The Denisovans who later interbred with modern human populations were carrying an older layer of mixed ancestry themselves, according to a new reconstruction of ancient genomes. UC Berkeley researchers estimate that 3 to 5 percent of Denisovan DNA came from an unidentified “super-archaic” population whose lineage diverged from the modern human lineage about 1.8 million years ago.
Some of that older DNA then traveled inside Denisovan-derived chromosome segments into the ancestors of living people. The conclusion comes from a statistical analysis of genome family trees, not from a newly sequenced fossil belonging to the unidentified donor.
This is one study, not settled consensus.
The DNA took a two-stage route
The proposed route involves two distinct episodes of interbreeding separated by a long interval. First, descendants of the deeply diverged population mixed with Denisovans in Eurasia, probably more than 200,000 years ago. Later, Denisovans interbred with modern human populations. Those encounters passed a small fraction of the older contribution onward for a second time.
Calling the Denisovans “already hybrids” captures that nested inheritance. It should not be taken to mean that one cleanly bounded species suddenly became a fixed mixture. Ancient human groups divided, moved and exchanged genes repeatedly, and the labels used for them compress that population history.
The UC Berkeley account of the work is explicit that the 3 to 5 percent estimate describes the super-archaic share of Denisovan DNA. It does not mean that 3 to 5 percent of any living person’s entire genome comes from this source.
Living people received only a fraction of that fraction.
Denisovan ancestry itself represents a minority of present-day genomes, with the largest known shares in some Oceanian populations. The still older component nested inside it is smaller again.
TRACE inferred ancestry without a donor genome
Yulin Zhang, Arjun Biddanda and colleagues published the result in a peer-reviewed paper in Science. Their method is called TRACE, short for Tracking Archaic Contributions via Ancestral Recombination Graph Estimation.
A chromosome does not preserve one family tree from end to end. Recombination reshuffles DNA in each generation, so neighboring stretches can have different genealogies. An ancestral recombination graph represents that changing web of local trees across a genome.
TRACE searches those reconstructed trees for branches deep enough to suggest a population separated from the sampled ancestors for a long period, attached to segments consistent with later gene flow. It does not require a sequenced ancient genome from the donor or an external comparison population.
The team first tested the approach in simulated histories and asked whether it could recover known Neanderthal and Denisovan ancestry. The main data then included 503 phased whole-genome sequences from British, Han Chinese, Indian Telugu, Yoruba and Luhya participants in the 1000 Genomes Project.
For the super-archaic question, the researchers separately analyzed 92 high-coverage Oceanian genomes. These were especially informative because Oceanian populations retain more Denisovan ancestry than most other living groups, giving the method more inherited Denisovan sequence in which to look for an older layer.
The signal was nested inside Denisovan regions
The deepest unidentified lineages in the Oceanian analysis were enriched within chromosome regions already classified as Denisovan-derived. The team did not see comparable enrichment within Neanderthal-derived regions. That asymmetry is why the authors favor a route through Denisovans rather than direct transfer from the super-archaic population to modern humans.
The study did not observe either encounter. It inferred the sequence from the location and genealogical depth of surviving DNA segments. Other demographic histories can sometimes generate similar patterns, which is why simulation choices and validation against known ancestry matter.
ScienceBlog has separately examined the broader TRACE finding of two unidentified lineages. The other signal in that paper is different: a “ghost” population that appears to have mixed directly with modern human ancestors in Africa and contributed roughly 0.5 to 1 percent of the genomes analyzed.
The super-archaic route is one step more remote. It entered Denisovans first, then reached only modern populations that later inherited Denisovan ancestry.
The two dates describe different events
The estimate of roughly 1.8 million years refers to divergence, the point at which the super-archaic lineage separated genetically from the branch leading toward modern humans. It is not the date when its descendants interbred with Denisovans.
UC Berkeley’s summary places that later contact more than 200,000 years ago. Denisovan gene flow into modern humans occurred later again. The headline’s “hundreds of thousands of years” therefore describes the span across a chain of population splits and contacts, not one continuous interbreeding event.
These timings are model estimates. Mutation rates, assumed generation lengths, population sizes and errors in reconstructed genealogies can move them. The paper’s full methods manuscript details the simulations and filters used to limit false discoveries, but those safeguards do not make one demographic explanation inevitable.
There was already evidence for deep mixing. A 2020 study in Science Advances inferred that the common ancestors of Neanderthals and Denisovans interbred with a very old Eurasian population. That model estimated a super-archaic separation closer to 2 million years ago and placed the contact earlier than the new Berkeley account does.
The two studies use different models and do not draw an identical family tree. Their agreement is broader: the ancestry of Neanderthals and Denisovans may itself include DNA from populations that had been separate for an exceptionally long time.
The donor still has no secure species name
A divergence date near 1.8 million years overlaps with early populations often grouped under Homo erectus. That makes an H. erectus-related source a plausible hypothesis, not an identification. The study has not genetically matched the segments to a named fossil population.
The distinction matters because a genetic branch and a fossil species are not interchangeable. Similar-looking fossils can belong to populations with different histories, while related populations may leave bones that vary enough to receive different names. DNA rarely survives across the full timescale involved.
Additional high-quality Denisovan genomes would test whether the super-archaic signal is consistently nested inside Denisovan ancestry. Ancient proteins recovered from older Asian fossils may eventually offer comparisons when usable DNA is absent.
For now, the evidence supports a two-step genetic route more strongly than it identifies the people at its beginning. The number of contacts, their locations and the fossil population behind the oldest contribution remain open.