DNA.Land and New Ancestry Inference
RSS
Search Text Case Sensitive Exact Words Include Comments
List of Bookmarks
DNA.Land has a new ancestry report. Above is my own. I’m pretty aware that designing these consumer-oriented services/applications isn’t easy. But I want to express a little skepticism that two of the three South Asian populations which they used for their “Dravidian” reference are not Dravidian speaking. In fact, geography and ethno-linguistic affinity in South Asia is only modestly correlated with genetic variation because of caste. Something more generic, such as “South Indian” is what I would have gone with.
Follow @razibkhan
English
Arabic
Chinese (Simplified)
Dutch
French
German
Hebrew
Hindi
Italian
Japanese
Korean
Persian
Polish
Portuguese
Russian
Spanish
Swedish
Turkish
Vietnamese
“Central Indoeuropean” is another perhaps misnamed component. Most of the reference populations for it are non-IE speaking West Asians, including the population in which it reaches 100% (Georgians).
A lot of Hispanics who put their DNA results on Youtube said they were surprised to find out they have some East Asian ancestry. I wonder if that East Asian in some Hispanics is actually just Amerindian admixture and the DNA company made a mistake.
Do you mean South Indian, as in Ancestral South Indian? So would Gujarati stand for Ancestral North Indian?
In any case, it doesn’t make sense as ASI is different from West Eurasian.
Hispanic, the US Anglo word to describe those “de habla hispana”, as it were, are products of mestizaje/mulataje meaning they are bi-racial and tri-racial implying varying degress of European, Native American and African ancestry. If you are in Bolivia, Peru, Ecuador, etc then you will show more Native American (Asian in some cases of recent Chinese or Japanese arrivals), African (Colombia, Caribbean, parts of southern Mexico (there is an Afro-Mexico?) and those of more European ancestry who came in after the Africans or Native Americans were killed, died of disease or just ended up in another country with borders being redrawn as in Argentina, Chile, Paraguay when the Germans, Italians excaped persecution in their own countries in the 1930s.
How common is it for South Asians like Indians and Pakis to have East Asian admixture?
a lot of NA admixture tests have difficulty removing E asian admixture from certain types of natives (e.g., Na Dene).
It was strange seeing East Asian admixture popping up in the DNA of some Mexicans, when Mexico never received a large wave of East Asian immigrants like The U.S and Canada did. East Asians make up well below 1 percent of Mexico's population.
not too common unless you have bengali or munda or himalayan origin.
Are you one of those?Replies: @Razib Khan
“not too common unless you have bengali or munda or himalayan origin.”
Are you one of those?
Are you one of those?Replies: @Razib Khan
yes. i’m bengali. i’ve mentioned this numerous times.
” A lot of NA admixture tests have difficulty removing E asian admixture from certain types of natives (e.g., Na Dene).
It was strange seeing East Asian admixture popping up in the DNA of some Mexicans, when Mexico never received a large wave of East Asian immigrants like The U.S and Canada did. East Asians make up well below 1 percent of Mexico’s population.
So their “Dravidian” does not stand for ASI?
There is no pure ASI around anymore.Replies: @BB753
No, it stands for modern South Indians like Tamils and Telugus.
There is no pure ASI around anymore.
There is no pure ASI around anymore.Replies: @BB753
That doesn’t mean that they won’t be able in the future to break down ASI and ANI components in commercial DNA tests.
My own results have gotten worse from their first version. Went from 98% Ashkenazi / 2% other to 90% Ashkenazi / 5% Italian / %5 other.
I agree this stuff is hard, but getting that old error confusing Italians and Jews seems a bit sloppy to me. Their reference samples look OK so I don’t really know why they are getting such large false positives.
I was thinking perhaps this is a hint at substructure within Ashkenazis, with some being more European-shifted than others. But again, their sample are American Ashkenazi Jews, mixed from all over Europe, so that doesn’t really make sense.
I agree this stuff is hard, but getting that old error confusing Italians and Jews seems a bit sloppy to me. Their reference samples look OK so I don't really know why they are getting such large false positives.
I was thinking perhaps this is a hint at substructure within Ashkenazis, with some being more European-shifted than others. But again, their sample are American Ashkenazi Jews, mixed from all over Europe, so that doesn't really make sense.Replies: @Shaikorth
These tools, meant to measure ancient ancestry, have limited separation capacity in case of very similar populations (in terms of old ancestry). An IBD analysis would easily reveal both the relationship of Ashkenazi and Sephardic Jews and the separation of Italians from both, but that is for relatively recent ancestry.
Also 'Ashkenazi' shouldn't even be a category and many other categories would need to be reorganized / relabeled if they are indeed measuring ancient ancestry. But I don't think they are going specifically for that - quite the opposite in fact (see blog post). And if they *are* going for ancient ancestry then they should use ancient DNA of course. One of the reasons why Davidski's tests at Eurogenes tests are the best out there IMO, though I look forward to more. :)Replies: @Shaikorth
No, it’s not too difficult to separate Italians and Jews (and similarly related pops) using exactly the tools and methods they are using — which means that their test simply isn’t as accurate as some others, either because their samples aren’t representative or for whatever other reasons. 23andme used to have this Italian/Jewish problem (though if I recall it was mostly the converse — Italians scoring Ashkenazi), but their newer version eliminated that mistake for the most part.
Also ‘Ashkenazi’ shouldn’t even be a category and many other categories would need to be reorganized / relabeled if they are indeed measuring ancient ancestry. But I don’t think they are going specifically for that – quite the opposite in fact (see blog post). And if they *are* going for ancient ancestry then they should use ancient DNA of course. One of the reasons why Davidski’s tests at Eurogenes tests are the best out there IMO, though I look forward to more. 🙂
https://dna.land/faq#ancestry_algorithm
It simply is an issue of the method (ADMIXTURE and STRUCTURE-based stuff) failing to completely separate closely related populations, or in this case perhaps more accurately populations composed of similar elements.
Also 'Ashkenazi' shouldn't even be a category and many other categories would need to be reorganized / relabeled if they are indeed measuring ancient ancestry. But I don't think they are going specifically for that - quite the opposite in fact (see blog post). And if they *are* going for ancient ancestry then they should use ancient DNA of course. One of the reasons why Davidski's tests at Eurogenes tests are the best out there IMO, though I look forward to more. :)Replies: @Shaikorth
23andMe uses a different method than DNA-land. I’m especially certain it isn’t a sampling issue, their program and samples are public and the result is repeatable.
https://dna.land/faq#ancestry_algorithm
It simply is an issue of the method (ADMIXTURE and STRUCTURE-based stuff) failing to completely separate closely related populations, or in this case perhaps more accurately populations composed of similar elements.