In the interview:
A lot of the difficulty in talking about race has been a lack of agreement on what “race” means. In the past, the idea of pure races also included an ordering of certain races as inherently superior to others. We reject this idea absolutely. However, that doesn’t mean that there are no genetic differences between populations of different ancestral origin. A few of our features use the genome-wide data of reference populations from around the world to trace the origin of pieces of an individual’s genome. Some customers have complex patterns depending on where their ancestors originated. These reference populations aren’t “races”; they’re representative samples of peoples who have lived in a single place for a very long time and have thus accumulated different sets of genetic variants over time.
That's a tricky piece of wordcraft -- they're not 'races'; "they're representative samples of peoples who have lived in a single place for a very long time and have thus accumulated different sets of genetic variants over time."
Uhh....I'm thinking that's pretty much the definition of race in a lot of textbooks...
The points she starts with -- both true -- are that human populations aren't isolated ("pure races"), and you shouldn't rank them ("inherently superior"). But those ideas conflict with the process, since the software commonly used in human genetics (like STRUCTURE and other programs) assumes a model in which originally isolated groups (otherwise known as pure races) mix together.
It is true that STRUCTURE (and similar tools) use a model in which each individual is assumed to have a share of genes from each of K different populations, that differ from each other in the frequency (and in some case co-occurrence, due to linkage) of different gene variants.
These K populations, are not, however, assumed to be "pure races", but rather simply populations that differ from each other in their genetic characteristics. The program itself does not make any assumptions as to why they differ from each other: it could be due to different types or intensity of gene flow, or to reproductive isolation. Distinctive gene frequencies may arise both due to "race purity" or in the presence of gene flow, provided that it occurs at a low enough level so that changes that occur in one population are not reflected immediately in the other, and their distinctiveness is maintained.
But, more importantly, the "pure races" model is an approximation whose validity can be checked. If gene flow is substantial, then our simplifying STRUCTURE view of the world will not result in distinct clusters, in which the great majority of individuals from a particular population have the greatest share of their ancestry from the same racial clusters.
In other words, the fact that we assume distinct "pure" clusters is no guarantee that we will get distinct races via an application of STRUCTURE. Try STRUCTURE with K=2 in a homogeneous population, and you won't get 2 distinct clusters, even though you made that assumption: you will get a bunch of individuals that belong -in various proportions- to the 2 clusters, but no 2 solid blocks of individuals, one of which belongs predominantly to the first, and the other to the second cluster.
What STRUCTURE and similar programs have repeatedly shown, is that humans do, in fact, have genomes that resemble each other in the way we would expect if there were pure races which have occasionally mixed in their peripheries.
Our understanding of genetic processes helps us understand, that this is not the result of separate creation of human races, but of long-term gene flow limitations due to geography and culture, that have allowed originally related human populations to evolve apart.
In contrast with a Gobineau-istic view of set primordial races becoming ever more mixed and indistinct, our modern understanding is that human races evolved in historical time, becoming ever more distinct and separate. This is not inevitable, and human history has been punctuated by episodes of intermixture as well as separation, but the overall thrust has been one of diversification and increased distinctness, which may, perhaps, be set back in the modern age due to increased ease of transporation.