• Don't want to see ads? Install an adblocker like uBlock Origin or use a Europe-based privacy-friendly browser like Vivaldi or Mullvad.

Admixtools My qpAdm Results Please Share Yours!

1769113223476.png

With those rights:

Mbuti.DG,
Russia_Sunghir_UP.SG
Russia_Yuzhniyoleniyostrov_Mesolithic.AG
Turkey_Marmara_Barcin_N.AG
Italy_Epigravettian.AG.BY.AA
Morocco_Iberomaurusian.AG

To me this is fine because the p-value is nearly 10x 0.05 and the sources are related (similar steppe/ANF portions)
 
Last edited:
Stefano said:
May i suggest you to try Croatia_Cetina, Montenegro_mlba, or Albania_ia?
Click to expand...
I've done some experimentation with more proximal and historically plausible sources, but it comes close, but fails. The reason is likely qpAdm is just rewarding genetic geometry in a generic sense, rather than a literal sense with these lefts and rights. Which is often the case.

With my right panel, I can demonstrate a broad Local southern Italian MBA source and a broad Balkan/Carpathian source with Romania and Bulgaria MLBA. But it cannot demonstrate further precision with these options I currently have.

Perhaps I need a better local Apulian MBA source (fairly low coverage), or even a third source to make it work, or a more particular Balkan/Carpathian source.
 
Last edited:
There is also the fact that when you are trying to model a singleton with more genetically proximal sources, the chances of getting a higher SE is especially the case. It tends to work better with large sample sets. This could also be observed with my modeling of GBR.DG with prioximal France_IA + Denmark_IA vs the one with qh777.

When your Left sources are genetically proximal, qpAdm becomes weakly identified (many mixtures fit nearly equally well), so the weight estimates are unstable and SE inflates. This is amplified for singleton targets.
 
With more tinkering I can get the SEs down.

1769188109699.png
 
Alessio said:
I wonder if this may have some truth to it.

The Normans did come to my part of Italy, and one of my grandparents had a Latinized Norman surname.
Click to expand...
I think the samples are supposed to be Medieval Irish people not full blooded Normans or Anglo-Saxons. IllustrativeDNA has them labeled as such:

1769537571797.png

1769537680480.png
 
Natufian Isolation model

1769828938006.png


1769828984246.png

1769829021721.png
 
Alessio said:
In that case, it may serve as a good proxy for general NW European enrichment perhaps.
Click to expand...
Is it assigning some of your steppe ancestry as medieval Irish I wonder. It's cool, I think. It just reinforces west Eurasian interconnectivity. Specifically European.
 
Using a single source left group vs the listed right panel.

Una faccia una razza!

1770136860716.png


1770136353168.png



1770135704000.png
 
Last edited:
1770322170835.png

1770322211358.png


References:

1770322237515.png
 
Alessio, I had to make some adjustments to the list you provided a few posts ago. I tried it with the classic 3-way and it wasn't working for me. I had to add Italy Epigravettian, Russia YuzhniyOleniyOstrov. Then I had to remove Kotias and replace it with Satsurblia lateUP.

1770873984313.png

1770874002700.png
 
qh777 said:
Alessio, I had to make some adjustments to the list you provided a few posts ago. I tried it with the classic 3-way and it wasn't working for me. I had to add Italy Epigravettian, Russia YuzhniyOleniyOstrov. Then I had to remove Kotias and replace it with Satsurblia lateUP.

View attachment 19197
View attachment 19198
Click to expand...
Look solid!
 
qh777 said:
Alessio, I had to make some adjustments to the list you provided a few posts ago. I tried it with the classic 3-way and it wasn't working for me. I had to add Italy Epigravettian, Russia YuzhniyOleniyOstrov. Then I had to remove Kotias and replace it with Satsurblia lateUP.

View attachment 19197
View attachment 19198
Click to expand...
However, there's one thing to be weary of.

When setting the outgroups, it is important to know which are low-resolution (low SNP count). If it is F2-block, it will filter the SNPs through the lowest one, and collapse the the overall SNP count. There by passing models because of low SNP count. Nevertheless, if it is ALLSNPS = True, it avoids this. It would be worth checking if this method uses F2-Block or ALLSNPS=True. Ultimately, I would just go with ALLSNPS=TRUE as that is common practice in many aDNA papers.

I would assume it is by default on ALLSNPS=TRUE, as to avoid that complication. At least that how I would set it up by default if I created such an tool.
 
Interesting!

1771415275524.png
 
My results for ANF, Steppe, WHG and Taforalt.
7iHc14H.png

The reference populations cover all the bases : Deep Eurasian (Ust-Ishim, Kostenki, Sunghir, MA1), Anatolian Epipaleolithic, CHG (Kotias, Satsurblia), Iran_N, EHG (YuzhniyOleniyOstrov, Sidelkino) , WHG (Bichon) and Marocco Epipalaeolithic for Taforalt.

In a PCA of ancient west-eurasians populations I’m shifted slightly right and higher than Iberia BA, in the direction of the periphery of Europe_LNBA.
6eSRq4b.png
 
French Bell Beaker + Roca Vecchia MBA (Apulia)


1772391804552.png


1772391826458.png
 
Normans?

One of my grandfathers had a Latinized-Norman Surname.

1772471115080.png
 
You must log in or register to reply here.
Back
Top