Try Boncuklu ANF too. That group seems to have low Natufian.Quote:
Originally Posted by Opie
Printable View
Try Boncuklu ANF too. That group seems to have low Natufian.Quote:
Originally Posted by Opie
I will :thumb001:Quote:
Originally Posted by Gannicus
https://i.postimg.cc/MZnSZhcJ/image.png
I'm experimenting again with modern populations.
What reference populations did you use? I’m still trying to figure out which ones work best for modern samples, so far my results have been a bit inconsistent.
Admixture modeled some ANFs and a Greek EEF:
Greece NeaNikomedeia EN:
Attachment 142786
Catalhoyuk:
Attachment 142787
Boncuklu:
Attachment 142788
Barcin:
Attachment 142789
And this is a Barcin ANF individual that I have used myself with my own models. His/her genome and mine go together well.
Attachment 142790
My results with Bar25:
Attachment 142791
And another run with a few tweaks in the references:
Attachment 142785
G25:
Target: Gannicus_MergedFile_officialDavidski_scaled
Distance: 0.0450% / 0.04499545
47.2 Russia_Samara_EBA_Yamnaya
41.6 Turkey_Marmara_Barcin_N:Bar25__BC_6366__Cov_99.86%
11.2 Luxembourg_Mesolithic:Loschbour_snpAD__BC_6075__Co v_92.30%
Edit: Had to fix the screenshots.
I've produced successful models using Mesolithic/Neolithic sources like EHG,AHG,WHG, Natufian etc. And I'm still experimenting, but so far when I use the Karelia EHG the model fails. The model works with Veretye EHG though. And this latest model is with a slight adjustment to the references used in the previous runs.
Attachment 142793
This modern one is the closest. I don't know how accurate these are due to no Irish in the samples. I think Norway.HO is the best to use because there are more samples. Also Norway.HG is too close to me and seems too Isles shifted. With the selection of populations there is not a lot of others I can use. All other combinations come out FALSE.
https://i.postimg.cc/tCHZnNMt/Screen...-20-111350.png
I tried English with Spanish but this is an example of a FALSE model for me.
https://i.postimg.cc/xjvPpRBq/Screen...-20-133906.png
Here is my fst list with mostly European populations. I think the Norwegian.HG and also the Polish one might be outliers (both are only 1 sample). So I think Norway.HO is more accurate for distance and to model with.
https://i.postimg.cc/d1NMP6TM/FSTDist1.png
https://i.postimg.cc/jq9VM8bP/FSTDist2.png
https://i.postimg.cc/fW3WWSW-7/FSTDist5shb3.png
https://i.postimg.cc/wB9n1ZX8/FSTDist3.png
https://i.postimg.cc/hjjwrmZ2/FSTDist4.png
Your model is close to your G25. Good to know. I'll try this for myself.Quote:
Originally Posted by Gannicus
I'm doing some ancient models
I want to see what my medieval mix is so I'm going with this. I know I'm a sort of Gaelic/Scandinavian mix from G25. Although I rationalise this as having high Steppe because I think G25 and also qpAdm is inflating Scandinavian. Is there any samples that I could use to resolve this?
This model was accepted.
https://i.postimg.cc/vHdjh3qq/Screen...-20-221011.png
With Anglo-Saxon it was rejected.
https://i.postimg.cc/g22gt5vD/Screen...-20-224317.png
I'll experiment further with models. There is a lack of Irish ancient samples and no modern day Irish reference so I feel a bit frustrated about that. I wish there was more Gaelic samples. I'll see if there are Picts and if they would be more suitable.
I'll look at some bronze age models over the coming week.
I haven't had much luck using two ancient Northwest Europeans as sources. Even if I use a third "Southern" population. It does the same thing you are seeing. I guess qpAdm struggles to figure out where the Steppe is coming from.Quote:
Originally Posted by Grace O'Malley
This one is working for me. Scotland_Viking is just a Gael so this appears good. More realistic as well. Irish are mostly Gael with a little bit of Germanic.Quote:
Originally Posted by Gannicus
https://i.postimg.cc/0ytKZXf7/Screen...-20-231450.png
I'm not satisfied with the standard errors that I get. I might be able to experiment with different references to fix it. I noticed Drb was able to model with two or more NW Euros and get decent standard errors. He said it involved using related Iron Age and Bronze Age populations in the references. I'm not sure if he kept the older West Eurasian populations in his references though such as AHG (Pinarbasi), EHGs etc.Quote:
Originally Posted by Grace O'Malley
Attachment 142812
Edit: This one isn't too terrible for SEs but still higher than I'd like. Especially since the Gallo-Roman admixture isn't small. I hear that a higher standard error is more excusable for smaller admixture percentages like 5-10% due to more uncertainty, because it's small.
So that bit at the side with percentage is standard errors 21.38%? Yes most likely the references are the key.Quote:
Originally Posted by Gannicus
Yes, that plus or minus X% to the left of the admix bar is the standard error. The interface doesn’t show it but there is another metric with these models and that is Z-Scores. The Z-score is your admixture divided by the standard error. Some hobbyists think that a Z-score of less than 3 means you don’t have that source. Through email correspondence with Andrei I don’t think that’s correct. And the reason I think that is because there have been published qpAdm admixture models in academic papers with Z scores less than 3. Btw I learned this stuff with the help of AndreiDNA, plus looking around myself and discussions on another forum.Quote:
Originally Posted by Grace O'Malley
This model would satisfy that hobbyist that held the belief that Z-Scores of 3 or better should only be considered:
Attachment 142815
My 20.55% admixture of EGY TIP divided by the standard error of 6.32% would equal a Z score of ~3.25.
It seems the general approach in peer-reviewed work is p-value + archaeological plausibility* > strict Z cutoff.
*I think that this can be interpreted with nuance as well
Attachment 142860Quote:
Originally Posted by Xavier727
For this model, I used these 30 references:
Attachment 142861
Attachment 142862
Attachment 142863
Quote:
Originally Posted by Gannicus
Thanks, mate, any idea why the Human Origins dataset is so limited, there's only like two Germanic populations to choose from, it seems a bit restrictive, which could affect reliability. I’ve been getting some weird FST results, when using Scottish as the target, French, Spanish and Hungry are showing closer distances than Orcadian.
That is one of my criticisms of that dataset is the lack of Germans. No Germans, Austrians etc. Have you tried doing FST with just individuals from a population?Quote:
Originally Posted by Xavier727
Quote:
Originally Posted by Gannicus
There aren’t very many Celtic populations either, which isn’t ideal since it could probably skew the results. I haven’t gone through the individual samples yet, I’ll check them now
I wonder where the russian.ho samples are from. Is it just the western part of Russia?
This is the best for me so far. I'm going by the percentage result on the left. Could someone let me know? I'm just trying to understand what is a good result in these models.
https://i.postimg.cc/ZKq60Qz1/Screen...-22-165020.png
High Yamnaya. How feasible is this model?
https://i.postimg.cc/NMnJtzMv/Screen...-23-144018.png
I think it's a very good model, considering the low errors.Quote:
Originally Posted by Grace O'Malley
Looks fine to me. P-value is way above 0.05.Quote:
Originally Posted by Grace O'Malley
You could experiment with different WHGs in the references and see if your own WHG admix increases.
Here are some of my recent results
https://i.postimg.cc/R0QXS3Cs/image.png
https://i.postimg.cc/V6Kgt3CX/image.png
https://i.postimg.cc/kX0cWNBY/image.png
there is still something to improve here
https://i.postimg.cc/0QXGjDTf/image.png
VK202 from Scotland is apparently a local Pictish individual with no Scandinavian ancestry. So, I did some models of myself with him.
Here is how he is labeled in the dataset:
Attachment 142986
Here you all will see a bit of a difference between G25 and qpAdm.
Attachment 142979
Distance to: Gannicus_MergedFile_officialDavidski_scaled
0.02821931 Scotland_Viking:VK202__AD_950__Cov_57.86%
0.18920718 Lebanon_EjJaouze_Phoenician.SG
Target: Gannicus_MergedFile_officialDavidski_scaled
Distance: 0.0228% / 0.02275620
91.8 Scotland_Viking
8.2 Lebanon_EjJaouze_Phoenician.SG
Attachment 142980
Distance to: Gannicus_MergedFile_officialDavidski_scaled
0.02821931 Scotland_Viking:VK202__AD_950__Cov_57.86%
0.16307509 Italy_Pompeii_Roman:f1R__AD_79__Cov_33.78%
Target: Gannicus_MergedFile_officialDavidski_scaled
Distance: 0.0240% / 0.02397695
91.6 Scotland_Viking
8.4 Italy_Pompeii_Roman
Attachment 142981
Distance to: Gannicus_MergedFile_officialDavidski_scaled
0.02821931 Scotland_Viking:VK202__AD_950__Cov_57.86%
0.11782058 Italy_Basilicata_Venosa
Target: Gannicus_MergedFile_officialDavidski_scaled
Distance: 0.0218% / 0.02181276
86.6 Scotland_Viking
13.4 Italy_Basilicata_Venosa
Attachment 142982
Distance to: Gannicus_MergedFile_officialDavidski_scaled
0.02821931 Scotland_Viking:VK202__AD_950__Cov_57.86%
0.25568171 Syria_TellQarassa_Umayyad
Target: Gannicus_MergedFile_officialDavidski_scaled
Distance: 0.0232% / 0.02324438
94.1 Scotland_Viking
5.9 Syria_TellQarassa_Umayyad
Attachment 142983
Distance to: Gannicus_MergedFile_officialDavidski_scaled
0.02821931 Scotland_Viking:VK202__AD_950__Cov_57.86%
0.22606617 Egyptian_LateKingdomPeriod_JK2134_JK2911_avg
Target: Gannicus_MergedFile_officialDavidski_scaled
Distance: 0.0228% / 0.02284658
93.1 Scotland_Viking
6.9 Egyptian_LateKingdomPeriod_JK2134_JK2911_avg
And then I modeled VK202's Steppe+ANF/EEF+WHG admixture:
Attachment 142984
With the exact same references, I modeled my own Steppe+EEF+WHG admixture:
Attachment 142985
Pre Yamnaya model:
Attachment 143029
I'm having great difficult getting good models for Iron Age and Medieval periods. All my models have a large statistical error.
Attachment 143031Quote:
Originally Posted by Grace O'Malley
Attachment 143032
Attachment 143033
You can try these ^
With those references I did this model:
Attachment 143034
Attachment 143035
Attachment 143036
Attachment 143037
You can also try this reference list too. It was with this list that I did this model:
Attachment 143038
If you try either of those two lists, you may need to remove or substitute a related population one at a time.
First model didn't work but 2nd one did. :thumb001:Quote:
Originally Posted by Gannicus
https://i.postimg.cc/DZ7W04Np/Screen...-25-131642.png
https://i.postimg.cc/sxybtsvh/Screen...-25-133037.png
I no doubt have real ancestry from Kilteasheen as my father has long term ancestry from close to where these genomes were buried and is R1b-M222. I just can't find the right model for them.
On MyHeritage Ancient Origins I'm closest to Kilteasheen KIL001.
My No 1 two way fit on Illustrative dna is below.
https://i.postimg.cc/QdT281tp/Screen...-25-160036.png
My No 2 is with the Pict and No 3 with Kilteasheen.
https://i.postimg.cc/HkMkqJ1N/Screen...-25-161530.png
That's something you can do. You can try to replicate the two-way models you see on the Deep Ancestry section of IllustrativeDNA. With those references like I mentioned, you may have to remove some references or substitute a related group.Quote:
Originally Posted by Grace O'Malley
My best attempt so far, can’t seem to get those standard errors down
Attachment 143041 Norway Medieval is French
Attachment 143042
Attachment 143043
Attachment 143044
I did another Steppe/ANF/WHG model of myself. The goal here was to maximize p-value and tighten up the standard errors as much as possible.
Attachment 143054
References:
Attachment 143055
Attachment 143056
Attachment 143057
Some of my recent random results using modern populations as a reference
https://i.postimg.cc/sXskpy4n/image.png
https://i.postimg.cc/Z0tFdYrn/image.png
https://i.postimg.cc/3xZVYzTm/image.png
I did a number of models with this IA British individual 6DT3. I looked up what studies have been done on him, and it appears that he has about 25% Iron Age Scandinavian ancestry:
https://www.nature.com/articles/s41586-024-08275-2?utm
"Twigstats substantially improves models of admixture between ancestries from Iron Age Britain and northern Europe in early medieval England9, halving standard errors from 9% with SNPs to 4% when using time stratification (point estimates 80% and 79% Iron Age Britain-related ancestry, respectively). We used this improved resolution to demonstrate that an earlier Roman individual (6DT3) dating to approximately second to fourth century ce from the purported gladiator or military cemetery at Driffield Terrace in York (Roman Eboracum), England60, who was previously identified as an ancestry outlier61,62, specifically carried approximately 25% EIA Scandinavian Peninsula-related ancestry (Fig. 2c). This documents that people with Scandinavian-related ancestry already were in Britain before the fifth century ce, after which there was a substantial influx associated with Anglo-Saxon migrations9. Although it is uncertain whether this individual was a gladiator or soldier, individuals and groups from northern Europe are indeed recorded in Roman sources both as soldiers and as enslaved gladiators63,64."
Attachment 143284
Target: Gannicus_MergedFile_officialDavidski_scaled
Distance: 0.0393% / 0.03932320 | R2P
79.8 England_IA_Roman:6DT3__AD_200__Cov_73.36%
20.2 Greece_Kastrouli_IA:I17962__BC_644__Cov_41.00%
Attachment 143285
Target: Gannicus_MergedFile_officialDavidski_scaled
Distance: 0.0455% / 0.04545825 | R2P
89.0 England_IA_Roman:6DT3__AD_200__Cov_73.36%
11.0 Egyptian_LateKingdomPeriod_JK2134_JK2911_avg
Attachment 143286
Target: Gannicus_MergedFile_officialDavidski_scaled
Distance: 0.0453% / 0.04526365
83.2 England_IA_Roman:6DT3__AD_200__Cov_73.36%
16.8 Italy_Imperial.SG
Attachment 143287
Target: Gannicus_MergedFile_officialDavidski_scaled
Distance: 0.0482% / 0.04815272
87.5 England_IA_Roman:6DT3__AD_200__Cov_73.36%
12.5 Italy_Pompeii_Roman:f1R__AD_79__Cov_33.78%
Here are some more models:
Attachment 143423
Target: Gannicus_MergedFile_officialDavidski_scaled
Distance: 0.0407% / 0.04067256
81.0 Wales_MIA_LIA
19.0 Italy_Imperial.SG
Attachment 143424
Target: Gannicus_MergedFile_officialDavidski_scaled
Distance: 0.0350% / 0.03498707
64.9 Wales_MIA_LIA
35.1 Italy_IA_Republic.SG
Attachment 143425
Target: Gannicus_MergedFile_officialDavidski_scaled
Distance: 0.0353% / 0.03532272
88.3 Ireland_EBA.SG
11.7 Egyptian_LateKingdomPeriod_JK2134_JK2911_avg
Attachment 143426
Target: Gannicus_MergedFile_officialDavidski_scaled
Distance: 0.0296% / 0.02958242
82.8 Ireland_EBA.SG
17.2 Greece_Koufonisi_Cycladic_EBA.SG
Comparison of me and Ukrainian.HO samples
I am aware of the negative p-value, but this is meant to illustrate the difference in drifts. It is clearly visible how the Ukrainian sample is heading southeast while I am heading west. Unless I did something wrong? I would also like to finally improve my margins, as this problem only arises when modeling my kit.
https://i.postimg.cc/TwF5xddL/image.png
my recent models:
https://i.postimg.cc/4yvc1CGY/image.png
https://i.postimg.cc/HL4CHBTQ/image.png
https://i.postimg.cc/TwYBxx76/image.png
https://i.postimg.cc/1RLPGg7V/image.png
https://i.postimg.cc/VLdr6brS/image.png
GermanyEarlyMedieval.SG sample admixture model:
https://i.postimg.cc/288ytLFc/image.png
Has anyone tried modeling themselves yet as 100% of one source?
I can model myself as 100% Iron Age Central Gaul from Yonne:
Attachment 143484
Models I've made using admixtools 2 (the faster but less accurate version of admixtools 1 that i usually use)
Rights:
Mbuti.DG,Russia_UstIshim_IUP.DG,Georgia_Dzudzuana_ UP.SG,Russia_Sidelkino_HG.SG,Israel_Natufian.AG,Ru ssia_MA1_UP.SG,Switzerland_Bichon_Epipaleolithic.S G,Russia_DevilsCave_N.SG
lefts:
Estonia_IA.SG,Ukrainian.HO,Kazakhstan_Sarmatian_Tu rk_IA.AG
Attachment 143491
Attachment 143492
Attachment 143493
Attachment 143494
Attachment 143495
Attachment 143496
Did another one population source model with France_HautsDeFrance_IA2.SG:
Attachment 143548