Every one of us is carrying a genetic record of relatives no one has ever seen. A new analysis of human genomes has identified traces of two previously unknown archaic hominin lineages — neither one represented by a single bone, tooth, or fossil fragment. They were found the way a growing share of paleogenomics discoveries now are: not by digging, but by computing.
In a study published in Science on July 30, 2026, Yulin Zhang and colleagues at the University of California, Berkeley and Johns Hopkins University describe a new method for detecting archaic ancestry hidden inside the genomes of living people, and used it to identify two lineages that have left no fossil trace at all.
Two ghosts in the genome
The first is a “ghost” lineage that appears to have diverged from the ancestors of modern humans roughly 800,000 years ago, then interbred with Homo sapiens more than 50,000 years ago. Its genetic fingerprint shows up broadly across living populations — the researchers found it is not confined to African populations, as some earlier ghost-lineage hypotheses proposed, but present across modern humans generally, contributing roughly 0.5 to 1% of the average person’s genome.
The second lineage is older and stranger still. It appears to trace back around 1.8 million years, making it substantially more ancient than the divergence of Neanderthals and Denisovans from the modern human line. This “super-archaic” population appears never to have interbred with Homo sapiens directly. Instead, the researchers detected its signal indirectly: it interbred with Denisovans at some point in the past, and that mixed ancestry was then passed down into the Denisovan DNA that later entered human populations — primarily visible today in the genomes of Oceanian populations, where Denisovan ancestry runs highest.
Together, these two previously uncharacterized lineages account for a meaningful share of the roughly 2% of the modern human genome long known to derive from archaic hominins — ancestry that, until now, researchers could attribute to Neanderthals and Denisovans but not fully resolve into distinct sources.
How you find an ancestor with no bones
Fossils and ancient DNA extraction are the traditional tools of paleoanthropology, and both have obvious limits: bones have to survive, and ancient DNA degrades. Most of the deep human past involving small, dispersed populations in tropical and subtropical regions — where DNA preserves poorly — has left almost nothing for either method to work with.
Zhang and colleagues took a different route. They built a computational method called TRACE (TRacking Archaic Contributions via ARG Estimation), which reconstructs the ancestral recombination graph — essentially, the branching genealogical history — underlying hundreds of living human genomes. Because recombination reshuffles chromosomes every generation, different short stretches of anyone’s genome carry independent genealogical histories. Some of those stretches turn out to be far older than they should be if they came only from the known human, Neanderthal, and Denisovan lineages. TRACE flags exactly those anomalously ancient segments and estimates when and from whom they entered the human gene pool — all without a single ancient bone or preserved DNA sample.
An open-data discovery, not a fieldwork discovery
This is the detail that makes the study stand out from most “new ancient human relative” headlines: no excavation, no cave, no newly sequenced fossil specimen sits behind either finding. The entire discovery rests on statistical inference applied to genomic reference datasets that were already public — large, shared collections of modern human genome sequences that many research groups can access, reanalyze, and build on.
That is a reproducibility story as much as a paleogenomics one. A method like TRACE only works, and only produces a result other scientists can check, because the underlying population genomic data is openly available and consistently curated. Nobody needed privileged access to a rare specimen; anyone with the computational method and access to the same reference genomes can, in principle, rerun the analysis. That is precisely the kind of open, reusable-data foundation that research data management standards — shared reference panels, consistent metadata, durable public repositories — exist to make possible. A discovery like this is what “FAIR” data pays off as, in practice: not an abstract compliance requirement, but the reason a genuinely new finding about human origins was even reachable from a desk rather than a dig site.
What it means, and what it doesn’t — yet
Neither lineage has a name, a described skeleton, or a confirmed geographic range in the way Homo erectus or Homo naledi do. What the study establishes is a genetic signature and an estimated age and admixture timing — strong evidence that something distinct existed and interbred (directly or, in the second case, indirectly) with our ancestors, without yet being able to say what that population looked like or where exactly it lived. Confirming and characterizing either lineage further will likely require new ancient DNA finds or fossils that can eventually be cross-checked against the genomic signal — fieldwork and computation converging on the same answer from two directions, rather than one substituting for the other.
For now, the result adds two more branches to an already more complicated human family tree than the textbook Neanderthal-and-Denisovan picture suggested, and it does so as a demonstration of what well-maintained, openly shared genomic data can still reveal, years after it was first collected, to anyone with a good enough method to ask it new questions.
Sources
- Zhang, Y. et al. “Recovering signatures of archaic hominin introgression using ancestral recombination graphs.” Science, July 30, 2026. DOI: 10.1126/science.aef8874
- Sci.News: “Scientists Find Two Previously Unknown Archaic Hominin Lineages Hiding in Human DNA”
- ScienceDaily: “Scientists Found Two Mysterious ‘Ghost’ Ancestors Hiding in Our DNA”
