Recently I was looking at a 3-D PCA animation which Zack generated from the Harappa Ancestry Project data set. Click the link and come back. Notice the outlier clusters? The Burusho are straightforward, they seem to have low levels of Tibetan admixture. But what about the Gujarati cluster? Again, we see what we’ve seen before, the fractioning out of the Gujaratis in PCA into two groups, one a tight cluster, and the other relatively widely distributed. This prompted me to look more closely at the HapMap Gujarati sample. Today I was exploring the question with Plink’s identity-by-descent feature. First I’ll start out with a smaller data set, my family (father, mother, sibling 1, sibling 2, and myself), and an Indian (from Uttar Pradesh) and Pakistani as unrelated individuals. I merged out 23andMe derived genotypes, and with ~900,000 markers calculated pairwise IBD:
The figure to the left is a three dimensional representation of principal components 1, 2, and 3, generated from a sample of Gujaratis from Houston, and Chinese from Denver. When these two populations are pooled together the Chinese form a very homogeneous cluster. They don’t vary much across the three top explanatory dimensions of genetic variance. In contrast, the Gujaratis do vary. This is not surprising. In the supplements of Reconstructing Indian population history it was notable that the Gujaratis did tend to shake out into two distinct clusters in the PCAs. This is a finding you see over and over when you manipulate the HapMap Gujarati data set. In reality, there aren’t two equivalent clusters. Rather, there’s one “tight” cluster, which I will label “Gujarati_B” from now on in my data set, and another cluster, “Gujarati_A,” which really just consists of all the individuals who are outside of Gujarati_B cluster. Even when compared to other South Asian populations these two distinct categories persist in the HapMap Gujaratis.
Zack has already identified a major difference between the two clusters: Gujarat_A has some individuals with much more “West Eurasian” ancestry. To be more formal about this in the future I simply assigned individuals in my merged data set to one of the two Gujarati clusters based on their position in the first two PCs. Yesterday night I ran ADMIXTURE K = 2 to 10, with 75,000 SNPs. I also removed the Native American groups, and added more European and East Asian samples from the HapMap. Below are some populations at K = 4: