Amphora
SWE_Megalithic
EST_Corded_Ware

best coefficients: 0.028 0.198 0.774
Jackknife mean: 0.038621794 0.188676160 0.772702046
std. errors: 0.140 0.129 0.03
Tail prob 0.187945

Moreover, there was very little GAC in the Estonia or Finland route, hence the debate is somewhat moot. This highlights the importance of contextual awareness.

BAx.png

If this is the case, this means that GAC and Ansarve have very little, almost no differentiation from each other (c.f. the f4s). Or this may not be the case (c.f. the standard errors of your current qpAdm, and the p-vals in your previous rotating outgroup qpAdm, which differ very little, by less than one order of magnitude). Either way, Allentoft is most likely wrong ("almost none from Ansarve") due to their source populations in qpAdm being algorithmically generated by IBD clustering as opposed to all coming directly from the GAC itself, which means you're right.

What is the circle labeled '10'?

**Kunig** · 02-07-2023, 05:34 AM

Originally Posted by Ryukendo

If this is the case, this means that GAC and Ansarve have very little, almost no differentiation from each other (c.f. the f4s). Or this may not be the case (c.f. the standard errors of your current qpAdm, and the p-vals in your previous rotating outgroup qpAdm, which differ very little, by less than one order of magnitude). Either way, Allentoft is most likely wrong ("almost none from Ansarve") due to their source populations in qpAdm being algorithmically generated by IBD clustering as opposed to all coming directly from the GAC itself, which means you're right.

What is the circle labeled '10'?

The dot-shaded 10 is the Lublin- Podol group of CWC , the circle is Kristiansen's hypothesised 'departure point' for CWC.

Agree, categoric claims need to be made with caution "no ancestry...' & '.... all ancestry'', because history played out in more subtle ways, and we need to acknowlegde the limitations of our tools

**VladimirTaraskin** · 02-07-2023, 05:41 AM

Originally Posted by Ryukendo

If this is the case, this means that GAC and Ansarve have very little, almost no differentiation from each other (c.f. the f4s). Or this may not be the case (c.f. the standard errors of your current qpAdm, and the p-vals in your previous rotating outgroup qpAdm, which differ very little, by less than one order of magnitude). Either way, Allentoft is most likely wrong ("almost none from Ansarve") due to their source populations in qpAdm being algorithmically generated by IBD clustering as opposed to all coming directly from the GAC itself, which means you're right.

What is the circle labeled '10'?

"10" does not refer to the circle, but to the shaded area in the circle. These are Subcarpathian CWC groups. And actually the circle is the center of the CWC formation according to Miroslav Buchvaldek.
As for the GAC, they were in Belarus. Moreover, according to archaeologists, they came there not from Poland, but from Ukraine. There were especially many GAC along the Neman River, this is western Belarus.

**Ryukendo** · 02-07-2023, 05:47 AM

Originally Posted by Kale

I'm assuming you've read deeper into this subject than I. My understanding based on skimmed reading is that a sample with missing data in a region has it's nearby regions analyzed and compared to a set of references to determine the most likely values for the missing data, yes? In that case, the references are mostly going to be modern or at the very least much younger when samples of great age are analyzed. To that end how biased are the results going to be towards mutation/selection events that post-date the sample? Was there any sort of trial such as... take Ust-Ishim (who is superb coverage) > remove portions of the data > run the imputation program > compare results?

Originally Posted by Ryukendo

People do do this, its standard practice. The match rate used to be >99% already pre-2020 for aDNA with coverage >0.4, now its even better, and with increasing size and geographic diversity of modern biobanks/genome datasets its only going to get even better.

The trick here is to see that the gaps between snps are often very short, and so the haplotypes that cover the gap are extremely old and so only get shuffled around among populations and don't really "disappear" so to speak (i.e. drift to freq of 0 everywhere, or gain so many new mutations that the freq of the original haplotype is now 0. Most genetic drift occurs instead when haplotypes get replaced by other haplotypes, or you get a change in freq between them). So with larger and larger sizes of biobanks/genome datasets, sooner or later almost every single short haplotype at every single locus will be represented in the biobank/dataset (i.e., we have near-saturation). Yes, even Ust-Ishim haplotypes, he will have combinations of haplotypes that are low-frequency today and not strongly overrepresented among any given population today but he will not have haplotypes that are completely missing today.

Issues arise when the alleles in the ancient genome you are imputing are supposed to be rare ones, and indeed sites around snps with low minor allele frequency among moderns are always the ones that are the least properly imputed and are most biased toward the majority haplotype in the reference panel. As your reference dataset size increases though this problem progressively disappears. And if you are worried, you can always exclude the sites with lowest minor allele frequency/you detect the strongest bias toward the majority haplotype, and this still gives you a lot of data to work with.

Edit:
P.S. thinking about it, Ust-Ishim had recent Neanderthal ancestors and much higher Neanderthal ancestry and so may have Neanderthal haplotypes that are missing today, so this is a special case. But snps that fall in the Neanderthal segments that are missing today will have extremely low minor allele frequency in the reference panel and so will get purged in the quality control step, which is very common in imputation-based pipelines, and you end up only working with human segments from the Ust-Ishim genome.

Something else, in case it wasn't clear--you can also impute 1240k genomes to .SG equivalents, i.e. fill in the gaps between the snps to recover the whole genome. Not sure about the tests they ran on Glimpse, but a very good way to test the accuracy of the genotyping-->sequencing imputation is to check the imputed genomes against reads in the BAM files (i.e. parts of the genome that are covered by reads but are not among the 1240k snps), the accuracy has become quite impressive even for this comparison nowadays.

**Kunig** · 02-07-2023, 06:05 AM

Originally Posted by VladimirTaraskin

"10" does not refer to the circle, but to the shaded area in the circle. These are Subcarpathian CWC groups. And actually the circle is the center of the CWC formation according to Miroslav Buchvaldek.
As for the GAC, they were in Belarus. Moreover, according to archaeologists, they came there not from Poland, but from Ukraine. There were especially many GAC along the Neman River, this is western Belarus.

the Pripet river is not in the 'sub-carpathian region".
The presence of GAC in Belarus is fleeting. In any case, Belarus is not in the East Baltic. Obvoiusly, GAC is not from Ukraine, but migrated there from Poland. We should refer to Marzena Szmyt , not vlad

**VladimirTaraskin** · 02-07-2023, 06:13 AM

Originally Posted by Kunig

the Pripet river is not in the 'sub-carpathian region".
The presence of GAC in Belarus is fleeting. In any case, Belarus is not in the East Baltic. Obvoiusly, GAC is not from Ukraine, but migrated there from Poland. That's a comical idea, even for vlad

There are many different classifications. One of them is as follows: The Subcarpathian archaeological culture includes four local groups:
Krakow-Sandomierz;
Lyubashevskaya;
Upper Dniester;
Podolskaya.

Thread: Paleogenomics of Upper Paleolithic to Neolithic European hunter-gatherers

Thread Tools

Display

The Following 3 Users Say Thank You to Kunig For This Useful Post:

The Following 2 Users Say Thank You to Kale For This Useful Post:

The Following User Says Thank You to Ryukendo For This Useful Post:

The Following User Says Thank You to Ryukendo For This Useful Post:

The Following 2 Users Say Thank You to Kunig For This Useful Post:

The Following 3 Users Say Thank You to VladimirTaraskin For This Useful Post:

The Following 2 Users Say Thank You to Ryukendo For This Useful Post:

Similar Threads

Hunter-Gatherers&Neolithic (scaled)

Upper Palaeolithic hunter-gatherers from San Teodoro cave Sicily (discussion)

Neolithic farmers coexisted with hunter-gatherers for centuries in Europe

Chahoud Paper on Stone Age Armenian Economies: Upper Paleolithic through Neolithic

Eurogenes K15 Results For Upper Paleolithic, Mesolithic and Neolithic Individuals

Posting Permissions