Announcement

Collapse
No announcement yet.

MAJOR recalibration of Near Eastern DNA in Geno 2.0+ European reference populations??

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #16
    Originally posted by Georgeanna7 View Post
    I don't know Leo76. That is a heck of a lot of subsuming! It is akin to "turning adipose tissue into muscle tissue". It is a physiological impossibility. One replaces the other. In this case, at least with my results, Scandinavian has completely replaced Mediterranean. The only logical explanation would be - what they thought was Mediterranean in Gen 2.0 should have actually been attributed to Scandinavia; a "contribution or deposit" of the Vikings raids that went into North African and Mediterranean. Still, as you stated, they should have given an explanation. I am still learning...
    No they wouldn't turn Mediterranean into a 1:1 swap into Scandinavian.

    Geno 2.0 did not even have a Scandinavian component. It used 8 basic ancient components. If you were pure Scandinavian your result might have looked something relatively close to what their Danish reference sample's mix of the basic ancient components looked like.

    On Geno 2.0 NG a bunch of the components, like British and Irish, Western and Central European, Eastern European, Scandinavian and perhaps to a tiny degree even Finnish and Siberian each contain within them a sort of base average for those regions of ancient farmer signals (Mediterranean and Southwest Asian basic ancient components in Geno 2.0) plus the base average amount of the type of hunter gather for that region (in a sense, they might be measuring something else directly to be able to make it stand out directly, but in the end it covers that stuff).

    Comment


    • #17
      Originally posted by Leo76 View Post
      Yeah, I was just giving wombat's hypothesis the fairest possible shake. It doesn't make much sense, does it.

      In my case, the first version of the test (Geno 2.0) had me at 38% Mediterranean and 17% SW Asian, for a total of 55% vaguely "near Eastern" origin. (The rest--the majority--was Northern Europe. Those combined admixtures matched almost perfectly with the same components in the German reference population at the time.) With the second version of the test (Geno 2.0+), ALL OF THAT is gone in my results, with the exception of a 2% "Arabian" component.

      My results in Geno 2.0+ are now:

      Great Britain and Ireland: 52%
      Western and Central Europe: 32%
      Scandinavia: 14%
      Arabia: 2%

      This makes MUCH more sense given my known ancestry, but again, it doesn't explain the huge divergence between Geno 2.0 and 2.0+. I'm starting to think more and more that they were just REALLY way off on Geno 2.0 and are just trying to "gloss over it" now. Hence the lack of communication and explanation.

      Kind of like, "Oh hey, you know all that DNA we thought was from SW Asia and the Mediterranean? Yeah, turns out it was native European all along, almost in its entirety. Let's COMPLETELY recalibrate the new test, LOL."
      No Geno 2.0 was not off at all. As they say on the Geno 2.0 website, all of Europe saw an influx of early farmers a few thousand years ago. If you look at the latest papers they know for sure that farming didn't just spread culturally into all of Europe but it was brought direct along with farming people, who often lived uncomfortably along side the original hunter gatherer people for long periods of time before eventually, eventually everyone mixed. Pretty much all of Europe then ended up a mix of various original hunter gatherer tribe populations plus mixed in to varying degrees it appears two major migrations of farming people from the Fertile Crescent region. After the influx of farmers, native European WAS a mix of old hunter gatherer plus the two farming components.

      And the Geno 2.0 NG site does say, if you over all of the reference population stuff that some of the new categories included a mix of signals all at once. So Western and Central Europe component includes the mix that area had on average of those farmers and the earlier hunter-gatherers for that region.

      (Things are even more mixed since there was also an invasion by Proto-Indo-European Language speaking horsemen who mixed in and whose languages eventually won out over almost all of Europe. Finland and Estonia have a different language type as do Basques, maybe a few others. Lithuanian retains the most of the original archaic Proto-Indo European language forms (with Latvian next) of all the living Indo-European languages (ones like English, French, German are much more removed). Apparently these people were very patriarchal and in most cases when they came in traces of old female Pagan gods were wiped out (apparently in the Baltics the PIE culture didn't take over quite as deeply and it ended up more of a 50:50 blend of what had been before and what was new).

      And so on.)
      Last edited by wombat; 3 May 2016, 12:10 AM.

      Comment


      • #18
        Probably they should have provided a chart showing how much early farmer each of the new categories subsumed since that would give more info for those who tested only on the new test to go by and maybe avoided a few confusions for people who had taken the old test (although it seemed to me they did sort of explain it if you read all the stuff) and actually even for those who took the old test it's hard to be 100% what the exact mix they subsumed into each group is for 100% sure and it takes a bit of work to sort of figure it out as you have to look at what the old reference populations scored for all the places that the new components cover and do a bit of an educated weighted average.

        They probably got sick of all the complaints from people confused by the old test going but I'm not 30-60% from Southern Europe and so didn't break down the Northern European into new bits but just created some new composite categories more similar to what other companies use instead. It's also possible they had to do that in order to test them they had to do it differently and couldn't have done it the other way and would've just had to robotically split it after the fact. It's also likely that since they had FTDNA program the autosomal result sifter in part now and FTDNA hadn't been using such component types that the code was just setup to handle it this way.

        Comment


        • #19
          Respectfully, wombat, I genuinely don't think you're comprehending the problem. "Subsuming" does not explain anything.

          The first test said that a massive amount of a typical European's DNA is of near-Eastern origin. [1]

          The second test says that nearly none of a typical European's DNA is of near-Eastern origin. [2]

          Now, there are only three possibilities:

          1. The first test was deeply flawed, and that Mediterranean/SW Asian DNA is actually native European after all, which Geno 2.0+ corrected for.

          2. The second test is deeply flawed--probably deliberately so--in the sense that it completely "ignores" near-Eastern DNA altogether. (This can't really be the case, though, since my own results show a 2% "Arabian" admixture.)

          3. There is, as you suggest, some degree of "subsumption" that has occurred: proportional absorption of near-Eastern DNA into the current European regional profiles in Geno 2.0+. But again, this isn't really possible, because you're talking about HUGE chunks of a genome--anywhere from about 46% upwards, even amongst a Nordic Danish population to 53%, for a German [1]--somehow suddenly becoming natively "European" almost literally overnight. If we were talking about percentages on the order of 6% or 8%, I could get on board with your hypothesis. But the differences between the respective test population profiles are too vast.

          So if you ask me, it all comes down to option 1. They did a LOT of correcting between test 1 and test 2.


          [1] Geno 2.0 reference populations: https://genographic.nationalgeographic.com/reference-populations/

          [2] Geno 2.0+ reference populations: https://genographic.nationalgeograph...ions-next-gen/
          Last edited by Leo76; 3 May 2016, 12:26 PM.

          Comment


          • #20
            Originally posted by wombat View Post
            And the Geno 2.0 NG site does say, if you over all of the reference population stuff that some of the new categories included a mix of signals all at once.
            Where exactly do they say this? Could you please point me to the specific web page and section? Or perhaps copy and paste the exact language they use? I've scoured the NG site, and haven't found anything like this.

            Not doubting you, I'd just really like to read it for myself. Thanks.

            Comment


            • #21
              Originally posted by Leo76 View Post
              Respectfully, wombat, I genuinely don't think you're comprehending the problem. "Subsuming" does not explain anything.

              The first test said that a massive amount of a typical European's DNA is of near-Eastern origin. [1]

              The second test says that nearly none of a typical European's DNA is of near-Eastern origin. [2]
              Yes, the first test says that (and it is correct, if you test Europeans by hunter-gather and ancient farming components, you will find that Europeans had a lot of genes get mixed in when the farmers came in and spread across European. Farming was spread physically by new people arriving. It was not just culturally transmitted without any migration and mixing of people. You can find scientific papers of recent years verifying this. And there are some GEDMATCH tools that will show things in a similar vein, they may be a little less stable and use slightly different reference samples and components, but you can clearly see such a mixing.)

              However, no, that is not what the second test is saying at all and it's not at all contradictory with the first test. They are just using different components. In this case the northern and eastern European components include a mix of the components from the first test. This time they are testing for gene patterns in different regions of western, northern, eastern parts of Europe and not testing the components they did the last time. And you can see results more of this type for those categories also on Ancestry or 23.

              Now, there are only three possibilities:

              1. The first test was deeply flawed, and that Mediterranean/SW Asian DNA is actually native European after all, which Geno 2.0+ corrected for.
              Not this one. And you are not getting that there is no such thing as singular pure native European. If you looked 12,000 years ago you'd mostly find different populations of hunter-gatherers. But if you wanted to look at more recent Europeans they would have been heavily mixed in with a couple of major migrations out of the Fertile Crescent (so Europeans more like of I think something like 3000-7500 years back, in some areas the farmers reach earlier). So a modern Western and Central European say, keeping it in simplified terms, is a mix of early hunter-gatherer plus two major farming migrations (plus also Indo-European language speaking migration plus all sorts of mixings of this and that on small scales over the last 1000 years.)

              2. The second test is deeply flawed--probably deliberately so--in the sense that it completely "ignores" near-Eastern DNA altogether. (This can't really be the case, though, since my own results show a 2% "Arabian" admixture.)
              I wouldn't say it is deeply flawed either. Some people will have gotten some bits that match extra strongly with people from the farming areas either due to a recent to pseudo-recent ancestor or the part of the region they came from and their various lines there happened to tie into tribes that had a bit closer and stronger ties and more mixing way back when from certain areas than average for people in the whole huge regions covered by each of the European categories.

              3. There is, as you suggest, some degree of "subsumption" that has occurred: proportional absorption of near-Eastern DNA into the current European regional profiles in Geno 2.0+. But again, this isn't really possible, because you're talking about HUGE chunks of a genome--anywhere from about 46% upwards, even amongst a Nordic Danish population to 53%, for a German [1]--somehow suddenly becoming natively "European" almost literally overnight. If we were talking about percentages on the order of 6% or 8%, I could get on board with your hypothesis. But the differences between the respective test population profiles are too vast.
              This is it. And it's not shocking because you forget that native European is not some singular pure thing, native European since around 300-7500 years ago not just on micro-scales but on mega-scales is a composite category consisting of huge chunks of a few different basic ancient broad-level heritages.

              So if you ask me, it all comes down to option 1. They did a LOT of correcting between test 1 and test 2.
              no. In some ways the first test is probably even more stable since it used very simple, ancient, pretty orthogonal components.

              If what you suggest were to be true to that level then both half the scientific papers written about humans in Europe are wrong and either all of Geno 2.0 plus much of GEDMATCH are pure garbage or all of Geno 2.0 NG, FTDNA MyOrigins, Ancestry and 23 are pure garbage.

              Comment


              • #22
                Originally posted by Leo76 View Post
                Where exactly do they say this? Could you please point me to the specific web page and section? Or perhaps copy and paste the exact language they use? I've scoured the NG site, and haven't found anything like this.

                Not doubting you, I'd just really like to read it for myself. Thanks.
                Actually, my apologies on this, it's not quite as directly explained there as I thought (I had skimmed over the reference populations results stuff too quickly and it's actually not made as directly clear there as I thought since) so I can definitely see how it could be pretty confusing for someone who tested on the first test.

                You can also see hints of it in text about the components though too like:
                "This component of your ancestry originates in the plains that extend from the Danube River and the Black Sea north to the Baltic Sea and east to the Volga River and the Ural Mountains of Russia. Your ancestors who lived in this region thousands of years ago were likely hunters and gatherers who gradually adopted agriculture from their neighbors to the south and west." (although, granted, this doesn't make it very clear as adopted could mean culturally only and not through mixing of people)

                "Your prehistoric Scandinavian ancestors most likely survived from hunting, gathering, and fishing, and it wasn’t until a few thousand years ago that farming first reached the area." (granted once again it doesn't make it very clear exactly what is going on for someone who tested on the first test and hasn't read lots of background and isn't used to dealing with component assignments and orthogonal and mixed basis and so on)

                "This component of your ancestry is associated with the western European islands of Great Britain and Ireland, but traces can also be found along the northern and western coasts of continental Europe. As modern humans first entered Europe, this part of the world was uninhabitable and covered in ice sheets. As the ice sheets retreated, settlers moved to the islands. The earliest settlers likely survived on fishing, but farming eventually reached the islands in the past several thousand years." (maybe a little bit clearer in this one)

                The new test simply uses a more complex basis and one that is mixed in terms of the components used in the first test.

                You could imagine a simple 3D X,Y,Z plot with each axis orthogonal and they could be hunter-gatherer, farmer 1, farmer 2, the dot product between any two basis vectors is zero and the component vectors for the basis are not mixed. You could then tilt some of the basis vectors and form a different, non-orthogonal basis and the components would be mixed. With the new test though it's even more complex.

                One quick thing to think of is imagine if the first test was a 3D cube with axis X=HG,Y=Farmer1,Z=Farmer2. The second test would have a hyper-cube, but to keep it simpler I'll just use three of new components: X=B&I, Y=W&C E, Z=SE (only it's a bit trickier in that these basis are likely not quite as stable and have more overlap with themselves in various ways). If you put that cube inside the 3D cube of the first test and put those new basis vectors in terms of the components used in the first test you'd find they'd not match along the original X,Y,Z axis. You'd see that a value of (1B&I,0,0) in the new system might be say (0.6HG(X),0.4F1(Y),0.2F2(Z)) in terms of the basis of the old test. So the one single component in the new test would be a mix of parts of three different axis on the old test.

                Between using so many more components and trying to find isolated key bits to key in on the new components probably had to made so different that there was no way to still talk in terms of split European plus the old original farmer components, things were probably too mixed up in odd ways and that might've required even more basis and become more unstable perhaps.

                Also I'm not sure ancestry composition is actually done that way though. I think they actually have use a very different technique that in the end is more like look at each little bit and see how well it matches each component so some spot might seem 70% like W&C E and 25% like SE and 5% like EE and they probably just call that spot W&C E and maybe forget the rest and then they maybe add up the totals of what component 'won' each section. So if it tested three spots and they were 70% W&C E, 25% SE, 5% EE and 60% W&C E,40 SE and 45% W&C E, 55% SE, 0 % EE maybe you'd get 66% W&C E and 33% SE at the end (and not 58.3% W&C E, 40% SE, 1.6% EE as this might actually be more unstable and misleading in the end due to each basic component itself being unstable to some degree and mixed and so on and so forth) ? Maybe some weighting too etc. I'm not actually sure. Haven't looked into the details of how they actually carry it out. But just noting that the actual process may be a bit different than the cube stuff I mentioned above, but that cube stuff just does show one way how components can be mixed.
                Last edited by wombat; 3 May 2016, 05:14 PM.

                Comment


                • #23
                  I'm wrong. D'oh!

                  Actually, you know what? You responded before I was able to come along and eat crow and take back pretty much everything I said, LOL.

                  I went in and read in much more detail the regional population descriptions in Geno 2.0+ in combination with what you were saying, and yeah, now I get how they "subsumed" the near Eastern admixtures into the modern European components.

                  So, sorry! Sometimes I am a Bear of Very Little Brain.

                  The only mystery to me now, is, if all the near-Eastern, neolithic migration stuff has been accounted for in the new test (which it certainly seems to be), where in the hell does my 2% "Arabian" DNA come from??

                  Comment


                  • #24
                    Originally posted by Leo76 View Post
                    All my family has been in America a looong time (since before 1770, on all branches), but yes, mainly UK, with some French and German.

                    My Geno 2.0+ breakdown is:

                    Britain and Ireland: 52%
                    Western and Central Europe: 31%
                    Scandinavia: 14%
                    Arabia: 2%

                    Or, more easily visualized:

                    http://i.imgur.com/ctqrRIS.jpg
                    They were not able to entirely subsume everything for everyone as there are sub-regional variations.

                    Few people exactly match the average reference component. As you can see, on average, people in England also get differing mixes of other bits depending upon whether their very ancient, semi-ancient, semi-recent, recent ancestry got more or less mix-in from this area or that. It's such a totally mixed up area it seems the mix any individual gets varies somewhat if they are from the UK or Germany in particular.

                    You look like their British reference but a little different (a touch on the lower side of B&I; hard to say if that is a hint of some within tree-tracing times heritage from the continent or just the particular chance mix your heritage from your parts of England had).

                    The Arabian could be some few times great-grandparent directly from there, all sorts of stuff happens, but it's probably more likely your equivalent to the Southern European component seen in the average British reference population where, especially for southeast English they often need some more ancient farmer or semi-old mix-ins from Southern Europeans. Either by chance or heritage going way back, some of your ancient times or more recent time but still a ways back (farther back than you could trace a tree too, maybe Moors had more influence than average for your ancestors than for the average person in the UK maybe (assuming you are from the UK or America/Canada but with lots of heritage ultimately tracing back to the UK)) southern and farmer stuff mix-ins end up closer to the gene patterns in their Arabian reference component than to their Southern European reference gene pattern.

                    I forget, but what is your known background if any?
                    Last edited by wombat; 3 May 2016, 05:35 PM.

                    Comment


                    • #25
                      The vast majority of my identified ancestors are English. (Not just "British", but English, proper.) I do have French and German ancestors, but they all left the Continent between 1750 and 1800, and it's only single lines here and there. So I'd reckon my ancestral breakdown would be something like 82% English, 18% Continental "other".

                      Which actually is ANOTHER mystery. It makes me wonder why my Western and Central European admixture is a full whopping 31%, as measured by Geno 2.0+. Clearly there would be some anyway, even if I were full, native English. But my most recent Continental ancestor left there around 1780, which was 7 generations back. That's long enough ago to dilute the genetic inheritance down to statistically zero, after intermarrying here in America with my English ancestors.

                      But anyway, on to the Arabian thing:

                      it's probably more likely your equivalent to the Southern European component seen in the average British reference population
                      That HAS to be it. Unless somebody was gettin' bizzy recently with somebody middle Eastern (HIGHLY improbable, given how well documented my genealogy is all round, going back at least 6 generations), it's probably a near-Eastern admixture artifact that just happens to match the "Arabian" profile better than the "Southern European" or "Asia Minor".

                      Anyway, thanks for thinking through all this with me.

                      Comment


                      • #26
                        Originally posted by Leo76 View Post
                        Respectfully, wombat, I genuinely don't think you're comprehending the problem. "Subsuming" does not explain anything.

                        The first test said that a massive amount of a typical European's DNA is of near-Eastern origin. [1]

                        The second test says that nearly none of a typical European's DNA is of near-Eastern origin. [2]

                        Now, there are only three possibilities:

                        1. The first test was deeply flawed, and that Mediterranean/SW Asian DNA is actually native European after all, which Geno 2.0+ corrected for.

                        2. The second test is deeply flawed--probably deliberately so--in the sense that it completely "ignores" near-Eastern DNA altogether. (This can't really be the case, though, since my own results show a 2% "Arabian" admixture.)

                        3. There is, as you suggest, some degree of "subsumption" that has occurred: proportional absorption of near-Eastern DNA into the current European regional profiles in Geno 2.0+. But again, this isn't really possible, because you're talking about HUGE chunks of a genome--anywhere from about 46% upwards, even amongst a Nordic Danish population to 53%, for a German [1]--somehow suddenly becoming natively "European" almost literally overnight. If we were talking about percentages on the order of 6% or 8%, I could get on board with your hypothesis. But the differences between the respective test population profiles are too vast.

                        So if you ask me, it all comes down to option 1. They did a LOT of correcting between test 1 and test 2.


                        [1] Geno 2.0 reference populations: https://genographic.nationalgeographic.com/reference-populations/

                        [2] Geno 2.0+ reference populations: https://genographic.nationalgeograph...ions-next-gen/
                        yeah the first geno 2 test was flawed (it's results) cause i compared them to the closest i were and it was completely different (just by substracting each part european etc from all the ref models and adding all these substractions), it appear im closest to lebanon and not sardinian as they wrote it... and geno 2 took like 6 months to give me the results (cause the sample failed), strangely enough it didn't fail with 23 and ftdna

                        Comment


                        • #27
                          Originally posted by boutrosdu93 View Post
                          yeah the first geno 2 test was flawed (it's results) cause i compared them to the closest i were and it was completely different (just by substracting each part european etc from all the ref models and adding all these substractions), it appear im closest to lebanon and not sardinian as they wrote it... and geno 2 took like 6 months to give me the results (cause the sample failed), strangely enough it didn't fail with 23 and ftdna
                          the components have different mixes

                          Comment


                          • #28
                            Originally posted by Leo76 View Post
                            The vast majority of my identified ancestors are English. (Not just "British", but English, proper.) I do have French and German ancestors, but they all left the Continent between 1750 and 1800, and it's only single lines here and there. So I'd reckon my ancestral breakdown would be something like 82% English, 18% Continental "other".

                            Which actually is ANOTHER mystery. It makes me wonder why my Western and Central European admixture is a full whopping 31%, as measured by Geno 2.0+. Clearly there would be some anyway, even if I were full, native English. But my most recent Continental ancestor left there around 1780, which was 7 generations back. That's long enough ago to dilute the genetic inheritance down to statistically zero, after intermarrying here in America with my English ancestors.
                            You expect 18% but get another 13%. The British reference populations gets 9% on average. So you have 4% extra there, but might be well within margin of error since that reference might have decent variation. Now it's also true that those continental ancestor's DNA might show as a mix itself and not expect to test fully 18% Western and Central European, although who knows what bits you got, but maybe you still have a bit more than expected. It's probably not too crazy though as the the whole UK,French,German thing is such a mish-mash of genes.

                            Comment

                            Working...
                            X