What are diversity partitioning and clustering analysis, applied to hominids & race? [closed]

Question

Can anyone EMLI5 (or an undergrad) the terms diversity partitioning and clustering analysis please? From Winther paper over my head "The genetic reification of race" re: Lewontin-Edwards debate Thanks

Answer 1

Looking at the paper it seems to be talking about diversity partitioning as beginning with pre-identified groups & looking at gene variations. And clustering analysis as finding groups from the data.

Diversity partitioning is a technique used in community ecology, to simplify & make sense of trait diversity, by grouping variations that go together, so you don't have to talk about every single variation separately, & because variations very often do become grouped together. See another paper introducing the term, also. And this about resource partitioning is interesting, how specialising in certain strategies leads to divergence.

Cluster analysis is part of the area of finding patterns in data. In community ecology it can for instance be used to examine relative speciation between individuals. Have a look at the hominid branch of this diagram of human speciation as example. What are considered separate species, neanderthals & denisovans, are now understood to have merged with homosapiens, even though their differences would have made reproductive success much less frequent between species than within them. The point of no return in speciation is when chromosome numbers change, a merging of two chromosomes marked the divergence of hominids from other apes, & like the difference between horses & donkeys would prevent species merging permanently. But below that scale of variation, speciation occurs by degree.

We only have small numbers of gene sequences for hominids. We do have geographical information. But we have to infer the separation of groups from this kind of clustering analysis. There have been attempts to infer whether other unknown diverged-then-merged hominid groups existed, & there are indications of one in Southern Africa (probably climate variation isolated for some amount of time).

Humans are always going to be the most contentious species to apply tools like this to. The term race in biology has changed & sharpened, as DNA has helped us understand life. But culturally, it is still largely an atavism, that hugely over-prioritises skin colour, and largely ignores mixing between preconceived groups. You might like to have a look at some discussions in this forum about these issues:

Is biological ancestry distinct from race?

Are there philosophically serious moral arguments against eugenics?

Do IQ tests measure intelligence?

Human variation is unusually small, there is less total variation, than within just some individual breeds of dog. This is because of long bottleneck about 70,000 years ago that reduced hominid population numbers probably to below 1,000,maybe less. But, seems to have been linked with the emergence of more complex religious practices, & longer distance trading. Despite this small variation, there have been groups determined to emphasise difference over similarity, basically through all of human history - reification of folk intuitions about race, often going with beliefs that groups shouldn't mix. It does seem Europeans have a regionally important but low amount of neanderthal DNA that Afrucans don't, & that Asiatic peoples have this and denisovan traits, while Australia seems to have been reached directly from Africa along island chains around the equator & stayed isolated while some humans returned to Africa. So we can make sense of broad varuatiins of for instance facial features. But things like the rapid spread of lactose and alcohol tolerance, point to how traits are not isolated to racial groups. This is critically important in the most controversial area, intelligence, because it has to be understood that genes for greater intelligence have always been the most mobile of all, & environmental factors are gigantically more important than race in predicting intelligence. That's another way of saying, intelligence is not partitioned with race, they aren't correlated with each other in a statistically significant way (but, culture is partitioned with race).

This is a good & interesting paper, examining how preconceptions about groups can impact the use of these mathematical methods in ways that will match our intuitions, rather than be derived from the data. As it concludes, the functional roles of specific genes associated with different groups are largely poorly understood. But I would note, using AI to understand protein folding & a host of other technologies are going to answer many questions in the bear future. Our understanding of hominids has been transformed beyond recognition over the last decade, so we should take care to remain humble about what we have evidence for, & optimistic about we will understand soon.