Page 1 of 2 12 LastLast
Results 1 to 10 of 11

Thread: Are dna companies like 23&Me, Ancestry, MyHeritage purely based on Modern clusters?

  1. #1
    Gold Class Member
    Posts
    1,902
    Sex
    Ethnicity
    Irish
    Nationality
    Australian & Irish
    Y-DNA (P)
    M222 (S588)
    mtDNA (M)
    J1c3f

    Ireland Australia Ireland County Tipperary Ireland Munster

    Are dna companies like 23&Me, Ancestry, MyHeritage purely based on Modern clusters?

    I find this a very interesting question and I'd like other people opinions?

    My opinion is that it is based on their panels and that people that get ancestry from other categories is because they don't match the representative panels well enough so the algorithm will go to the another cluster to compensate.

    Many people attribute this to some ancient similarity i.e. Vikings, Celts, Anglo-Saxons etc.

    I think these are all based on modern day population clusters.

    I'll give the example of the British & Irish cluster from 23&Me because I get 94% of this cluster being Irish. The British & Irish population panels are composed of Irish and British populations but with how clustering and admixture works the Irish are assigned the majority of this cluster and the more outlying Southeast English are assigned the least of this and will get other categories because they are more an outgroup but many Southeast English have long term ancestry there so getting amounts of French & German, Scandinavian etc the fault of them being put in a panel with Irish so the clustering will assign Irish the most representation in that cluster and they will get the least.

    I personally don't think that 23&Me and similar companies are actually showing ancient admixture.

    These are some maps showing some of these components.

    You can see how it reaches a maximum in Ireland and then decreases from there.



    Iberian appears to be centred on Basque so the same result with outlying groups getting less.



    I'd be interesting in other's people's opinions on this?

  2. The Following 5 Users Say Thank You to Jessie For This Useful Post:

     Aroon1916 (07-27-2020),  JMcB (07-25-2020),  jstephan (07-25-2020),  msmarjoribanks (07-25-2020),  Trelvern (07-25-2020)

  3. #2
    Gold Class Member
    Posts
    1,902
    Sex
    Ethnicity
    Irish
    Nationality
    Australian & Irish
    Y-DNA (P)
    M222 (S588)
    mtDNA (M)
    J1c3f

    Ireland Australia Ireland County Tipperary Ireland Munster
    I find this quote from Generalissimo quite informative.

    Quote Originally Posted by Generalissimo View Post
    It's largely an artifact of the ADMIXTURE method.

    What often happens is that the program creates clusters based on the most isolated/drifted populations, and then the other, more outbred groups are given membership in near and far related clusters to compensate for this problem.

    So you really have to check properly whether the appearance of some Middle Eastern or even South Asian admixture reflects recent gene flow. Remember that bullshit with Dienekes' Gedrosia cluster? Same thing.

  4. The Following User Says Thank You to Jessie For This Useful Post:

     JMcB (07-25-2020)

  5. #3
    Registered Users
    Posts
    1,597
    Sex
    Location
    France
    Ethnicity
    Breton & Maineau
    Nationality
    French
    Y-DNA (P)
    R-BY145002 (DF13+)
    mtDNA (M)
    U4c1a

    France France Bretagne
    Quote Originally Posted by Jessie View Post
    I find this quote from Generalissimo quite informative.

    I am not sure that this response from Generalissimo is suitable for your question.

    I think there is genetic continuity with populations.
    The English in the South East will be closer than the Irish, to the Dutch.
    It is important to locate the place where the component is most important! It is even essential.
    Myheritage results are inconsistent between inhabitants of the same country, and therefore, nothing can be learned.
    But the results are consistent on 23andMe, so let's take them as an example.
    Irish & British, the maximum is in Ireland / Scotland.
    French & German, the maximum in South Germany.
    If your ancestors are from Northwestern Europe, but neither Ireland nor Southern Germany, then you risk having multiple categories in your results, but also a lot of Broadly Northwest Europeans.

    However, sometimes it is an ancestral population that is the cause of confusion.
    Especially the Iberian component which maybe due to the population of the European Neolithic.
    Especially true in France.
    Many French people will have Iberian in their results and yet, they have no Spaniards among their ancestor, neither recently, nor ever.
    This problem is probably due to a lack of reference from the local population, and that as you say, Iberian has the maximum in the Basque Country. But the Basque Country is Franco-Spanish!
    Y haplogroup: R1b: L21 --> DF13 --> BY145002
    The oldest L21 known are I2457 et I2565 from Stonehenge (Beaker Culture, 2400-1900 BC)

    MTDNA: U4c1
    The oldest U4c1 known are "poz224", Yamnaya culture (2882-2698 BC), and 2 Bell-Beaker in Germany (Karsdorf, 2314-2042 BC)

    Paternal MTDNA: K1b2b
    The oldest K1b2 are Eastern European Mesolithic: Kunda Donkalnis5 (Lithuania), 6000 BC and Meso-Ene Lepenski Vir Lepe28 in Serbia, 5900 BC.
    The oldest K1b2b is Alt-3, Corded-Ware Germany (2500 BC)

  6. The Following 7 Users Say Thank You to Tolan For This Useful Post:

     Aroon1916 (07-27-2020),  Jessie (07-25-2020),  JMcB (07-25-2020),  jstephan (07-25-2020),  msmarjoribanks (07-25-2020),  Ruderico (07-25-2020),  vettor (07-26-2020)

  7. #4
    Gold Class Member
    Posts
    1,902
    Sex
    Ethnicity
    Irish
    Nationality
    Australian & Irish
    Y-DNA (P)
    M222 (S588)
    mtDNA (M)
    J1c3f

    Ireland Australia Ireland County Tipperary Ireland Munster
    Quote Originally Posted by Tolan View Post
    I am not sure that this response from Generalissimo is suitable for your question.

    I think there is genetic continuity with populations.
    The English in the South East will be closer than the Irish, to the Dutch.
    It is important to locate the place where the component is most important! It is even essential.
    Myheritage results are inconsistent between inhabitants of the same country, and therefore, nothing can be learned.
    But the results are consistent on 23andMe, so let's take them as an example.
    Irish & British, the maximum is in Ireland / Scotland.
    French & German, the maximum in South Germany.
    If your ancestors are from Northwestern Europe, but neither Ireland nor Southern Germany, then you risk having multiple categories in your results, but also a lot of Broadly Northwest Europeans.

    However, sometimes it is an ancestral population that is the cause of confusion.
    Especially the Iberian component which maybe due to the population of the European Neolithic.
    Especially true in France.
    Many French people will have Iberian in their results and yet, they have no Spaniards among their ancestor, neither recently, nor ever.
    This problem is probably due to a lack of reference from the local population, and that as you say, Iberian has the maximum in the Basque Country. But the Basque Country is Franco-Spanish!
    But this is based on their panels and modern populations so it will depend on how you match the panel. What I'm saying is that it has nothing to do with ancient admixture which some people think it does.

    The quote from Generalissimo was to explain how clustering works.

    Thanks for your response.
    Last edited by Jessie; 07-25-2020 at 10:39 AM.

  8. The Following 2 Users Say Thank You to Jessie For This Useful Post:

     Aroon1916 (07-27-2020),  JMcB (07-25-2020)

  9. #5
    Registered Users
    Posts
    69
    Sex

    They use modern populations as references , also its up to them how they rephrase it, for example an european would be mixed of multiple prehistoric migrations or ancestries but
    his reference if used wouldn't show all this it would only show the last 700 yr ancestry of his average population as "Spanish" for example...so its good at detecting recent admixture.
    also the accuracy of the test depends on the company and how good the reference they have , in my opinion some companies should never be used internationally but target specific groups
    or its accuracy would be very wrong.

  10. The Following 2 Users Say Thank You to Ramses For This Useful Post:

     Aroon1916 (07-27-2020),  Jessie (07-25-2020)

  11. #6
    Gold Class Member
    Posts
    1,909
    Sex
    Location
    Chicago, Illinois
    Ethnicity
    mixed European
    Nationality
    USA
    Y-DNA (P)
    Dad: R1b/L21/DF63
    mtDNA (M)
    K2b2
    Y-DNA (M)
    R1b-M269
    mtDNA (P)
    K2b1a1a

    United States of America England Wales Sweden Germany
    Quote Originally Posted by Jessie View Post
    I find this a very interesting question and I'd like other people opinions?

    My opinion is that it is based on their panels and that people that get ancestry from other categories is because they don't match the representative panels well enough so the algorithm will go to the another cluster to compensate.
    Agreed.

    Many people attribute this to some ancient similarity i.e. Vikings, Celts, Anglo-Saxons etc.

    I think these are all based on modern day population clusters.

    I'll give the example of the British & Irish cluster from 23&Me because I get 94% of this cluster being Irish. The British & Irish population panels are composed of Irish and British populations but with how clustering and admixture works the Irish are assigned the majority of this cluster and the more outlying Southeast English are assigned the least of this and will get other categories because they are more an outgroup but many Southeast English have long term ancestry there so getting amounts of French & German, Scandinavian etc the fault of them being put in a panel with Irish so the clustering will assign Irish the most representation in that cluster and they will get the least.

    I personally don't think that 23&Me and similar companies are actually showing ancient admixture.
    I'm not totally sure what you are saying. You mean that SE English get less English and Irish than the average Irish person NOT because it is detecting ancient inputs like Viking and Anglo Saxon, but because the Irish are the most distinctive or representative of the cluster made up of modern people (because the Irish component of the panel is the most distinctive vs the rest of NW Europe)?

    Or are you suggesting there isn't a lot of ancient similarity between the British Isles and Ireland populations and NW Europe (post Bell Beaker)?

    I think NW Europe and England are hard to tell apart because of that ancient similarity AND various quite significant NW European inputs into much of Eastern England since and that Ireland both lacked as much NW European input (although there seems to be significant Viking, at least in some areas), and also did have more genetic drift in some areas (same with Wales). But I would agree that the test is not so much picking up "German" from A-S or "Scandinavian" from Vikings, but seeing the overall similarity in those areas and that the SE English are probably part way between Ireland and, I dunno, the Netherlands or Denmark.

    I should have a substantial portion of my English from SE England (hard to say, as much was 1630s emigration, but the early settlers of Mass Bay colony were largely from East Anglia, and any specific ancestor is too distant to focus on, and then my gg-grandfather who emigrated in 1870 was 50% East Anglia (the other 50% Welsh borders area). My Irish is who knows (could be all Scots and English, again colonial era emigration for the most part, the only known later example was born in Antrim and had non Irish names), and I do have more recent (1850) Welsh ancestry, half from North Wales, half from the center.

    My 23andMe probably overstates my German/French vs. my English, but only by a little bit (49.5 vs. 21.2). I think it's common for actual testers with longstanding family in SE England to get even more F&G than that. (The rest of mine is almost all Scandinavian, which is right, and broadly NW Europe.)

  12. The Following 2 Users Say Thank You to msmarjoribanks For This Useful Post:

     Aroon1916 (07-27-2020),  Jessie (07-26-2020)

  13. #7
    Registered Users
    Posts
    204
    Sex
    Nationality
    Brazilian
    Y-DNA (P)
    R1b
    mtDNA (M)
    H1

    They focus on the American market, because that's the place where most people buying DNA tests live. DNA testing companies use only recent clusters going back 500 years at most, even though I think MyHeritage and FTDNA go back a little bit more, maybe 1000 years. If you want neolithic, pre-historic and other types of inferrences you should try Global 25 by Davidski.

  14. The Following 4 Users Say Thank You to Caius Agrippa For This Useful Post:

     Aiden (07-26-2020),  Aroon1916 (07-27-2020),  Jessie (07-26-2020),  vettor (07-26-2020)

  15. #8
    Registered Users
    Posts
    97
    Sex

    This is from Eupedia

    23andMe claims that its Ancestry Composition only assesses ancestry over the last 500 years, but our analysis showed that it really distinguishes between Late Ancient and Early Medieval populations. The modern appellation like 'French & German' have been mislabled and should read more like 'ancient continental Germanic'. This explains why their Ancestry Timeline is completely unreliable, as they greatly underestimate the number of generations elapsed.

  16. The Following User Says Thank You to Aiden For This Useful Post:

     Nino90 (07-30-2020)

  17. #9
    Gold Class Member
    Posts
    1,902
    Sex
    Ethnicity
    Irish
    Nationality
    Australian & Irish
    Y-DNA (P)
    M222 (S588)
    mtDNA (M)
    J1c3f

    Ireland Australia Ireland County Tipperary Ireland Munster
    Quote Originally Posted by msmarjoribanks View Post
    Agreed.



    I'm not totally sure what you are saying. You mean that SE English get less English and Irish than the average Irish person NOT because it is detecting ancient inputs like Viking and Anglo Saxon, but because the Irish are the most distinctive or representative of the cluster made up of modern people (because the Irish component of the panel is the most distinctive vs the rest of NW Europe)?

    Or are you suggesting there isn't a lot of ancient similarity between the British Isles and Ireland populations and NW Europe (post Bell Beaker)?

    I think NW Europe and England are hard to tell apart because of that ancient similarity AND various quite significant NW European inputs into much of Eastern England since and that Ireland both lacked as much NW European input (although there seems to be significant Viking, at least in some areas), and also did have more genetic drift in some areas (same with Wales). But I would agree that the test is not so much picking up "German" from A-S or "Scandinavian" from Vikings, but seeing the overall similarity in those areas and that the SE English are probably part way between Ireland and, I dunno, the Netherlands or Denmark.

    I should have a substantial portion of my English from SE England (hard to say, as much was 1630s emigration, but the early settlers of Mass Bay colony were largely from East Anglia, and any specific ancestor is too distant to focus on, and then my gg-grandfather who emigrated in 1870 was 50% East Anglia (the other 50% Welsh borders area). My Irish is who knows (could be all Scots and English, again colonial era emigration for the most part, the only known later example was born in Antrim and had non Irish names), and I do have more recent (1850) Welsh ancestry, half from North Wales, half from the center.

    My 23andMe probably overstates my German/French vs. my English, but only by a little bit (49.5 vs. 21.2). I think it's common for actual testers with longstanding family in SE England to get even more F&G than that. (The rest of mine is almost all Scandinavian, which is right, and broadly NW Europe.)
    Yes I think with clustering when you put the Irish and British together that the Irish been a smaller and less admixed population will be the more representative so that people like the Southeast English will end up with getting other French & German, Scandinavian etc because they will be more an outgroup but this does not mean they have recent ancestry from those population it is just an artifact of how clustering works IMO.

    I just think that people try to make sense of this by attributing these admixtures to ancient populations like Anglo-Saxons, Celts, Vikings etc when these tests are not picking up ancient components as any ancient contribution will be part of the population now.

    I think all NW Europe has similarities stemming back to the Bronze Age so if you look at cluster maps they all cluster together and all have close genetic distances so I also think that people make inferences i.e. that the Irish are somehow some ancient and unmixed population which I don't agree with. The fact that they get so much Irish & British does not tell anyone about the admixture that is in Ireland.

    In short I'm saying that these companies are based on modern populations and also what populations you get will depend on their panels and algorithms. I hope that is clear what I'm saying so I would just like people's opinions on this. I think looking at history populations like Southeast English are of course more similar geographically to Continental Europeans than the Irish are so I understand why they would not end up matching a component that has Irish in it.

  18. The Following 3 Users Say Thank You to Jessie For This Useful Post:

     Aroon1916 (07-27-2020),  fabrice E (07-26-2020),  msmarjoribanks (07-27-2020)

  19. #10
    Registered Users
    Posts
    305
    Sex
    Omitted
    Ethnicity
    Irish/British
    Nationality
    Canadian
    Y-DNA (P)
    I-Y6375 (Fatherís)
    mtDNA (M)
    H17a

    Canada England Ireland Ireland County Tipperary
    Thank you Jessie.

  20. The Following User Says Thank You to Aroon1916 For This Useful Post:

     Jessie (07-28-2020)

Page 1 of 2 12 LastLast

Similar Threads

  1. NEW K35 Calculator based on modern G25 spreadsheet
    By michal3141 in forum Autosomal (auDNA)
    Replies: 140
    Last Post: 07-27-2020, 05:28 PM
  2. Replies: 9
    Last Post: 02-15-2020, 10:14 PM
  3. Replies: 32
    Last Post: 12-28-2019, 10:34 PM
  4. Scaled modern G25 distances to K=42 clusters
    By michal3141 in forum Autosomal (auDNA)
    Replies: 130
    Last Post: 05-15-2019, 07:44 PM
  5. Myheritage results based on phased kits.
    By Tz85 in forum MyHeritage
    Replies: 0
    Last Post: 08-16-2017, 01:53 PM

Posting Permissions

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