Post by history91 on Dec 25, 2016 8:13:35 GMT -5
Are All DTC Firms Underestimating African Ancestry in Southern Europeans?
www.unz.com/gnxp/are-all-dtc-firms-underestimating-african-ancestry-in-southern-europeans/
quote:
At the time Dienekes had a pretty strong critique of the paper: the authors assumed that Northern Europeans were an unadmixed reference population, when in fact these populations may have had mixture from East Asians. He presented PCA plots which illustrated the fact that CEU (whites from Utah with British and German ancestry) sample were shifted toward Chinese in comparison to Sardinians. All these years later I think that in fact there is another explanation besides East Asian admixture to explain this: the Chinese themselves have some admixture from West Eurasians. The Sardinians are notably lacking in some admixture components common among mainland Europeans, so the West Eurasian admixture into Chinese probably is closer to mainland Europeans than it would be to Sardinians.
So what about the Sardinian admixture from Africans? The same dynamic might be at play here: old admixture from “Early European Farmers” into Africans might explain why they’re closer to Africans.
But let’s assume that the Sardinian admixture from Africans is legitimate. What does that mean for estimates of ancestral derivation from continental populations than you see in the DTC personal genomics firms that report ancestry results? It can only mean that among people of Southern European ancestry a few percent of African ancestry is being “masked” because it is part of the reference population set. This came to my mind because a half Cuban friend of mine had ~2% African ancestry. Reasonable. When I checked by running unsupervised ADMIXTURE he had ~4%. Then I noticed that the Sardinian reference set was often in the 1-2% range. One explanation for the discrepancy then would be that a few percent of African ancestry in his genome was simply swallowed up by the reference population in the supervised learning framework.
www.unz.com/gnxp/are-all-dtc-firms-underestimating-african-ancestry-in-southern-europeans/
quote:
At the time Dienekes had a pretty strong critique of the paper: the authors assumed that Northern Europeans were an unadmixed reference population, when in fact these populations may have had mixture from East Asians. He presented PCA plots which illustrated the fact that CEU (whites from Utah with British and German ancestry) sample were shifted toward Chinese in comparison to Sardinians. All these years later I think that in fact there is another explanation besides East Asian admixture to explain this: the Chinese themselves have some admixture from West Eurasians. The Sardinians are notably lacking in some admixture components common among mainland Europeans, so the West Eurasian admixture into Chinese probably is closer to mainland Europeans than it would be to Sardinians.
So what about the Sardinian admixture from Africans? The same dynamic might be at play here: old admixture from “Early European Farmers” into Africans might explain why they’re closer to Africans.
But let’s assume that the Sardinian admixture from Africans is legitimate. What does that mean for estimates of ancestral derivation from continental populations than you see in the DTC personal genomics firms that report ancestry results? It can only mean that among people of Southern European ancestry a few percent of African ancestry is being “masked” because it is part of the reference population set. This came to my mind because a half Cuban friend of mine had ~2% African ancestry. Reasonable. When I checked by running unsupervised ADMIXTURE he had ~4%. Then I noticed that the Sardinian reference set was often in the 1-2% range. One explanation for the discrepancy then would be that a few percent of African ancestry in his genome was simply swallowed up by the reference population in the supervised learning framework.