For fifteen years Denisovans were a population with genomes and almost no anatomy. Everything known about their bodies came from a finger bone, a few molars, a partial jaw from the Tibetan Plateau and a molar from Laos. Two studies published in 2025 changed that by identifying a nearly complete cranium — the Harbin skull from northeastern China, more than 146,000 years old — as belonging to a Denisovan.
Neither study found a new fossil. Both applied molecular methods to one that had been described in 2021 as a new species, Homo longi. One recovered ancient proteins from the bone; the other recovered mitochondrial DNA from the dental calculus on a tooth. Both pointed the same way.
The evidence comes from The proteome of the late Middle Pleistocene Harbin individual in Science, which retrieved 95 endogenous proteins, and its companion Denisovan mitochondrial DNA from dental calculus of the >146,000-year-old Harbin cranium in Cell.
The short answer: the Harbin cranium carries three Denisovan-derived amino-acid variants and clusters with Denisova 3 in protein-based analysis, while mitochondrial DNA from its dental calculus falls within Denisovan variation and sits near an early Denisovan branch known from Denisova Cave in Siberia. Together the results give the group a well-preserved skull and extend its known range across Asia. Molecular identification places a fossil in a lineage; it does not settle how species should be named.
Why the skull had no name that stuck#
The Harbin cranium has an unusual history. It was reportedly recovered in the 1930s near a bridge in Harbin, Heilongjiang province, hidden for decades, and only handed to scientists much later. It is remarkably complete for a Middle Pleistocene hominin — a braincase, a face, a heavy brow — and it does not fit neatly into the categories the region's fossils are usually sorted into.
In 2021 a team described it as the type specimen of a new species, Homo longi, meaning "dragon man" after the Heilongjiang region. Others suspected it might be Denisovan, on the reasoning that a large-brained Middle Pleistocene East Asian hominin was exactly what a Denisovan skull ought to look like. Nobody could test the idea, because the group was defined genetically and the skull had yielded no DNA.
That gap is the whole point of the 2025 work. It is an attempt to attach a molecular identity to an anatomy, in a case where both had been known for years and could not be connected.
What ancient proteins can and cannot do#
DNA is fragile. In warm or temperate conditions it degrades beyond recovery within tens of thousands of years, which is why ancient genomes cluster in cold, dry and cave environments — and why the first whole genome from Old Kingdom Egypt was such an outlier.
Proteins are tougher. Sequences of amino acids survive in bone and enamel far longer than DNA does, and because the genetic code specifies them, a protein sequence carries a lossy copy of genetic information. Where a species-diagnostic position differs between lineages, a preserved protein can report which variant an individual had.
The trade-off is resolution:
| Ancient DNA | Ancient proteins | |
|---|---|---|
| Survival | Tens of thousands of years in favourable conditions | Hundreds of thousands of years, more widely |
| Information | Whole genome, millions of variable sites | A few dozen proteins, a handful of diagnostic positions |
| Best use | Population history, admixture, kinship | Assigning a fossil to a lineage |
| Typical result | "This individual descends from these populations in these proportions" | "This individual belongs to this clade" |
The Harbin proteome delivered exactly what that second column promises: 95 endogenous proteins, three Denisovan-derived amino-acid variants, and a clustering with Denisova 3 — the individual from Denisova Cave that defines the group.
The DNA came from plaque#
The Cell study is the more surprising of the pair, because of where the DNA was found. Attempts on a tooth and on the petrous bone — the dense inner-ear bone that is the usual first choice for ancient DNA — both failed. Mitochondrial DNA was recovered instead from dental calculus: mineralised plaque, hardened onto the tooth surface during life.
Calculus is normally studied for what it traps from outside the body, such as food particles and oral microbes. Here it acted as a container for the host's own DNA, sealing a small quantity of it away from the degradation that had destroyed the rest.
The recovered mitochondrial sequence falls within Denisovan variation and is related to an early Denisovan mtDNA branch previously seen at Denisova Cave in southern Siberia. That is a specific and useful result: it places Harbin not just inside the group but near a particular part of its early diversity.
The methodological implication is broader than the fossil. If host DNA survives in calculus where it does not survive in bone, then teeth that have already been written off as sterile may be worth revisiting — including in regions where preservation has been the field's binding constraint.
Mitochondrial DNA is one lineage, not a whole ancestry#
One caveat deserves emphasis, because it recurs throughout ancient DNA. Mitochondrial DNA is inherited only through the maternal line. It traces a single thread back through an individual's mother, her mother, and so on, and says nothing about the rest of the family tree.
An individual can carry Denisovan mtDNA and still have had substantial ancestry from another group — the pattern is well documented among Neanderthals, whose mitochondrial and nuclear histories do not always match. Neither does mtDNA on its own establish nuclear admixture in either direction.
That is precisely why the protein evidence matters alongside it. The proteome samples nuclear-encoded loci, so the two lines of evidence come from different parts of the genome and are not simply a single result reported twice. Their agreement is what makes the identification convincing, rather than either study alone.
What the identification changes#
A face for the group. Denisovans now have a nearly complete skull. Any reconstruction of their anatomy previously rested on a jaw, some teeth and inferences from genome-wide methylation patterns; there is now an actual cranium to reason from.
A larger range. Confirmed Denisovan remains had come from southern Siberia, the Tibetan Plateau and Laos. Harbin extends the securely identified distribution into northeastern China, consistent with a group occupying an enormous span of Asia and a wide range of environments.
A rethink of Chinese Middle Pleistocene fossils. Several other Chinese specimens have long resisted classification, sitting awkwardly between Homo erectus and later forms. With one of them anchored molecularly, the others can be compared to a known Denisovan rather than to a hypothesis.
Naming remains unsettled. Whether the group should be called Homo longi, Homo denisova, or nothing at all is a taxonomic argument that molecular data cannot resolve. What the studies establish is which lineage the individual belonged to; what to call that lineage is a separate question about how species are defined.
Limitations#
| Limitation | Why it matters |
|---|---|
| The recovery context is poor | The cranium was not excavated by scientists; its find spot rests on reported testimony. |
| The date is a minimum | "More than 146,000 years" is a floor, established indirectly rather than from a secure layer. |
| Proteins carry few diagnostic sites | Three variants is a strong signal, not a genome. |
| mtDNA is a single lineage | It cannot describe the individual's overall ancestry. |
| One individual is not a population | Harbin shows what one Denisovan looked like, not the range of the group. |
| Taxonomy is unresolved | Lineage assignment does not decide what the species should be called. |
Frequently asked questions about the Harbin skull#
Is the Harbin skull definitely a Denisovan?#
Two independent molecular lines — nuclear-encoded proteins and mitochondrial DNA — both place it within the Denisovan clade. That is strong evidence. It rests on a small number of diagnostic positions rather than a genome, which is the honest limit of the claim.
How old is it?#
At least 146,000 years. The figure is a minimum age, derived indirectly because the specimen lacks a documented excavation context.
What happened to the species name Homo longi?#
It still exists in the literature. If Harbin is a Denisovan and the name has priority, some researchers argue Denisovans should be called Homo longi; others prefer to keep using "Denisovan" informally for a genetically defined group. The molecular evidence does not decide the naming convention.
Why did the tooth and the petrous bone fail when calculus worked?#
Preservation is local and unpredictable. Mineralised plaque appears to have shielded a small amount of DNA from the degradation that destroyed it elsewhere in the specimen. It is a promising route for other poorly preserved fossils, not a guaranteed one.
Does this tell us what Denisovans looked like?#
It gives a well-preserved cranium to work from, which is far more than existed before. Soft tissue, skin, hair and build are not recoverable from a skull, and any illustrated reconstruction of a face is an interpretation rather than a measurement.
Does it change how much Denisovan DNA living people carry?#
No. Those estimates come from comparisons between modern genomes and sequenced archaic genomes, and Harbin adds no nuclear genome. Its contribution is anatomical and geographic. The 200,000-year-old Denisovan genome is the study that changes how that ancestry is resolved.
Molecules where DNA runs out#
The Harbin result belongs to a pattern that is reshaping palaeoanthropology. Ancient DNA transformed the study of the last hundred thousand years, and then hit a wall: in most of the world, in most conditions, it simply does not survive far beyond that.
Proteins push past that wall. They carry less information, but they carry it further, and for the specific job of asking which lineage a fossil belongs to, less is often enough. A Denisovan mandible on the Tibetan Plateau, a Denisovan molar in Laos and now a Denisovan cranium in northeastern China were all identified this way.
What that leaves is a group known across a continent and still without a nuclear genome from anywhere but Siberia. The molecular map of the Denisovans is expanding faster than the genomic one — which is a reasonable description of where the field stands, and of why a well-preserved skull from a river in Heilongjiang mattered enough for two journals to publish on the same day.
Sources and further reading#
- Fu, Q., Bai, F., Rao, H. et al. (2025). The proteome of the late Middle Pleistocene Harbin individual. Science 389, 704–707. DOI: 10.1126/science.adu9677.
- Fu, Q., Cao, P., Dai, Q. et al. (2025). Denisovan mitochondrial DNA from dental calculus of the >146,000-year-old Harbin cranium. Cell 188, 3919–3926.e9. DOI: 10.1016/j.cell.2025.05.040.
- Ji, Q., Wu, W., Ji, Y. et al. (2021). Late Middle Pleistocene Harbin cranium represents a new Homo species. The Innovation 2, 100132 — the original species description.
Editorial note: this article was written as a source-based synthesis and distinguishes lineage assignment from species naming throughout. Its hero and section artwork was generated with AI as an interpretive landscape reconstruction, not as scientific evidence or a facial reconstruction.
