Population-specific SNPs and archaic admixture in Homo sapiens

A population-specific SNP is one which both alleles occur in a population X and only one in the rest of mankind. The existence of such SNPs can be explained in three different ways:

1. Both alleles belong to the ancestral gene pool of H. sapiens but were lost (by selection or drift) in most populations except one.
2. Ancestral humans were monomorphic, but a new allele appeared by mutation in one population, and there has not been enough time or opportunity for it to spread to the others.
3. An additional allele introgressed into a population by admixture with a regional population of archaic humans; this is equivalent to (2), with the new allele appearing through admixture, rather than new mutation.

In order to identify alleles that are specific to the major human groups, I carried out an ADMIXTURE analysis of the Harvard HGDP set with K=5. The results can be seen below, and as is expected, five major clusters emerge:

ADMIXTURE outputs a *.P file of allele frequencies in the inferred components. I used this file to identify population-specific alleles. Specifically, I identified, for each of the 5 components, SNPs where they were polymorphic, but all the other 4 components were fixed. This is the harvest of such component-specific alleles:

Asian: 2,321
West_Eurasian: 4,516
African: 50,835
Australasian: 1,726
Amerindian: 62

Note, that these numbers do not reflect on the relative number of population-specific SNPs across the genome. Nonetheless, they do appear concordant with what we know about the diminution of genetic diversity away from Africa.

Note also that these SNPs were identified only on the basis of comparisons between modern populations.

In all the following experiments, I will calculate D-statistics of the form D(Pop1, Pop2, Neandertal, Chimp) and D(Pop1,Pop2, Denisova, Chimp) to assess whether Pop1 matches Neandertal/Denisova more than Pop2 does.

Using Asian-specific SNPs

Han matches Neandertal 69% more than Sardinian does
Japanese matches Neandertal 71% more than Orcadian does
Dai matches Neandertal 69% more than Mandenka does
Cambodian matches Neandertal 52% more than San does
She matches Neandertal 5% more than Miao does

These results are consistent with a large number of Asian-specific alleles having been inherited from Neandertals or a Neandertal-like population.

Using West Eurasian-specific SNPs

Sardinian matches Neandertal 63% more than Han does
Orcadian matches Neandertal  61% more than Japanese does
Italian matches Neandertal 55% more than Mandenka does
French matches Neandertal 57% more than San does
Tuscan matches Neandertal 11% more than Basque does

Again, these results are consistent with many West Eurasian-specific alleles having been inherited from Neandertals or a Neandertal-like population.

Using African-specific SNPs


Dai matches Neandertal 55% more than Mandenka does
Cambodian matches Neandertal 41% more than San does
Italian matches Neandertal 55% more than Mandenka does
French matches Neandertal 41% more than San does
BantuKenya matches Neandertal 11% more than MbutiPygmy does

It thus appears that a substantial number of African-specific SNPs make Africans appear less Neandertal-like than Eurasians. This is unexpected if African-specific SNPs are common human SNPs that were retained in Africa but lost Out-of-Africa due to a bottleneck, or if they are SNPs that appeared recently by mutation in the African population.

They are, however, consistent with the following model:

This model is consistent with the evidence: Eurasian-specific alleles tend to match Neandertals, consistent with Neandertal introgression into the population of Eurasians. But, African-specific alleles tend not to match Neandertals.

This can be explained by the presence of archaic African admixture that stems from before the common ancestor of modern humans and Neandertals.

Denisova admixture in Australasian-specific SNPs

Papuan matches Denisova 84% more than Mandenka does
Melanesian matches Denisova 85% more than San does
Papuan matches Denisova 84% more than Sardinian does
Melanesian matches Denisova 85% more than Orcadian does
Papuan matches Denisova 84% more than Japanese does
Melanesian matches Denisova 85% more than Dai does
Papuan matches Denisova 24% more than Melanesia does

It will thus appear that in many Australasian-specific SNPs, Papuans/Melanesians match the Denisovan allele much more often than other populations.

(As a sanity check for my calculations, I note that Reich et al. 2011 (Table 2) inferred that Bougainville Melanesians have 82% of the Denisova ancestry that Papuans do on the basis of genotype data, and hence, Papuans have (100-82)/82 = 22% more, which closely matches my 24% figure)

African diversity in perspective

It is often said that non-Africans harbor a subset of African genetic variation, but that is not really generally true in the strict mathematical sense of "subset". For the 50,835 African-specific SNPs it is true that Africans are polymorphic whereas all other populations are monomorphic. But, for the 2,321 Asian-specific, and 4,516 West Eurasian-specific SNPs it is Asians and West Eurasians respectively that are polymorphic, whereas Africans and all remaining populations are monomorphic.

By examining these special sets of SNPs, where a regional human population is polymorphic and all the rest of mankind is not, we have been able to show that relationships with archaic humans are amplified. The implication is direct: regional-specific variation in humans is in part the heritage of regional continuity in both Africa and Eurasia. However it was that H. sapiens came to dominate our planet, it was not by extinction of archaic humans.

Human population differentiation: tree-like divergence or admixture between divergent demes?

What can account for differences between continental human populations? There are two explanations: (1) Tree-Like Divergence and (2) Incomplete Admixture between divergent demes.

(1) Until quite recently, it was near-universally thought that modern humans are the descendants of a single recent African population. Differences between human groups were ascribed to the operation of genetic drift, natural selection, and new mutation, as modern humans left their primordial Eden and expanded to populate the rest of the globe. According to this model: humans became more different from each other over time.

(2) But, there is a different idea, the reverse of the previous one: that modern humans are descended from many regional groups of earlier hominins that were very highly differentiated from each other. These earlier groups did of course diverge from common ancestors, but in the remote past; perhaps they represent long branches stemming from Homo heidelbergensis. Gene flow between them intensified as humans became more numerous and more mobile, According to this idea, humans around the world became more similar to each other over time.

We are now in a position, through the power of ancient DNA, to answer this question empirically. I am pretty sure that we will answer it over the next decade. And, the way to answer it is simple: take an UP West Eurasian, an UP East Asian, and a LSA African. Under the model of tree-like divergence, these ought to be genetically closer than a living European, a living East Asian, and a living African are to each other, because ~30-40ky of drift and selection acted independently on the three branches:

I suspect that we will be quite surprised when we look at the data, for a number of reasons:

1. Low genetic diversity of Denisova hominin, consistent with a model in which human diversity is generated by admixture between populations with low intra-group (as in Denisova), but high inter-group diversity (see point #2).
2. Higher genetic divergence between Denisova and Vindija (across ~6Mm of distance) than between any two living humans from the entire globe.
3. Diminution of human cranial variability over time, and disappearance of archaic forms.
4. Possibility of greater Neandertal admixture in UP Europeans than in recent ones.

In any case, if the the UP/LSA folks look like close relatives, we will know for sure that tree-like divergence did indeed shape our differentiation: we started very similar, and ended up different. But if they're not, then the door will open for a greater appreciation of how substantial archaic admixture, tempered by gene flow, has been a major cause of differences between living populations of mankind.

Looking forward to finding out...