Eupedia Forums
Site NavigationEupedia Top > Eupedia Forum & Japan Forum
Results 1 to 11 of 11

Thread: Admixture studio question (smaller percentages)

  1. #1
    Regular Member
    Join Date
    07-05-21
    Posts
    15

    Y-DNA haplogroup
    I2a1b (L161-Isles)

    Country: UK - England



    Admixture studio question (smaller percentages)

    So I bought the Pro version of admixture studio, and I have to admit it's great fun running my data through all the calculators. Most of them seem to be generally accurate for the bigger chunks, such as NW Europe, SW Europe etc.
    Also what's great is that almost all of the calculators are giving me a consistent Berber chunk of between 3 to 5%, which is consistent with my dad being NW Spanish, but which Ancestry DNA (whose raw data I am using) did not pick up at all.

    However I'm also getting some outlandish results that come and go depending on the calculator. One calculator will give me 0.4% Papuan, which will then not appear at all in any of the other calculators. One will give me 0.3 Inuit, which in turn will not appear in any of the other calculators. One calculator will give me a sliver of Omotic, which will not appear...etc etc etc...

    I'm no expert, and am not sure how to interpret these minute portions. Is there a cut-off point percentage-wise, below which I should not put too much stock in the results? Should I be focusing mainly on the larger chunks?

    Thanks...

  2. #2
    Regular Member Archetype0ne's Avatar
    Join Date
    11-06-18
    Posts
    988

    Y-DNA haplogroup
    J2b2-L283/FT29003

    Country: Albania



    Hey mate.

    Great pur chase of the pro. Have had it for almost a year and thoroughly enjoy it, best investment I did genetics tool wise.

    Some of the calculators are not very reliable, don't want to mention names.

    However the trace results <2% sometimes are just noise, added there to pull your sample in a particular direction for a better fit. If it consistently pops up across calculators there might be more to it. But I would not pay much attention to trace amounts of nearly impossible ancestries.
    “Man cannot live without a permanent trust in something indestructible in himself, and at the same time that indestructible something as well as his trust in it may remain permanently concealed from him.”

    Franz Kafka

  3. #3
    Regular Member
    Join Date
    07-05-21
    Posts
    15

    Y-DNA haplogroup
    I2a1b (L161-Isles)

    Country: UK - England



    Yeah, it's a great little tool. Really interesting and allows you to go a bit deeper than just the recent ancestry.

    Thanks for the answer! I'll not pay too much attention to anything under 2% then. Seems like a good rule of thumb...

  4. #4
    Regular Member Archetype0ne's Avatar
    Join Date
    11-06-18
    Posts
    988

    Y-DNA haplogroup
    J2b2-L283/FT29003

    Country: Albania



    Quote Originally Posted by I_too_am_I2 View Post
    Yeah, it's a great little tool. Really interesting and allows you to go a bit deeper than just the recent ancestry.
    Thanks for the answer! I'll not pay too much attention to anything under 2% then. Seems like a good rule of thumb...
    No problem.

    Nonetheless some trace amounts need to be pa yed attention too if they pop up on most calculators.

    Say on FTDNA and other mainstream DNA analysis sites I come up 100% European.
    Nonetheless doing some private analysis with these tools as well as pa ying for an analysis through some service, I noticed that a lot of calculators give me 0.6-1.8% Han or some sort of East Asian ancestry. In this case I do not ignore it, since whether it pops up as Han or any other East Asian ancestry it consistently does with a large number of calculators (30-50%), and also the private service I pur chased showed the same trace amount of EA ancestry.

    Nonetheless asking the geneticist that did my analysis, he said that it could be just an overfitting of the model, hence it could or could not be an actual genetic contribution.

    A friend ran my results, with turkic, roma and mongolian... but none matched the model to substitute that 1.6% East Asian, so I am guessing it has to be some "ancient genetic" input that was not "traditionally" on the steppe if its not just noise.

    An example:


    South China Neolithic Farmer :O ?
    Tibetan Plateau Late Neolithic ?

    I still cant wrap my head around this.

  5. #5
    Regular Member
    Join Date
    07-05-21
    Posts
    15

    Y-DNA haplogroup
    I2a1b (L161-Isles)

    Country: UK - England



    That's interesting. So is there any percentage that you would say can be ignored even if it comes up on multiple calculators? Something like 0.1 or 0.2%? I'm getting South Asian on a few calculators in miniscule amounts, but can't see where it would come from.

    The other thing I was thinking, is if for example Indians have ancient DNA from Iranian farmers as I believe is the case, and say I've also picked up some ancient DNA from Iranian farmers by having some Med ancestry, is it possible the calculator is picking that up and linking me to South Asia and "cutting out the Iranian middle man" so to speak?

    By the way, is that calculator in the image one of the ones on the admixture studio?

  6. #6
    Regular Member Archetype0ne's Avatar
    Join Date
    11-06-18
    Posts
    988

    Y-DNA haplogroup
    J2b2-L283/FT29003

    Country: Albania



    I would say all trace ammounts can be ignored and would not change much. However hey are interesting to keep in mind.

    It is possible with modern populations for that to happen, if you share some similar component from an ancient population. Thats why in the image above I went around that by actually providing just the ancient component calculator.

    About that calculator, it is based on g25 values that only Davidski currently provides, as far as I know, and when I bought it it cost $12.
    Then after getting the coordinates you either use https://dnagenics.com/g25-studio/ or http://g25vahaduo.genetics.ovh/. I use the second currently since its free. But it is the same thing. You have various calculators based on g25, some members in this forum has said that even Davidksi has said it is not accurate for the modern populations (but for me its the most accurate calculator (maybe due to Albanian being my nationality and Albania being quite well sampled for its population size)). Yet from what I can tell these g25 calculators have a big bonus, of including ancient genetic components not really available in admixture/gedmacht, components such as Yamnaya, CHG averages etc. I am sure that such things can be custom made given you find the files for the ancient populations and simply use Dodecad to create averages, similar to what Jovialis and Salento have done in some other thread for ancient Italian populations. But it is too much work for me.

    In your case I would just play around with admixture studio, the ancient calculators and see if you can pinpoint anything, since the Pro is already a great tool.


    Calculators that have such ancient components are helpful. Meaning non simply geogrpahical components: balkans, nw europe etc, but rather eef, chg, shg, whg. I just wish thes ecomponents were separated for this particular calculators.

    There should be a few such calculators in admixture studio.

    Good luck.

  7. #7
    Regular Member Archetype0ne's Avatar
    Join Date
    11-06-18
    Posts
    988

    Y-DNA haplogroup
    J2b2-L283/FT29003

    Country: Albania



    In this case for example, I would ignore Ameridian, since its likely impossible.



    Edit: Now if you notice the mixed mode results, such as Tuscan Karelia and Tuscan Samara. I suspect I get the Amerindian due to Siberian populations that contributed millennia earlier to Amerindian genetics and I suspect that such a component might have been found in those Neolithic China Farmers / Tibetans. But to actually investigate that I would need custom models and a bunch of autosomal data from ancient samples.

  8. #8
    Regular Member
    Join Date
    07-05-21
    Posts
    15

    Y-DNA haplogroup
    I2a1b (L161-Isles)

    Country: UK - England



    The mixed mode results, it's not saying you actually have Greek and Hungarian ancestry, just that you are best modelled with those percentages of those two populations right? Like the Yamnaya are neither Pakistani nor Scottish, but I read somewhere they can best be modelled as a two way mixture of Balochi and Highland Scottish.

    The mixed mode results are also giving me some weird populations it's unlikely I have any links to.

  9. #9
    Regular Member Archetype0ne's Avatar
    Join Date
    11-06-18
    Posts
    988

    Y-DNA haplogroup
    J2b2-L283/FT29003

    Country: Albania



    Quote Originally Posted by I_too_am_I2 View Post
    The mixed mode results, it's not saying you actually have Greek and Hungarian ancestry, just that you are best modelled with those percentages of those two populations right? Like the Yamnaya are neither Pakistani nor Scottish, but I read somewhere they can best be modelled as a two way mixture of Balochi and Highland Scottish.
    The mixed mode results are also giving me some weird populations it's unlikely I have any links to.
    Correct. These are only models, and the best fits populate the top of the list. But sometimes due to overfitting those are not the best models.

    Think of it in PCA terms, one component pulls the other in one of 4 directions to get closer to the sample.
    In this case however my model from the calculator is quite good.
    That is Bronze Age Hungarian (some sort of north paleo Balkanic) and Greek. Albanians and Greeks are in a genetic continuum, while Albanians and Tuscans are relatively close on PCAs just more eastern shifted (PCA).
    So I can be modeled as 78% modern Greek and 22% Bronze Age Hungarian, however that is likely overfit.
    Tuscan and Karelia Samara, likely overfit too, but it gives me some idea of where the rest of the component that ends up Amerindian could originate from. Namely some sort of very old Siberian Hunter Gatherers.

    And it is further supported by the Albanian 97% + 3% of these other populations. Take for example Motala, 9.7k hunter gatherers/foragers found in Motala Sweden in 2009. The timeframe is some 4-5k years away from certain groups of Amerindians reaching North America.

    https://www.ancient-origins.net/unex...bitants-002361

    In the case of Yamnaya being modeled as Scottish / Afghan. One component pulls the Yamnaya sample North West Europe, the other East of Iran.

    These is where the Yamnaya lived

    So it does not mean that a Scottish person and a Afghani person having children gave us Yamnaya. Rather that taking the NWE shift of Scottish and averaging it with the Eastern shift of Afghani gives you a good model to estimate Yamnaya. This is further corroborated by both populations stemming from some Indo European stock, at least in part.
    Similar case to me being modeled as Tuscan (West Shifted Albanian) + Karelian/Samaran (Higher Eastern Shift IE/Siberian HG) to model what I am.

  10. #10
    Regular Member
    Join Date
    07-05-21
    Posts
    15

    Y-DNA haplogroup
    I2a1b (L161-Isles)

    Country: UK - England



    1 members found this post helpful.
    Got it! Dude thank you so much for taking the time to explain all this. You have cleared up a lot and have really improved my understanding of the results. I appreciate it.

  11. #11
    Regular Member Archetype0ne's Avatar
    Join Date
    11-06-18
    Posts
    988

    Y-DNA haplogroup
    J2b2-L283/FT29003

    Country: Albania




    You're welcome! Enjoy your research ;)

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •