Monday, March 28, 2016

Elevated Levels Of Denisovan Ancestry In South Asians

Razib Khan captures the key findings of a new study of Denisovan DNA in a variety of populations succinctly (read every sentence twice and ponder it; he packs a lot of content in this paragraph):
The South Asian groups consistently jump well above the trend line for inferred Denisovan as a function of shared ancestry with Australians. Also, if you look at the admixture patterns for Denisovan ancestry in South Asia you see they follow the ANI-ASI cline. That is, it seems to come into the South Asian populations through the “Ancestral South Indians.” Interestingly, the Onge sample of Andaman Islanders has less Denisovan than low caste South Asian groups, reminding us that though the Onge and their kin are the closest modern populations to the ASI, they are not descended from the ASI. The highest fraction of inferred Denisovan is in the Sherpa people of Nepal. . . . . The proportion of Denisovan in low caste South Asians indicates that the fraction in ASI was about at the same level as the Sherpa. I suspect that ASI and the Tibetan groups got their Denisovan via different paths, but it doesn’t seem like we know yet.
(The comparison is actually as a function of non-West Eurasian ancestry and not Australian ancestry, but his description is otherwise on point.)

The paper notes with respect to this finding that:
[W]e were surprised to detect a peak of Denisovan ancestry estimates in South Asians, both in the Himalayan region and in South and Central India. The highest estimate is in Sherpas (0.10%), who have a Denisovan point estimate about one-tenth of that seen in Papuans (1.12%). Although this is notable in light of the likely Denisovan origin of the EPAS1 allele that confers high-altitude adaptation in Tibetans, EPAS1 is not sufficient to explain the observation as Sherpas have the highest point estimate even without chromosome 2, on which EPAS1 resides. To determine whether the peak of Denisovan ancestry in South Asia is significant, we tested whether the Denisovan ancestry proportion in diverse mainland Eurasians can be explained by differential proportions of non-West Eurasian ancestry (as it is already known that there is more Denisovan ancestry in East Eurasians than in West Eurasians). For each Eurasian population X, we computed an allele frequency correlation statistic that is proportional to eastern non-African ancestry. . . . South Asian groups as a whole have significantly more Denisovan ancestry than expected (block jackknife Z score for residuals = 3.2, p = 0.0013 by a two-sided test for the null hypothesis that the Denisovan ancestry estimate in South Asians is predicted by their proportion of non-West Eurasian ancestry[.] . . . The signal remains significant (Z = 3.1) when we remove from the analysis five populations that have ancestry very different from the majority of South Asians (Tibetan, Sherpa, Hazara, Kusunda, and Onge); however, the signals are non-significant for Central Asians (Z = 1.2) and Native Americans (Z = 0.1). Taken together, the evidence of Denisovan admixture in modern humans could in theory be explained by a single Denisovan introgression into modern humans, followed by dilution to different extents in Oceanians, South Asians, and East Asians by people with less Denisovan ancestry. If dilution does not explain these patterns, however, a minimum of three distinct Denisovan introgressions into the ancestors of modern humans must have occurred.
To take this in stride, however, the estimated percentage of Denisovan ancestry in South Asians is only 0.06% +/- 0.03% (and 0.01% +/- 0.03% on the X chromosome) v. 0.06% +/- 0.02% (and 0.00% +/- 0.01% on the X chromsome) for East Asians.  Native Americans and Central Asians are 0.05% +/- 0.01% (and 0.00% +/- 0.00% on the X chromosome).  West Eurasians average 0.02% +/- 0.01% (and 0.00% +/- 0.00% on the X chromsomes).

Thus, on average, South Asians, East Asians, Native Americans, Central Asians and the Onge all have mean Denisovan ancestry between 0.05% and 0.06%, while Oceanians are at 0.85% +/- 0.43% (and 0.18% +/- 0.17% on the X chromosome), and West Eurasians are at 0.02%.

The results are generally consistent with admixture non-West Eurasians at the same rate subject to minor further dilution by West Eurasians in Central Asians and Native Americans, with additional admixture in Papuans and Australians (which enhances the percentages of populations admixed with them) which was only mildly diluted, and additional admixture (pre-dilution) in Tibetans.  If so, this would imply that Denisovan admixture took place somewhere in the vicinity of Afghanistan and the Indian subcontinent, rather than further east, right around the time of the West Eurasian-East Eurasian population split.

Razib Khan also notes this language from the abstract of the paper (emphasis his):
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.
The paper explains that the raw data call for ridiculously recent dates of admixture (1000 +/- 8 generations for Denisovans and 1121 +/- 16 generations for Neanderthals) but explains that all likely sources of bias (mostly incomplete genome samples and dramatic population events) which could make the linkage disequalibrium estimates seem younger than they are would impact both admixture date estimates proportionately.  So, Denisovan admixture happened about 10% fewer years ago than Neanderthal admixture, which based upon other estimates of Neanderthal admixture dates would suggest a date about 6,000 years after Neanderthal admixture (ca. 44,000-54,000 years ago).

We know a lot about Denisovans (with some people having as much as 5% Denisovan DNA) for a population that we still have not yet managed to associate with skeletons that could tell us which archaic hominin species they belong to, or what they looked like in broad general outline.

The big new open question is "How did South Asians end up with elevated levels of Denisovan ancestry in a scenario in which the Onge did not?"

I don't have an answer to that one yet.

Raw Data By Population Below The Break


Populations Neanderthal Denisovan ancestry ancestry A X A X (%) (%) (%) (%)

Abkhasian 0.976 0.100 0.011 0.000
Adygei 1.126 0.119 0.020 0.000
Albanian 1.203 0.334 0.019 0.000
Aleut 1.357 0.368 0.044 0.000
Altaian 1.413 0.445 0.064 0.000
Ami 1.440 0.183 0.047 0.019
Armenian 1.077 0.121 0.013 0.000
Atayal 1.531 0.785 0.062 0.000
Australian 1.559 0.300 0.895 0.105
Balochi 1.070 1.046 0.026 0.000
Basque 1.100 0.098 0.011 0.000
BedouinB 0.858 0.386 0.007 0.000
Bengali 1.261 0.268 0.063 0.000
Bergamo 1.134 0.015 0.020 0.000
Bougainville 1.622 1.375 0.861 0.032
Brahmin 1.101 0.635 0.064 0.000
Brahui 1.099 0.261 0.018 0.000
Bulgarian 1.078 0.250 0.005 0.000
Burmese 1.334 0.427 0.057 0.000
Burusho 1.272 0.200 0.035 0.061
Cambodian 1.419 0.538 0.075 0.000
Chane 1.338 0.761 0.042 0.000
Chechen 1.019 0.000 0.025 0.000
Chipewyan 1.633 0.384 0.049 0.000
Chukchi 1.228 0.161 0.040 0.000
Cree 1.260 0.126 0.057 0.000
Crete 0.993 0.187 0.014 0.000
Czech 1.067 0.000 0.028 0.000
Dai 1.314 0.211 0.064 0.014
Daur 1.359 0.475 0.067 0.010
Druze 0.965 0.186 0.011 0.000
Dusun 1.438 0.312 0.086 0.000
English 1.085 0.210 0.015 0.000
Eskimo Chaplin 1.500 0.000 0.053 0.000
Eskimo Naukan 1.401 0.408 0.060 0.000
Eskimo Sireniki 1.491 0.265 0.051 0.000
Estonian 1.076 0.167 0.021 0.000
Even 1.411 0.229 0.064 0.000
Finnish 1.165 0.302 0.013 0.000
French 1.023 0.188 0.012 0.000
Georgian 1.134 0.000 0.012 0.000
Greek 0.975 0.579 0.005 0.000
Han 1.495 0.144 0.062 0.005
Hawaiian 1.342 0.184 0.117 0.000
Hazara 1.225 0.324 0.034 0.000
Hezhen 1.399 0.277 0.053 0.000
Hungarian 1.122 0.057 0.019 0.000
Icelandic 1.237 0.147 0.015 0.000
Igorot 1.399 0.503 0.048 0.000
Iranian 0.968 0.351 0.022 0.000
Iraqi Jew 0.926 0.231 0.020 0.000
Irula 1.199 0.212 0.089 0.000
Itelman 1.428 0.042 0.045 0.000
Japanese 1.308 0.444 0.058 0.000
Jordanian 0.810 0.282 0.005 0.000
Kalash 1.113 0.409 0.025 0.000
Kapu 1.069 0.705 0.055 0.000
Karitiana 1.374 0.120 0.037 0.000
Kashmiri Pandit 1.175 0.235 0.041 0.000
Kharia 1.133 0.380 0.085 0.000

Populations Neanderthal Denisovan ancestry ancestry A X A X (%) (%) (%) (%)

Khonda Dora 1.207 0.157 0.086 0.000
Kinh 1.448 0.433 0.052 0.000
Korean 1.457 0.539 0.062 0.000
Kurumba 1.313 0.751 0.081 0.000
Kusunda 1.256 0.581 0.075 0.061
Kyrgyz 1.306 0.101 0.040 0.000
Lahu 1.358 0.075 0.061 0.000
Lezgin 1.125 0.338 0.014 0.019
Madiga 1.126 0.795 0.073 0.000
Makrani 1.041 0.141 0.015 0.000
Mala 1.127 0.527 0.052 0.000
Mansi 1.311 0.091 0.040 0.000
Maori 1.252 0.000 0.136 0.000
Mayan 1.386 0.183 0.069 0.000
Miao 1.341 0.151 0.073 0.000
Mixe 1.342 0.222 0.048 0.000
Mixtec 1.252 0.414 0.044 0.000
Mongola 1.389 0.346 0.068 0.000
Nahua 1.332 0.263 0.046 0.000
Naxi 1.371 0.106 0.070 0.000
North Ossetian 1.079 0.226 0.013 0.000
Norwegian 1.157 0.297 0.001 0.000
Onge 1.325 0.533 0.057 0.000
Orcadian 1.132 0.077 0.004 0.000
Oroqen 1.399 0.540 0.059 0.000
Palestinian 0.909 0.074 0.010 0.000
Papuan 1.596 0.366 1.123 0.269
Pathan 1.097 0.469 0.041 0.000
Piapoco 1.318 0.236 0.053 0.000
Pima 1.437 0.266 0.052 0.000
Polish 1.086 0.240 0.036 0.000
Punjabi 1.156 0.156 0.061 0.000
Quechua 1.361 0.333 0.045 0.000
Relli 1.190 0.572 0.064 0.019
Russian 1.148 0.243 0.018 0.000
Saami 1.363 0.000 0.028 0.000
Samaritan 0.888 0.000 0.002 0.000
Sardinian 1.133 0.200 0.009 0.000
She 1.468 0.224 0.077 0.000
Sherpa 1.395 0.250 0.106 0.000
Sindhi 1.174 0.188 0.048 0.022
Spanish 1.031 0.130 0.018 0.000
Surui 1.446 0.011 0.050 0.000
Tajik 1.064 0.068 0.016 0.000
Thai 1.458 0.584 0.048 0.000
Tibetan 1.389 0.169 0.082 0.010
Tlingit 1.261 0.211 0.042 0.000
Tu 1.466 0.232 0.045 0.000
Tubalar 1.391 0.261 0.052 0.000
Tujia 1.430 0.266 0.092 0.010
Turkish 1.024 0.226 0.014 0.000
Tuscan 1.151 0.131 0.016 0.000
Ulchi 1.508 0.177 0.064 0.000
Uygur 1.170 0.398 0.057 0.019
Xibo 1.437 0.438 0.066 0.000
Yadava 1.157 0.469 0.047 0.000
Yakut 1.525 0.155 0.070 0.000
Yemenite Jew 0.947 0.277 0.012 0.000
Yi 1.387 0.036 0.070 0.000
Zapotec 1.360 0.329 0.051 0.000

6 comments:

Kodhambo Madhathakaja said...

Conjecture is some indian admixture from middle upper Paleolithic hominids like those found at jwalapuram by Petragila, Clarkson et al. They lived both, before and after Toba explosion (conjecture). Even if I am aware that denisovan dna may not be provided by the Paleolithic Neanderthals, note that all populations including, Punjabis, Brahmins, Bengalis have denisovan at 0.06 percentage. So I think it is something basal.

terryt said...

"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"

Interesting.

"If so, this would imply that Denisovan admixture took place somewhere in the vicinity of Afghanistan and the Indian subcontinent, rather than further east"

Not necessarily so. That idea comes from what I think is an over-simplification of modern human expansion.

"The comparison is actually as a function of non-West Eurasian ancestry and not Australian ancestry, but his description is otherwise on point".

Again not necessarily so. Note:

"The South Asian groups consistently jump well above the trend line for inferred Denisovan as a function of shared ancestry with Australians".

That may be looking at it in the wrong direction.

"if you look at the admixture patterns for Denisovan ancestry in South Asia you see they follow the ANI-ASI cline".

What if the ASI originates from a movement west from Australia/New Guinea? In other words what if the ASI is not a remnant of an 'original' migration east through South Asia but a product of an original movement east into South Asia. That actually fits the admixture evidence rather well. But it would mean that South Asia was not the only, or even mainly, the route modern humans took first in their eastward movement.

andrew said...

"What if the ASI originates from a movement west from Australia/New Guinea?"

This is pretty much ruled out by the uniparental haplogroup data.

terryt said...

"This is pretty much ruled out by the uniparental haplogroup data".

I don't think that is so at all. I agree Y-DNAs F and H show movement into South Asia from the west but anything derived from K shows the opposite, as does C. When we turn to mt-DNA nothing shows an eastward movenet through South Asia. Sure, some show movement into South Asia from the west but these get no further than South Asia itself. M is most likely to have entered from the east as is R. The uniparental haplotypes provide plenty of evidence for east to west movement into South Asia.

andrew said...

In fairness, I don't mean to say that there is never East to West migration. The Munda, the Malagasy people of Madagascar, and the Roma are three examples. And, it seems to be shaping up that Siberia has toggled back and forth between periods of Eastern to Western dominance (e.g. the Turks, the Mongols v. the Tocharians and the Russians) perhaps half a dozen to a dozen times since Out of Africa.

But, it is also pretty self-evident that the first wave had to be West to East. The evidence that humanity originated in Africa is pretty overwhelming. And, the evidence of any modern humans lived east of India on the Southern route pre-Toba is very thin and not very convincing. I know that there are a handful of alleged Chinese outliers, but none are sufficiently convincing yet to persuade me that these were real AMHs at the pre-Toba dates claimed.

terryt said...

"it is also pretty self-evident that the first wave had to be West to East".

Agreed. But the route is far from self-evident. In fact semi-open grassland was present through much of Central Asia but less so through much of South Asia at appropriate times.

"the evidence of any modern humans lived east of India on the Southern route pre-Toba is very thin and not very convincing".

And non-existent in South Asia.