Gorilla.HO 23andme format sample

What happens if you run a Gorilla through an ancestry DNA test designed strictly for humans? Does it break the system, or does it reveal something deeply fascinating about human history?

 First, we looked at Fst values, which measure genetic differentiation. Unsurprisingly, the gorilla is massively distant from all humans. But look closer at who it's closest to. The test ranks indigenous African populations like the Khoisan and Bantu as its 'nearest' human relatives, while groups like the Mixtec and Papuans are the absolute most distant.

 F2 values show mostly the same trends, with Sub Saharan Africans plotting closest to Gorillas. F2 is less susceptible to low heterozygosity, making it the ideal medium for the construction of a PCA. 

 Here is a 2D PCA constructed on the basis of an F2 matrix. As you can see, humans are very distant from apes, but, the sub-Saharan africans are marginally closer than eurasians. 

 Here’s something crazy.

 With a PCA based calculator designed for humans, Gorillas score entirely African. According to my andreimix non negative least squares model, the Gorilla sample can be modeled as over 90 percent south african hunter gatherer. Since andreimix has no anchor for non humans, unlike the F2 matrix, the Gorilla maxes out the African anchor, since it is marginally closer to gorilla than other human groups. 

 Gorillas are obviously quite different from Africans. But if you didn’t have access to F2 and Fst, you wouldn’t even know it. PCA based calculators for humans are prone to tunnel vision, because their dimensions are mathematically optimized to separate the specific populations you feed them, blinding them to anything that falls outside that narrow frame of reference.