July 17, 2011

A good idea for a new project

I was a little disappointed because the excellent new paper by Li and Durbin had not included full genomes from Palaeoafrican individuals, as these are, perhaps, the most interesting ones in terms of the deep ancestry of our species.

I was then reminded, that a full genome of Khoisan individual (KB1) was, in fact, published by Schuster et al. in 2010, and both the paper and the genome are freely available online.

Why is this interesting? Consider the following figure from Schuster et al. (2010):

Notice that the African hunter-gatherer (KB1) has 1,704 private SNPs compared to a Yoruba (NA19240) and Archbishop Desmond Tutu (ABT), and 2,038 SNPs compared to a European American (J. C. Venter), and a Chinese (YH).

This amount of private variation admits to two explanations:
  1. Higher effective population size in Khoisan
  2. Deep population structure followed by admixture
As I have noted in my review of Li & Durbin (2011) UPDATE III, the effective mutation rate is in fact dependent on the effective population size, and it seems almost certain that a lower effective mutation rate must be used in a population of higher effective size.

There is no mystery why this is the case: accumulated genetic variation is a consequence of the mutation rate (how aggressively variation is introduced), and the effective population size (which controls how severely variation is lost due to drift).

A substantial difference in effective population size means that almost certainly the indiscriminate use of a single 2.5x10-8 mutation rate for different human populations is unwise.

This is a serious limitation, as far as I can tell, of the PMSC method introduced by Li & Durbin, as it assumes a single mutation rate parameter which is then used to estimate past population sizes.

In any case, it would be interesting to see how far back the divergence of the Khoisan individual from other humans will be, even if the 2.5x10-8 rate is employed, how large the Khoisan effective population will be, and also what antiquity of population substructure followed by admixture within Africa will be sufficient to "save the phenomena."

Another interesting observation is that the genealogical autosomal mutation rate in humans (1.1x10-8) is actually lower than the estimated evolutionary rate from human-chimpanzee divergence (2.5x10-8)

Nothing in evolutionary biology can account for such a discrepancy, I think, unless there is extreme balancing selection maintaining variation across the entire genome.

So, either:
  1. There is a serious flaw in the genealogical rate as estimated from 1000 Genomes trios, or
  2. We are about to find out that quite deep population structure and admixture played a role in the history of the genus Homo, deep in a sense of human-ape interbreeding after Homo-Pan speciation 7 million years ago, an idea that was proposed, for different reasons, a few years ago
Calibration of the mutation rate is, of course, quite important for correlating genetic with archaeological events.

For example, Li & Durbin propose that gene flow between Eurasians could have been effected during the Ice Age, as they retreated southwards; such a proposal is necessary to account for divergence between Europeans and East Asians of ~20ky, which is about half the earliest known colonization of Europe. Halving the mutation rate harmonizes the genetic divergence with archaeology, but would push the divergence of Eurasians from West Africans to the dawn of anatomical modernity, and African hunter-gatherer antiquity well beyond it.

I predict that the next few years will reignite many old debates in anthropology.

7 comments:

terryt said...

"2.We are about to find out that quite deep population structure and admixture played a role in the history of the genus Homo, deep in a sense of human-ape interbreeding after Homo-Pan speciation 7 million years ago, an idea that was proposed, for different reasons, a few years ago"

That paper made sense to me then, and still does today:

http://genepath.med.harvard.edu/~reich/Patterson%20et%20al.2.pdf

Why would it not be so?

pos said...

It would definitely be interesting to apply the method to more African populations. However, the KB1 genome was identified by the authors as having recent admixture in its ancestry, something that could strongly bias inferences based on heterozygosity.

jes-r said...

The KB1 individual does seem to have very minor Bantu admixture, probably less than 10% judging from this PCA:
http://www.nature.com/nature/journal/v463/n7283/fig_tab/nature08795_F4.html

However, the NB1 and TK1 samples seem to be pure Khoisan.

Liken said...

Hi Dienekes. What do you think about this new interesting paper (non-Africans are hibridated with neandertals, X haplotype)

http://mbe.oxfordjournals.org/content/early/2011/01/25/molbev.msr024.short

German Dziebel said...

"such a proposal is necessary to account for divergence between Europeans and East Asians of ~20ky, which is about half the earliest known colonization of Europe. Halving the mutation rate harmonizes the genetic divergence with archaeology, but would push the divergence of Eurasians from West Africans to the dawn of anatomical modernity, and African hunter-gatherer antiquity well beyond it."

Even if we don't go as far as the out of America theory, in the New World, linguistic diversity seems to be a better indicator of population antiquity than archaeology because low population size and density affects archaeological visibility but not linguistic diversity. Judging by the level of linguistic diversity, Amerindian populations may be at least three times older (see Nichols, "Linguistic diversity and the first settlement of the New World." Language 66:3. (1990)) than the dates provided for the peopling of the Americas in many genetics publications (e.g., 12,000 years). Using linguistic evidence - language being a key attribute of behavioral modernity - puts the antiquity of modern humans in America roughly within the time frame of their archaeological attestation in, say, Europe.

We'll have to divide the mutation rate by three in order to harmonize linguistics and genetics in this case. This will create an entirely unrealistic dates for the divergence of West Africans from Eurasians and African foragers from West Africans.

In the Old World, quite naturally, the increase in archaeological visibility and the decrease in linguistic diversity are both functions of growing population size. Even if we halve the mutation rate to account for the archaeology-genetics discrepancy in the Old World, we still end up having Khoisans, who are absolutely and unequivocally behaviorally modern humans, separate from the rest of humanity hundreds of thousands of years before the emergence of archaeological signatures of behavioral modernity and do so without spawning any linguistic diversity of their own.

terryt said...

"linguistic diversity seems to be a better indicator of population antiquity than archaeology because low population size and density affects archaeological visibility but not linguistic diversity".

Wouldn't linguistic diversity be a function of population size? Languages are more likely to break up in a population scattered thinly across a large region than in a larger one with more contact between groups. You even hint at the possibility:

"In the Old World, quite naturally, the increase in archaeological visibility and the decrease in linguistic diversity are both functions of growing population size".

German Dziebel said...

"Wouldn't linguistic diversity be a function of population size?"

Yes, that's exactly what I'm saying. Regional differences in population size explain nicely 3 facts: 1) low intragroup and high intergroup genetic diversity in the New World vs. high intragroup and low intergroup genetic diversity in Africa; 2) high linguistic diversity in the New World (and PNG) vs. low linguistic diversity in Africa (and Europe); 3) high archaeological visibility of behaviorally modern humans in Africa and Europe vs. low archaeological visibility of behaviorally modern humans in the New World (and PNG/Australia). All these things are logically connected because observable, leave-behind traces of modern human behavior in the archaeological record correlate with observable attributes of modern human behavior that are inherited from generation to generation (languages).

"Languages are more likely to break up in a population scattered thinly across a large region than in a larger one with more contact between groups."

Sure, but languages don't emerge ex nihilo. It's not that Amerindians had bigger brains that could spawn languages at a faster rate. It takes time to accumulate diversity through stepwise branching. Languages in the New World and Oceania seem to have more conservative grammatical features (Nichols 1992) and kinship structures (Dziebel 2007) than languages in Africa and Europe. Human population size has only been steadily increasing throughout human history. Mid-Pleistocene hominid populations must have been low in population size, low on intragroup genetic diversity and high in the diversity of primitive speech forms. Hence, we arrive at a logical conclusion: behaviorally modern human populations in the New World and Oceania must be "older" and preserve the linguo-demographic parameters of Mid-Pleistocene hominid populations better than than populations in Africa and Europe.