Denisovans, Neanderthals, Yetis, oh my!

An excellent open access paper is out in Cell which explores the distribution of archaic hominin, and in particular Denisovan, ancestry, Analysis of Human Sequence Data Reveals Two Pulses of Archaic Denisovan Admixture:

Anatomically modern humans interbred with Neanderthals and with a related archaic population known as Denisovans. Genomes of several Neanderthals and one Denisovan have been sequenced, and these reference genomes have been used to detect introgressed genetic material in present-day human genomes. Segments of introgression also can be detected without use of reference genomes, and doing so can be advantageous for finding introgressed segments that are less closely related to the sequenced archaic genomes. We apply a new reference-free method for detecting archaic introgression to 5,639 whole-genome sequences from Eurasia and Oceania. We find Denisovan ancestry in populations from East and South Asia and Papuans. Denisovan ancestry comprises two components with differing similarity to the sequenced Altai Denisovan individual. This indicates that at least two distinct instances of Denisovan admixture into modern humans occurred, involving Denisovan populations that had different levels of relatedness to the sequenced Altai Denisovan.

Before you get caught up in the results, you should check out the methods. They’re pretty ingenious. Though with novel results like this people also really need to work their way through them as well (the authors present a lot of simulation results to validate the method, so I’m sure that will convince most; it certainly sways me).

The plots at the top of this post show the different distribution of Neanderthal and Denisovan admixture, by matching regions of the genome that they’ve identified as archaically introgressed. The ultimate logic is to look for variants which aren’t found in Africans, and are found in non-Africans, and scan over segments of the genome hoping that you can pick up the haplotypes that would slowly be chopped up over time through recombination that came in from Neanderthals or Denisovans.

At the top-left of the figure, you see “Northwest Europeans.” The segments tend to concentrate at the bottom-right of the panel. That means that they match the Neanderthal reference sequence to a high degree, but not the Denisovan. This makes sense since everything we know from earlier work indicates that Northwest Europeans don’t have Denisovan ancestry.

On the bottom-right you see Papuans. They’re very out of place because they are the only population in the list where Denisovan ancestry is greater than Neanderthal ancestry. This is visible in the match patterns.

South and East Asian populations exhibit a pattern with high (relative) levels of Neanderthal matches, but also a minor amount of Denisovan matching. This aligns with earlier work, which reported low levels of Denisovan admixture among populations with eastern Eurasian ancestry broadly.

The surprise is that the variation in matching to the Denisovan Altai genome exhibited a north-to-south cline. In particular, Northeast Asian populations seem to have a mix of two types of Denisovan. One, which is close to the Denisovan sequence that is normally used as a reference, and one which is diverged from it. The Papuans and South Asians seem to have Denisovan ancestry which is not so much like the Altai sample. This is not very shocking of course.

Finns barely miss the p-value cut-off (Bonferroni-corrected threshold), but they clearly have some Denisovan from East Asian gene flow, and some of it looks to be similar to the Altai Denisovan. Curiously, the Vietnamese (Kinh) don’t show any Altai Denisovan, but the Dai do. The Japanese have a lower proportion of the Altai Denisovan than the two Han Chinese samples. And very strangely the 1K Genomes samples from the New World, a substantial proportion of which have Amerindian admixture, show no Denisovan.

Pontus Skoglund immediately made a very interesting observation:

And Alexander Kim followed up:

In the thread to Skoglund’s original comment Africa Gomez notes that the authors suggest that high linkage disequilibrium in New World populations, due to recent admixture between diverged groups, may reduce the power to detect the Denisovan ancestry. So perhaps that’s that?

But for a moment, let’s set that aside. The best evidence right now is that the Denisovan admixture into Papuans, and therefore South Asians, occurred not too after the Neanderthal admixture event. That mixture is reasonably well dated because of ancient genomes which are closer to the period of admixture. But what about the second event with the Altai Denisovan? If what Skoglund says is true the date for that might be closer to the Last Glacial Maximum, and not when modern humans came to dominate the region. And I say dominate because there’s evidence that anatomically modern humans may have ventured quite far into eastern Eurasia before they finally swept aside more established lineages.

A few years back researchers found that one of the mutations that allow for Tibetan high altitude adaptation seems to have come in from a Denisovan genetic background. Spencer Wells, who knows a thing or two about Central Asia, has always half-seriously suggested that the legends of the Yeti derive from populations of archaic humans who persisted in the uplands of the heart of Eurasia.

But perhaps they weren’t pure Denisovans in any case. Work out of David Reich’s lab has suggested that Denisovans themselves, or at least the Alta Denisovan, harbors a deep ancient lineage diverged from modern humans, Neanderthals, and Denisovans, in low fractions. The “Altai Denisovan” admixture may have come into Northeast Asians via a mixed population, which arose when modern humans came to dominate eastern Eurasia, but only transmitted the Altai Denisovan ancestry later.


8 thoughts on “Denisovans, Neanderthals, Yetis, oh my!

  1. The Native American lack of Denisovan there is probably a methodological issue, whether ld-related or not. Previous studies like Mallick et al. on Simons Genome Diversity project detected Denisovan in Natives (Karitiana, Pima and many more). Supp. table 1, column AK here, it looks like they had as much Denisovan as Siberians.

    Siberians, some Central Asians like Hazaras and Native Americans had Denisovan estimates between 0.003-0.004, more western populations from Tuscans to Saamis and Samaritans 0.001-0.002. 0.005 and above in some Indian tribals like Irula as well as in Tibet beyond the Southeast Asian hotspot.

    This method, based on rate of nd10 sites appears to gives false Denisovan ancestry to African hunter-gatherers but should be informative OOA.

  2. Qin and Stoneking found small % but ubiquitous Denisovan in Native Americans. Racimo et al. report TBX15/WARS2 alleles in Inuit and other Native Americans that apparently introgressed from Denisovans and reached near-fixation probably via cold-adaptation selection in Beringia. Obviously it would be most interesting to see comparable plots for Inuit, Surui and Karitiana, and Tibetans.

  3. It seems particularly tricky when comparing these patterns with the current models of pop structure from latest papers on East Eurasia (Southeast Asia and Tianyuan ancient dna). Those models have a split like:

    Proto All East Eurasian -> split of Tianyuan from other East Eurasians -> split of Papuan from remaining East Eurasian -> split of Onge + East Asian (with admixture of Tianyuan back into East Asian)

    And Onge is presumed to be the most similar reference to the Indian population that contributed (or just simply was) ASI (the Ancestral South Indian).

    But this poses the problem that a single wave of south Denisovan admixture into Papuan+Onge&ancient Indian cannot really be possible without contributing to East Asians as well, since Onge splits later from most of East Asian ancestry than Papuans.

    Plus the absolute amounts in south Denisovan related ancestry Papuan and in any Onge&ancient Indian would have to be quite different, which again persuades against a single admixture event. (Neither did the Hoabinhians have an enriched Denisovan signal).

    On top of this, the different signals between fairly closely related East Asian ancestry in Japanese/Han/Native American/mainland South East Asian (both more heavily Austroasiatic Kinh and more Northeast Asian Japanese population lack signal, while ST speaking Han have it).

    Putting it all together, it seems like it might be really useful to look at high coverage genomes for heavily ASI populations from South Asia, the Onge and for Tibetans to really understand this.

    The best model that seems to fit that I can guess is for at least 3x discrete “Denisovan” admixture events for South Asians/Onge, Papuans and the High Altitude Adapted population with Denisovan ancestry that contributed known variants to Tibetans, and probably at a lower level to other Sino-Tibetan speaking populations. Probably only the latter event (admixture into the High Altitude Adapted) actually would be closely related to the specific Altai Denisovan population, but different “southern” Denisovan pulses may not be particularly closely related to each other.

    (Final complication, issues of possible selection against genome wide Neanderthal ancestry believed to be identified from ancient dna probably would have to apply to various Denisovan populations…)

  4. The best model that seems to fit that I can guess is for at least 3x discrete “Denisovan” admixture events for South Asians/Onge, Papuans and the High Altitude Adapted population with Denisovan ancestry that contributed known variants to Tibetans, and probably at a lower level to other Sino-Tibetan speaking populations. Probably only the latter event (admixture into the High Altitude Adapted) actually would be closely related to the specific Altai Denisovan population, but different “southern” Denisovan pulses may not be particularly closely related to each other.

    yes. i think so.

    at the very low ancestry % we r getting into tricky power territory….

  5. I begin to wonder after re-reading this study – – which shows a large portion of the DNA amongst Tibetans at a high altitude – is of unknown archaic ancestry. Up to 9%, I might add!! It also clearly says that it is neither of Neanderthal nor Denisovan origin. Yet only the Denisovan DNA is discussed when high altitude adaption is recited. Did Denisovans live at high altitudes? It is also “very complicated”, “complex” and evidently therefore quite delicate. Oh Elusive Yeti – where art thou?

  6. Madsen et al. 2017 (Archaeological Research in Asia 11 (2017) 15–26)

    We identified and dated 18 occupational events at eight sites dating to ~14.7–10.8 kiloannum before present (ka BP) during a decade of archaeological survey and excavation on the northeastern high Tibetan Plateau (TP)> 3200 m. The ephemeral nature of the earliest sites suggests they were created by small foraging groups during very short stays. By ~12 ka BP, larger foraging groups began to occupy sites> 4000 m leading to a more intensive occupation of the high TP after ~9.5 ka BP. This archaeologically-based chronology closely matches genetically-based Tibetan population histories showing an early growth in population size and initial split with Han populations ~15–9 ka BP, and a second spike in population growth during the early-mid Holocene. We found no evidence for occupation of the high TP prior to or during the Last Glacial Maximum (LGM), suggesting the initial separation of Tibetan and Han populations may have occurred at lower elevations in the TP margins or after the LGM in the high TP.

    There is no archaeological evidence of either Denisovans or crypto-hominins at high altitude before the arrival of Han-related people.

  7. @Stuart Fiedel

    Absence of evidence is not evidence of absence.

    The genetic evidence indicates that the effective population size of the Altai Denisovans was very small. And, distinguishing Middle Paleolithic lithic toolkits from random rocks and animal created kill remains is quite subtle. It doesn’t help that we have no real idea what we are looking for because we don’t have a good idea of what physical relics should be associated with Denisovans unlike H. Erectus and Neanderthals.

    And, given political conditions in Tibet, the presence of archaeologists there has been moderately suppressed relative to many other places.

    Denisovans make up ca. 4%-8% of the ancestry of Papuans, yet we have absolutely no evidence of anything that can be reliably associated with Denisovan hominins in SE Asia.

    So, I don’t think that you can read much into a lack of archaeological evidence in Tibet to date. We could be one curious goat herder away from a major break through.

  8. O,

    Middle Paleo toolkits made by Neanderthals are quite sophisticated. Given the Neanderthal/Denisovan divergence date, we should attribute similar mental/technological skills to them. No one would confuse a Neanderthal-made Levallois point with an Oldowan chopper, or either with “random rocks.” Some Russian archaeologists have been (probably mistakenly) ascribing some amazing artifacts from Denisova Cave to the Denisovans, particularly a polished stone bracelet fragment:

    Sometimes, after enough effort has been expended, absence of evidence really is pretty good evidence of absence.


Comments are closed.