search this blog

Monday, March 28, 2016

Higher than expected Denisovan admixture in South Asia


We've seen lots of papers in recent years on South Asian genetic history. Unfortunately, most of them have not been very useful. The only way to remedy this sad state of affairs is with at least a few prehistoric genomes from the Indian subcontinent. If this ever happens, we'll probably be treated to some surprises. A new paper at Current Biology suggests that inflated Denisovan-related ancestry in South Asian hunter-gatherers will be one of them:

Summary: Some present-day humans derive up to ∼5% [ 1 ] of their ancestry from archaic Denisovans, an even larger proportion than the ∼2% from Neanderthals [ 2 ]. We developed methods that can disambiguate the locations of segments of Denisovan and Neanderthal ancestry in present-day humans and applied them to 257 high-coverage genomes from 120 diverse populations, among which were 20 individual Oceanians with high Denisovan ancestry [ 3 ]. In Oceanians, the average size of Denisovan fragments is larger than Neanderthal fragments, implying a more recent average date of Denisovan admixture in the history of these populations (p = 0.00004). We document more Denisovan ancestry in South Asia than is expected based on existing models of history, reflecting a previously undocumented mixture related to archaic humans (p = 0.0013). Denisovan ancestry, just like Neanderthal ancestry, has been deleterious on a modern human genetic background, as reflected by its depletion near genes. Finally, the reduction of both archaic ancestries is especially pronounced on chromosome X and near genes more highly expressed in testes than other tissues (p = 1.2 × 10−7 to 3.2 × 10−7 for Denisovan and 2.2 × 10−3 to 2.9 × 10−3 for Neanderthal ancestry even after controlling for differences in level of selective constraint across gene classes). This suggests that reduced male fertility may be a general feature of mixtures of human populations diverged by >500,000 years.

Sankararaman et al., The Combined Landscape of Denisovan and Neanderthal Ancestry in Present-Day Humans, Current Biology (2016), http://dx.doi.org/10.1016/j.cub.2016.03.037

50 comments:

andrew said...

When you get down to the raw data, the implication seems to be not so much that South Asians have elevated Denisovan ancestry, as it is that all Asians without Papuan/Australian ancestry have similar levels of Denisovan ancestry in comparison to all West Eurasians who do not. All average about 0.05%-0.06% despite some lacking West Eurasian admixture and some having it.

The LD estimate of admixture for both Denisovan and Neanderthal is very recent (1000 +/- 8 v. 1121 +/- 16 generations), a result they attribute to incomplete ancient genomes. And, while it is clear that something is screwed up about the dates, their argument for the source of the problem has a lot of hand waving and isn't very convincing. Their explanation may be right but they don't prove it very powerfully.

Could archaic admixture actually have been ongoing in Asia until fairly recently (i.e. 30,000 years ago)? I don't see how it could in the case of Papuans and Australians whose admixture date should dominate the rest of the sample. It would have been interesting to see a breakdown of mainland v. non-mainland admixture date estimates.

Nirjhar007 said...

Uh I think its clearly related to the ASI/ASE Ancestry?.

Romulus said...

Would like to see Denisova scores for each pop, specifically tibetans who supposedly got their high altitude adaptaions from Denisova. Also Ainu. Would be neato to see individual Euro pop scores for Denisova and Neanderthal too.

capra internetensis said...

@andrew

They do say that lacking the proper recombination maps biases the date downwards. That's pretty far downwards, but I guess they know what they're talking about.

Also, it came out as basically one pulse of admixture, and since according to the other recent paper Oceanians shared in the same Neanderthal admixture as Eurasians, that would imply that it wasn't due to later interbreeding, and the uncorrected LD result really is that far off.

Davidski said...

Admixture dates based on LD decay and modern samples always look out of the ballpark.

Shaikorth said...

Romulus, Table S2.

http://www.cell.com/cms/attachment/2051794959/2059460386/mmc1.pdf

This study used Simon's genome dataset so unlike in most studies all genomes analyzed are high coverage sequences. The numbers look different to the usual percentages because of the methods they used, so look at the relative proportions instead of absolute.

andrew said...

The authors argument that any time adjustment would preserve relative dates is solid, but they barely discuss the recombination map biases (FWIW, it stuns me how little detail and analysis is present in papers in this field compared to other academic disciplines, even including supplemental materials.)

It isn't so much that I really think that the LD dates are wrong (although why report them with such tiny margins of error when you know they have massive systemic bias approaching 100% of the base number?). It is more that I'd like to see more analysis of why this possible source of error would provide such a big bias in one direction and that I'd like to see the magnitude of that error estimated from first principles instead of from other data points.

andrew said...

Correction: "I really think that the LD dates AREN"T wrong"

capra internetensis said...

@andrew

Yeah, I agree, I was kind of surprised they just left it at that.

Kristiina said...

Thanks Shaikorth, Sherpas have indeed a higher level of Denisovan than other East Asians. The highest are Papuans (1.123%), Australians (0.895%) and Bougainville (0.861%)! Then the percentage goes down considerably with Maori scoring 0.136%, Hawaiians 0.117% and Tibetan Sherpas 0.106%. All the rest are below 0.09. Many ASI populations are around 0.08%. The lowest of all in the table is Norwegian with 0.001%.

Chad Rohlfsen said...

Kristiina,
Papuans are 5% Denisovan.

rozenfag said...

@Chad Rohlfsen

She cites numbers from the paper. I don't know why they get lower percentage of a Denisovan admixture.

Shaikorth said...

The numbers are smaller because the percentages given on the table and fig. 2 are the proportion of confidently inferred SNP's of neanderthal/denisovan origin across the whole genome. It isn't contradictory to the previous estimates of total archaic ancestry which they also mention because there are also overlapping and low confidence SNP's that are not counted. So instead of looking at the absolute percentages, we should look at the proportions relative to each other (like "Papuans have X times more Denisovan than Tibetans").

Ryan said...

David - can you check if Sherpa ancestry correlates with Denisovan ancestry for South Asians / East Asians?

Rokus said...

Decay rates of DNA snippets, be it LD or Denisovan or Neanderthal or anything else, should depend heavily on the heterozygosity of a population. Homozygous populations simply wouldn't produce as much disruptive crossing overs, to the effect that segments of ancient admixtures would be preserved and still recognizable especially there, where modern admixture events were delayed - such as Melanesia and Australia.
It may easily be deduced this mechanism should complicate age estimation much more than admitted in this kind of studies. Most of the older segments would have been decayed beyond recognition and 'depleted' - save for a scatter of short ('atomic') IBD segments - by the impact of heterozygosity, much faster than merely warranted by dilution. Hence, without this effect clarified, dilution and the depletion of deleterious DNA by natural selection are just an assumption.
IMO South Asians once measured equal Denisovan admixtures as Australian aboriginals, and maybe the levels could still turn out comparable once we find a way to identify the bulk of decayed DNA impossible to recognize nowadays as part of an independent IBD segment.

Shaikorth said...

Don't think the modern South Asian numbers relative to Australians are going to turn out anything close, and the same goes for pure ASI if it's ever found. These are already full genome sequences with high coverage being compared, and the proportion of individual SNP's of reliably Denisovan origin in Australians is 10x the South Indian number.

Gill said...

Anecdotal but I got 3.8% Denisovan on Geno 2.0 and usually get some raised Papuan/Oceanian in most calculators. In West Eurasia K8 I had 16% ASE and 4% Oceanian. A couple other individuals got high Oceanian who had northernly origins in the subcontinent, near the mountains.

Davidski said...

Ryan,

I don't have the Sherpa dataset. I'd have to apply to get it.

Rokus said...

'proportion of individual SNP's of reliably Denisovan origin in Australians'
Define reliable origin. Being part of a segment long enough to have a reliable match? That would exclude all the individual SNP's of segments chipped up beyond recognition. As well the part of the original segment without marked SNP's that was shuffled into the genome like a deck of cards.

Shaikorth said...

It's not on a segment/IBD basis, more like SNP basis. The methods are explained in the supplementary data.

I'm actually certain that South Indian - or pure ASI - Denisova levels are nowhere near to that of Australians, and this can be seen in D-stats (African/PrimateOutgroup Denisovan ASI/Tribal Australian). Or IBS or anything that compares individual polymorphisms instead of segments. Especially when a high coverage dataset like in this study is used.

Ryan said...

David - might be worth getting whenever you have time I guess. They seem to be "special" so they may be worth including in various analyses.

terryt said...

"We document more Denisovan ancestry in South Asia than is expected based on existing models of history"

Perhaps the existing models of history are incorrect.

"Uh I think its clearly related to the ASI/ASE Ancestry?".

Yes, but it seems quite possible that the ASI is immigrant to South Asia from the east.

"Don't think the modern South Asian numbers relative to Australians are going to turn out anything close, and the same goes for pure ASI if it's ever found".

I think that is correct.

"the proportion of individual SNP's of reliably Denisovan origin in Australians is 10x the South Indian number".

Which is consistent with dilution outside Australia/New Guinea.

Rokus said...

It's not on a segment/IBD basis, more like SNP basis.
No, the study only focusses on SNP's detected in WINDOWS that are part of the genome whereof the 'union of detected Denisovan haplotypes spans 257 Mb in Oceanians'
Quote: 'We further restricted our analysis to windows in which the number of SNPs that pass filters is at least 1000 and over which the genetic length >= 2 cM'

My issue is just that I reject the author's presumption that 'at a slower linkage disequilibrium decay provides unambiguous evidence of more recent admixture'. It is just the opposite: that recent admixture speeds up decay of (e.g. Denisovan) haplotypes. Incomplete lineage sorting could then be responsible for some of the more extreme size differences between segments within the populations. IMO it is a conceptual error to assign admixture dates beforehand to segment sizes.

Hence my position that South Asia and Australia once harbored comparable high levels of Denisovan admixture from the same stabilized hybrid source (also confirmed by high Z scores), and that current geographic differences are NOT fully due to dilution.

Shaikorth said...

The study focuses on individual snp's when producing the proportions.

Quote: 'We further restricted our analysis to windows in which the number of SNPs that pass filters is at least 1000 and over which the genetic length >= 2 cM

is about a separate analysis of genomic regions depleted of archaic ancestry, not their analysis about the proportion of archaic SNP's. That may be confused by heterozygosity. The Denisovan polymorphisms are inferred using Africans and actual Denisovan, not from the combined Denisovan segments of Oceanians.

No mainland Eurasians even get significant Z-scores of Denisovan when compared to Han. I don't think there's much hidden in South Asians, at least not any more than in East Asians - which Sherpa, who peak the Denisovan in the mainland, are in a genetic sense. Certainly nothing approaching Oceanians.

terryt said...

"This suggests that reduced male fertility may be a general feature of mixtures of human populations diverged by >500,000 years".

I forgot to point out that this is normal for mammalian species. For example as the relationship between cattle species becomes greater the first thing to give out is male fertility. Female fertility remains for a little longer in the separation process. For example crosses between yaks and domestic cattle produce fertile females but infertile males. A similar phenomenon is so for horses as well. The half million year separation mark is obviously not a constant, in fact species separated by one million years are sometimes interfertile. Longer than that though infertility strikes.

Karl_K said...

Rokus is kind of right, but mostly wrong.

Imagine that there was a chimp human hybrid. 50% chimp at 100% of autosomal regions.

Looking at only SNPs, you would get something like 49.9% definite chimp.

Looking at all positions as if they were SNPs, without looking at IBD segments, you would get 99.9% human (or 99.9% chimp), and 0.1% the other.

So if it actually occurred that crossovers were occuring everywhere and only after they happened, then the small segment containing the human SNP was passed on, then 99.9% of the genome could actually be IBD from chimp, and only the SNPs from human (0.1%) were IBD from human. So the resulting population would genetically be 99.9% chimp, but identical to humans in every way.

But that is a really dumb thing to think about. It doesn't happen in natural systems. The SNPs in neutral regions should give a good estimate of % Denisovan or Neanderthal or anything else, as long as we have a few reference sequences.

Rokus said...

So if it actually occurred that crossovers were occuring everywhere
Don't they?

BTW. I think there is insufficient ground to distinguish modern human heterozygosity from archaic admixture. Not all archaic admixture is contained in multi-allele autosomal structures like segments or IBD's that could easily be identified. Desintegration of segments imply the circulation of debris.

Karl_K said...

@Rokus

" 'So if it actually occurred that crossovers were occuring everywhere'
Don't they?"

Over a long enough timespan, they could be almost anywhere. But in each generation they occur only once or twice per chromosome.

Nearly all of the selection or random drift for a one time hybridization occurs in just a few big chunks of DNA in the first few generations.

It is impossible for natural selection to sort out specific neutral SNPs. In the lab, when we blindly backcross 1 specific allele from one mouse strain onto another, it takes ~20 backcrosses to get it to 99% of the desired strain background. With whole genome SNPs for each animal, it can be done in 10 generations.

For 2 alleles, it is usually quicker to do them seperately, and then cross the resulting strains at the end.

To eliminate even 50% of the Denisovan SNPs from an Australian, while keeping the rest of the genome the same, would take thousands of generations even while screening full genomes for the least Denisovan at each generation.

It is a crazy proposition.

It is possible that 50,000 years ago people in East and South Asia had a similar rate of Denisovan ancestry as does Australia now.

But it would have to have been diluted by different modern human DNA lacking that ancestry. It couldn't have just been selected out.

terryt said...

"It is possible that 50,000 years ago people in East and South Asia had a similar rate of Denisovan ancestry as does Australia now".

I think that is quite likely.

"But it would have to have been diluted by different modern human DNA lacking that ancestry. It couldn't have just been selected out".

Again I think that is quite likely. We can be sure that such genes in SE Asia and on the nearby islands have been considerably diluted by migration south from more northerly regions in the Eastern Neolithic.

Karl_K said...

@terryt

". We can be sure that such genes in SE Asia and on the nearby islands have been considerably diluted by migration south from more northerly regions in the Eastern Neolithic."

I think the dilution would have been long before the Neolithic.

If we suppose that a fast early migration mixed with first Neanderthals, and then with Denisovans, and quickly hopped over the Wallace Line when they got that far...

Very soon after a very closely related second migration (or just a seperate group out of the same initial migration that didn't meet the Denisovans) became much more successful, but never made it over the Wallace Line.

This would be a reasonable scenerio.

Rokus said...

'But in each generation [chromosomal crossovers] occur only once or twice per chromosome.'
How sloppy of me. For 2000 generations (40ky) this would be 20 crossovers per cM. This would break down one cM into 21 pieces. For humans 1 cM is 1.000.000 base pairs, 21 pieces would be on average ~47k base pairs.

Rokus said...

However, the real lengths distribution of IBD segments with the Neanderthal genome (the part of the IBD segment that matches the Neanderthal genome) has a global peak at 18,000 bp (75 kya) and a smaller peak at 43,000 bp (43 kya).
I doubt the segments at the lower end of the curve can be detected as easily. Rather not. Hence, they would disappear from the radar just like this.

Karl_K said...

"I doubt the segments at the lower end of the curve can be detected as easily. Rather not. Hence, they would disappear from the radar just like this."

If they were just broken up by recombination, they only disappear as IBD segments. They do not disappear as SNPs.

If Denisovan SNPs were present in Asian populations, you could see them using statistics. And you can see some, at a very low level. Using them same methods with Auatralians gives a much higher level. That is the point.

"How sloppy of me."

I don't know if it is sloppiness or not, but I can't tell what you are even defending.

"BTW. I think there is insufficient ground to distinguish modern human heterozygosity from archaic admixture. Not all archaic admixture is contained in multi-allele autosomal structures like segments or IBD's that could easily be identified."

The multi-locus segments were only searched for after it was clear which populations had the SNPs. You do not need any IBD segments to tell how much admixture there was, as long as you have reference genomes.

It is not that complicated.

terryt said...

"I think the dilution would have been long before the Neolithic".

Not necessarily so. The only real evidence we have for movement into SE Asia after the very early movements into and through the region is the early Neolithic movement. That period coincides with the period of change from 'Melanesian' features to more 'Mongoloid' features in peninsula SE Asia. I'm not claiming that to be the only movement, but it is certainly the major movement.

"If we suppose that a fast early migration mixed with first Neanderthals, and then with Denisovans, and quickly hopped over the Wallace Line when they got that far..."

And presumably they left remnant populations along the way, whatever the 'way' was.

"Very soon after a very closely related second migration (or just a seperate group out of the same initial migration that didn't meet the Denisovans) became much more successful, but never made it over the Wallace Line. This would be a reasonable scenario".

But 'very soon after' doesn't fit the available evidence apart, perhaps, for the movement that managed to reach the Andaman Islands. They presumably did change the makeup of the mainland population but haploid genetics (Y-DNA D for example) shows no major presence in peninsula SE Asia. The most interesting further investigation would be to see if Australian Aborigines have more or less of the Denisovan element than do people from New Guinea. I understand that just Cape York inhabitants have been tested in Australia so far, and they almost certainly have genetic input from New Guinea.

Karl_K said...

"But 'very soon after' doesn't fit the available evidence apart, perhaps, for the movement that managed to reach the Andaman Islands."

Well I am pretty skeptical that these remnant populations with high Denisovan ancestry were very common. They may have persisted in some harsh isolated locations (such as high altitudes). But without further evidence from ancient DNA, it doesn't seem likely.

However, I totally agree that it is a possibility. I wouldn't have expected that populations like Oase 1 would have been hanging around even 37,000 years ago, yet they clearly were. So I wouldn't doubt that there were populations that were even mostly Denisovan in some locations.

But, this scenerio would mean that nearly all of the hunter-gatherer population of South Asia was wiped out in the Neolithic. And that these people contributed very little to any other living population.

Karl_K said...

"And that these people contributed very little to any other living population."

Which clearly is not the case, if you are going by mtDNA and Y haplogroups.

"Y-DNA D (for example) shows no major presence in peninsula SE Asia.)

Shaikorth said...

Here's some new stuff: Southern route through India gets attacked, new proposition is that Australia got populated through a northern route, and India has major introgression from Southeast Asia. Perhaps ASI is not really Indian?

http://biorxiv.org/content/early/2016/04/06/047456

"Founder ages of M lineages in India are significantly younger than those in East Asia, Southeast Asia and Near Oceania. These results point to a colonization of the Indian subcontinent by modern humans carrying M lineages from the east instead the west side. Conclusions The existence of a northern route previously advanced by the phylogeography of mtDNA macrohaplogroup N is confirmed here by that of macrohaplogroup M. Taking this genetic evidence and those reported by other disciplines we have constructed a new and more conciliatory model to explain the history of modern humans out of Africa."

terryt said...

"Southern route through India gets attacked, new proposition is that Australia got populated through a northern route, and India has major introgression from Southeast Asia. Perhaps ASI is not really Indian?"

Wow!! I have recently decided that is exactly what the haploid evidence shows. Some great details in the paper:

"Clearly, the fossil record in East Asia would be more compatible with a model proposing an earlier exit from Africa of modern humans that arrived to China following a northern route, around 100 kya. Indeed, this northern route model was evidenced from the relative relationships obtained for worldwide human populations using classical genetic markers (Cavalli-Sforza et al., 1988; Nei and Roychoudhury, 1993) and archaeological record (Marks, 2005)".

"The founder age of M in India is younger than in eastern Asia and Near Oceania and so, southern Asia might better be perceived as a receiver more than an emissary of M lineages".

That fits with mt-DNA having diversified and expanded from somewhere near the Burma/South China/Northeast India region.

terryt said...

"Well I am pretty skeptical that these remnant populations with high Denisovan ancestry were very common".

But such a population must have been present in SE Asia at least, although Shaikorth's recent link shows that an admixture event in the Altai region is completely possible. The possibility of there being two separate populations in SE Asia, one with Denisovan admixture and one without, seems unlikely.

"But, this scenerio would mean that nearly all of the hunter-gatherer population of South Asia was wiped out in the Neolithic".

Well mt-DNA M must have arrived in South Asia long before the Neolithic, and now we can see that a diluted Denisovan element was presumably carried into South Asia via mt-DNA M. If the Denisova element had been introduced to SE Asia through the arrival of mt-DNA N and Y-DNA C it would make sense that the early ASI carried far less Denisova than do Australian Aborigines. Although the problem remains of how New Guinea with its mt-DNA M and Y-DNA K have such a high proportion. Perhaps a product of admixture in SE Asia?

"And that these people contributed very little to any other living population."

Who said that? Did I say it? And what people are referred to?

terryt said...

Apologies. I see you were quoting yourself.

"Y-DNA D (for example) shows no major presence in peninsula SE Asia.)"

Yet interestingly there is a whole basal branch found only in the Philippines so far (D2). I'm fairly sure traces of some D or other have been found in Sumatra as well. But the settlement of the Andamans post-dates the crossing of Wallace's Line and D in South Asia is also a relatively recent arrival, being virtually confined the northeast predominantly in Tibeto-Burman speaking people.

"And that these people contributed very little to any other living population."

The paper actually points out that M in SW Asia came via South Asia and so the population did contribute to living populations.

Rokus said...

'I can't tell what you are even defending.'
Nothing important for people that already have all the answers, I am sure. IMO most don't even have all the questions, being of more concern.

- the current results are apparently based on parental heterozygosity for each generation, on the pretext that archaic parental events were rare and that archaic admixture remained rare, even within small communities. Because crossovers for homozygous alignments of archaic segments would be irrelevant for the decay of segment size over time.

- modern human heterozygosity tend to be fully credited to a rare combination of mutation rates and a single origin, what would be at odds with archaic evidence.

- it is impossible to make a distiction between archaic(X modern) and modern de novo SNP's that occurred since admixture, I think ~300/cM during 40k.

- distribution curves of archaic segments become very tricky at the lower end of segment sizes that per definition would be ever harder to distinguish of segments that have a modern or other origin. Due to the above the same applies albeit in lesser degree to SNP's.

The approach that Neanderthal and probably Denisovan (I wouldn't be surprised to find out Denisovan YDNA to have been as divergent/incompatible or even as Neanderthal as well) actually had an admixture issue with modern humans is interesting. However, it doesn't imply such admixture problems were the rule. Neanderthal and Denisovan might have been the exception during a longer period while other archaics merged into the modern gene pool without similar problems. But admixture problems don't rule out admixture events over a larger period (IMO most likely, despite the simplistic models of recent-only admixture currently presented), they just explain the larger segments. The smaller segments confirm that admixture happened over a much larger period of time and may have been partially stalled once genetic incompatibility emerged (as a - apparently failed - biological protection against admixture). This latter is a known mechanism in biology and speciation.

Karl_K said...

"Neanderthal and Denisovan might have been the exception during a longer period while other archaics merged into the modern gene pool without similar problems."

So you are suggesting that much or all of modern human diversity is the result of admixture with unknown archaics. The SNPs then became unlinked due to recombination over a very long time period. This makes it impossible to say which SNPs came from admixture vs from mutation in modern humans.

Is that correct?

That could certainly be true for some SNPs. The only issue is that as you suggest that more of human genetic diversity came from admixture, then the last common ancector of modern humans has to actually be much younger than it appears.

If half of human genetic diversity came from admixture with archaics, then the last common ancestor of humans would have been only 100,000 years ago.

Is this what you are suggesting?

How do you explain the mtDNA and Y chromosome trees? The dating is significantly off?

terryt said...

"The only issue is that as you suggest that more of human genetic diversity came from admixture, then the last common ancector of modern humans has to actually be much younger than it appears".

Wouldn't it mean that the last common ancestor would actually be much more ancient? Because you would have to include the ancestor of all the archaic humans in that 'common ancestry'.

"So you are suggesting that much or all of modern human diversity is the result of admixture with unknown archaics".

I'm inclined to agree with Rokus here, if that is what he is actually suggesting.

"How do you explain the mtDNA and Y chromosome trees? The dating is significantly off?"

As is the case with most mammals we now can see that male infertility appears quite early during the interbreeding of what are tending towards becoming separate species. In other words hybrid males would not be able to form F2 hybrids and so one Y-DNA line would become extinct as one population became the majority. With further diversification between the populations female sterility would develop, leading to the extinction of one mt-DNA line. As a result haploid DNA has a more recent common ancestor than does diploid DNA. There would be no correlation in dating at all.

Karl_K said...


So I will try to walk through you scenario.

The genetics of modern human diversity in SNPS genome wide suggests that the species is 200,000 years old (if you think the data suggests an older or younger date, then all the dates can be adjusted without any issue).

You suggest that 50% of human diversity is from admixture with archaic peoples. If this was true, then the date of the last common population would be 100,000 years ago.

Let's suppose that 100,000 years ago, there was an isolated human population somewhere that we will call proto-modern-humans, so we don't confuse them with actual modern humans.

We see fossils of modern human looking people in China and the Middle East at this time. We see that the Altai Neanderthal had what appears to be modern human admixture dating from around this time. We can suppose that they were widespread.

However, the current population of actual modern humans outside of Africa seems to have a simple origin. If we exclude the known Neanderthal and Denisovan admixture, then they appear to have a common ancestral population ~70,000 years ago. This also fits with what we know about colonization of Australia and New Zealand, and the change in the fossil record, and the radioactive dating of fossils that have yielded DNA, all across Eurasia. So Eurasia seems consistent with this dating.

So what this discussion is mostly about is Africa. That is where most of actual modern human diversity lies.

So your suggestion is the single proto-modern-human population that expanded across the rest of the world also started expanding in Africa at the same time. We can suppose that the geographic location of the proto-modern-human population was North East Africa (just to simplify things).

Within Africa, they mixed with some unknown Archaic groups that boosted their genetic diversity.

We could suppose that 50% of the African diversity is from 5% admixture with archaic populations that were diverged from the proto-modern-humans by 800,000 years.

The archaic population could not have been specifically related to either Neanderthals or Denisovans, because then they would appear to form a clade with those species to the exclusion of Eurasians. So this seems reasonable up to this point.

However, several lines of evidence argue against this.

How does that work out with the mtDNA and Y haplogroups across Africa? If the proto-modern-humans expanded from an extremely small population 100,000 years ago, then why is there so much more haplogroup diversity in Africa? Why is that haplogroup diversity segregated into populations that largely match their autosomal diversity? Why is there so much less linkage for nearby autosomal SNPS all across Africa, even within small populations?

It seems like you are looking for the solution for a problem that doesn't exist.




Karl_K said...

Sorry, there was a huge mistake in my post above.

I said:
"We could suppose that 50% of the African diversity is from 5% admixture with archaic populations that were diverged from the proto-modern-humans by 800,000 years."

Actually that would not be nearly enough. To get the extra genetic diversity (appearing to be ~200,000 years between some modern African populations, and also betweenbeach of them and the non-African populations), we would need at least two seperate events of >20% admixture from isolated populations that diverged ~1,000,000 million years ago. Or we could have had one event of 20% admixture with a population that diverged ~2,000,000 years ago, followed by immediate rapid growth and diversification of the population within Africa.

The diversity within Africa is actually less confined to non-genic regions, which is not what is seen with the Neanderthal and Denisovan introgression in non-African populations. Yet, you are supposing that the populations involved were much more divergent, and much more prone to selecting out DNP diversity within genes.

None of it adds up. In the end, to match the data, you will have to push back the date of admixture with the mystery archaics to a date that is the same or older than the dates already used without it.

The only solution would be that Africans had small population sizes, which will be at odds with the fact that the SNP diversity and linkage of those SNPs within Africa shows that they had to have maintained large effective population sizes for >200,000 years.

Usually scientists look for solutions to problems that arise from the data not making sense. In this case, you are putting the cart before the horse. There is not areally any data to suggest this scenerio.

When it arises, then it will be extremely interesting.

In my opinion, the truth is very different. Modern humans probably arose and diversified within Africa much earlier than 200,000 years ago. Small isolated populations moved out once and a while. About 100,000-70,000 years ago a small population expanded with a new success into Eurasia due to new technology or genes that allowed the new technology. They mixed with archaic humans there. Some of this population soon back-migrated to Africa, and their technology helped their genes to spread quickly and widely.

So in my opinion, the proto-modern-human population may have been older and within Africa. But with the back-migrations from Eurasia, the common ancestor appears to be a little bit younger.

This also makes it appear that the proto-modern-human ancestor was a bit closer to Neanderthals and Denisovans.



Rokus said...

'How do you explain the mtDNA and Y chromosome trees? The dating is significantly off'
This can't be explained by drift, that is for sure. However, now we know that evolution of human YDNA indeed occurred at an appreciable rate, possibly supplying a feasible new version each 150.000 years. Natural selection would have driven obsolete versions to extinction already long ago. The same may apply to mtDNA, where indications of a quite divergent version was recovered in LM3, a pleistocene modern human. Mind that relict mtDNA or YDNA are no reliable indication of origin, only of the local selective pressure.

'Usually scientists look for solutions to problems that arise from the data not making sense. In this case, you are putting the cart before the horse.'
Not entirely true. African Nanderthal/Denisovan-like IBD's are notoriously small. Some modals are close to the point where the average number of archaic SNP's equals one. Similar segments, of course, also occur elsewhere. IMO this feature merges with common modern human heterozygosity.
For a fully diverged archaic genome, assuming fully heterozygous crossovers for each generation, decay towards segments of this size would take up to ~500.000 years. Sure, this still needs a correction for the number of additional mutations during this period in case of full genetic isolation. Additionally, an 'opposite' correction should cope with the fact that full heterozygosity for each generation everywhere is hardly realistic. Hence this result is a feasible time span wherein a moderately isolated human subspecies would be fully incorporated within the fold of mainstream contemporaneous humanity.

This evidence of ongoing admixture is a rejection of any single origin at ~200.000 years ago.

Indeed there are also sufficient anthrolopogical indications of ongoing admixture: contemporaneous humans always are more similar in between than to their supposed direct ancestors/descendents. Though extended extinction scenarios may be proposed as an explanation, involving virtually all fossil subspecies, IMO the 'braided stream' hypothesis is more likely. Reproductive isolation may only have been required for some specialists that were required to survive in extreme conditions, such as Neanderthal, though not all of those subspecies could survive without the benefit of modest admixture, and probably none survived to exist within the last million years.

'The archaic population could not have been specifically related to either Neanderthals or Denisovans, because then they would appear to form a clade with those species to the exclusion of Eurasians.'
However, the diminutive size of African archaic IBD's indicate that both (early?) Neanderthal and (early?) Denisovans were geographically in close range with the ancestors of Africans.

Karl_K said...

"IMO the 'braided stream' hypothesis is more likely."

Before 300,000 years ago, I have no objection. The line leading to modern humans could have mixed up with dozens of streams that were seperated from Neanderthals and Denisovans by similar amounts of time (even 300,000 years). Without ancient genomes, it is impossible to tell.

But the mtDNA and Y chromosome at least tell us that there was a single major population that dominated starting from around 300,000 years ago. There are no million year old divergent version because they were lost by drift, because they were small minorities.

So. If you want to say that there was archaic admixture all the time, it is possible. But is was always a minor thing. No existing group that branched off since 200,000 years ago had a major (>10%) genetic contribution from an archaic species. It would be too obvious from genome sequences and haplogroups.

Rokus said...

'So. If you want to say that there was archaic admixture all the time, it is possible. But is was always a minor thing.'
I would say it differently. The admixture rate has always been enough to keep the species together, and to take advantage of subspecies genes or 'genetic inventions' that made a difference. Moreover, like I said somewhere, mainstream humanity may have been a much older phenomenon than the isolation of Neanderthal/Denisovan subspecies, implying that differences are most likely to have accumulated at a much lower pace at places without the same geographic or genetic barreers: It takes 45k years of isolation to generate a single divergent SNP on average at an interval of 40k, and this would also be the IBD size after 37k years of full heterogenous admixture- almost enough time to also generate one random de novo mutation that is actually sub-population derived. This may have been the whole stretch between China and Ethiopia.

One more comment about the strange phenomenon of diminutive Neanderthal/Denisavan IBD's in Africans: this would be explained if Modern Humans came from e.g. Turkestan and had met with Neanderthal and Denisovans already before some migrated to Africa. Eurasian IBD's that derive from the same admixture events would either be older and smaller and indistinguisable from normal human variability, or larger due to the lower degree of heterozygosity that should cause a slower rate of decay. Another possibility would be that admixture was originally higher for those that migrated to Africa while the 'Turkestan' homeland population remained 'pure' for a longer period due to -indeed- isolation up to a more recent wave of (Eurasian) expansion. Just some alternative scenarios that crossed my mind. The high modern human rate of heterozygosity would plead for scenario 1, and doesn't exclude scenario 2, but pleads against scenario 3.

Rokus said...

I mean: [...]implying that WITHIN-POPULATION differences are most likely to have accumulated at a much lower pace at places without the same geographic or genetic barreers

Karl_K said...

@Rokus

I am sorry. I completely misunderstood your position. I hope you refine it so that it is more easily clearly conveyed.

The major difference is surrounding the "common ancestor" of ALL modern humans. In this case we have to throw out the extra obvious admixture in any lineage.