Denisovans get a face from fossils dug up in China in the 1970s

Skull pieces, teeth and jaws that sat in a Chinese collection for decades belonged to the Denisovans, the extinct cousins of Neanderthals that science first met as a single finger bone. Proteins pulled from two of the bones carry a molecular signature that, among humans studied so far, only Denisovans have. That turns the Xujiayao fossils into one of the largest sets of Denisovan remains anywhere, and it lets researchers say, feature by feature, what these people looked like.

The answer is a long, low head with very thick bone, a heavy brow ridge that sweeps backward, a sloping forehead, squarish eye sockets, a broad nose opening and cheeks that sit flat rather than jutting forward. “We are finally starting to put a face to the Denisovan,” Sally Wasef, a palaeogeneticist at the Queensland University of Technology in Brisbane who was not part of the study, told Nature’s news team. Until recently, most of what anyone knew about this group came from DNA, not anatomy.

The work, led by Qiaomei Fu’s group at the Institute of Vertebrate Paleontology and Paleoanthropology in Beijing, also redraws a messy corner of the human family tree. Several East Asian fossils that experts had filed under different names, including a recently proposed species called Homo juluensis, may all be Denisovans.

How they did it

No DNA survived at Xujiayao, a site in the Nihewan Basin of northern China that yielded 20 human fossils. Proteins last longer than DNA, so the team went after those instead. They CT-scanned two specimens, a piece of the right parietal bone from the side of the skull and part of a lower jaw, then drilled a small hole in each inside a clean room to collect bone powder.

A mass spectrometer, a machine that weighs molecular fragments precisely enough to read their building blocks, then identified the protein pieces. Two separate laboratories, at Capital Medical University and Fudan University, ran the samples, and the results were checked with two different software packages. Proteins that sit in the ground for a very long time slowly change chemically, while modern contamination from skin or lab reagents does not, so the team used that wear to separate genuinely ancient material from stray modern proteins. After that filtering, 18 ancient proteins remained.

The decisive clue sits at one position in a collagen protein called COL1A2. At that spot, every Denisovan sampled so far, from Siberia’s Denisova Cave to the Tibetan Plateau, the Taiwan Strait and the Harbin skull, carries an amino acid called lysine, while Neanderthals and modern humans carry arginine. In the parietal bone, 57 different protein fragments showed lysine there and only one showed arginine. The jawbone gave 61 against one. A family tree built from eight proteins grouped both bones with the Denisovans, with a statistical confidence above 99 percent.

What a Denisovan face looked like

With Xujiayao confirmed, the team set its bones beside the Harbin cranium, a near-complete skull nicknamed Dragon Man that earlier protein and DNA work had already tied to the Denisovans. Together, the two sets of fossils give the first shared checklist of Denisovan anatomy rather than a description built from one specimen.

Some traits look old, like those of Homo erectus: a low, wide braincase and a vault that is 7.7 millimetres thick at the top of the head and 12 millimetres at the bulge of the parietal bone. Others look surprisingly modern, such as a strongly arched temporal bone above the ear. A few match late Neanderthals, including the layout of the inner ear’s balance canals and a small pit at the back of the skull. Then there is a combination the team calls distinctively Denisovan: a low skull with almost parallel side walls, a massive brow ridge, flat cheekbones and only a slight forward thrust of the jaw. The lower jaws from Xujiayao, Xiahe and Penghu also share a small hollow on the inner side of the bone that is rare in other humans.

For comparison, Neanderthals typically show a rounded skull seen from behind and a bulge at the back of the head known as an occipital bun. The Denisovan skulls lack both, so they were not simply Neanderthals living farther east.

Fewer branches on the family tree

The most provocative claim reaches beyond Xujiayao. One of its crania, Xujiayao 6, had been made the type specimen of Homo juluensis, a species defined partly by a huge brain volume of about 1,700 millilitres. The authors argue that size alone cannot carry a species. Brain volumes in their group range from about 1,120 to 1,800 millilitres, a spread that fits within the range seen in living humans, roughly 1,090 to 1,880 millilitres.

Comparing anatomy across the region, the team concludes that famous fossils such as Dali, Jinniushan and Hualongdong, long described as archaic humans with unsettled names, belong to the Denisovan lineage too. If that holds, Denisovans were not a rare group known from scraps but the main archaic human population of East Asia, present from at least 300,000 until about 45,000 years ago, from cold northern China to the subtropical south. “The study represents a major step forward” in understanding what these people looked like and where they lived, Fabrice Demeter, a palaeoanthropologist at the University of Copenhagen, told Nature.

What it doesn’t settle

The molecular evidence covers two bones out of 20, and it rests on a single amino-acid difference. No other variant separates individuals within the Denisovan group, so the proteins cannot say how the Xujiayao people relate to those from Siberia, Tibet, Taiwan or Harbin. The reference set that makes the marker “Denisovan-only” is also small, five individuals compared with Neanderthals and modern humans.

The wider reclassification is weaker than the Xujiayao result itself. Dali, Jinniushan, Hualongdong and Xuchang have no protein data in this study; they are linked to Denisovans by anatomy alone. The authors concede that the Denisovan suite “does not appear simultaneously in individual specimens but manifests in different forms or combinations across various specimens”, which makes the definition flexible and harder to test. Demeter also points out that a group spread over such a vast area and time span probably varied in appearance.

The face itself is partly inference. The facial bones that complete the picture come from Harbin and other skulls, not from Xujiayao. Skin colour, hair and soft tissue cannot be read from bone, so any lifelike portrait remains an artist’s interpretation.

Common questions about the Denisovans

Who were the Denisovans?

They were an extinct group of archaic humans, closely related to Neanderthals, first identified from the DNA of a finger bone found in Denisova Cave in Siberia. They interbred with our species, and people in Oceania and parts of Asia still carry their DNA.

What did Denisovans look like?

Based on the Xujiayao and Harbin fossils, they had long, low skulls with thick bone, a massive brow ridge, a sloping forehead, squarish eye sockets, a broad nose opening and flat cheeks. Their jaws projected only slightly. Skin, hair and eye colour remain unknown.

How can proteins identify a human species?

Proteins are built from amino acids in an order set by genes, so a change in a gene can leave a lasting mark in the protein. Collagen in bone can survive far longer than DNA. A known Denisovan-specific change in the protein COL1A2 works as a molecular fingerprint even when no DNA is left.

Is Homo juluensis a real species?

It was proposed for the large-brained Xujiayao and Xuchang fossils. The new study argues that the size difference is normal variation within one population and that these fossils are Denisovans, though other researchers may still defend the separate name.

What comes next

The test the study invites is direct: if Dali, Jinniushan or Hualongdong are Denisovans, any of them that still holds collagen should show lysine at the same spot in COL1A2. Protein or DNA from more Xujiayao individuals, ideally from different layers of the site, would show whether one population lived there through the roughly 60,000 years its deposits span. The assemblage was dug up between 1976 and 1979 and is dated to between about 199,000 and 138,000 years ago, which places it among the earliest molecularly confirmed Denisovans. The study was published in Nature on 21 October 2026, a month after a separate Nature paper on Denisovan remains from Bianfu Cave in Yunnan.

Reference: Xing et al., “The proteome of Middle Pleistocene Xujiayao and common Denisovan morphology,” Nature, 2026. DOI: 10.1038/s41586-026-11132-z

Tags: archaeology, human evolution, Neanderthals, Denisovans, Harbin, Homo juluensis