The Ghosts in Your Genome

A Signal That Matched No One

Geneticists have known for two decades that modern humans carry DNA from extinct relatives. When Homo sapiens left Africa roughly 50,000 years ago, they met Neanderthals and Denisovans in Eurasia and had children with both. Most people outside sub-Saharan Africa carry 1 to 2.4 percent Neanderthal DNA today. People across Asia and Oceania carry anywhere from a trace to 6 percent Denisovan DNA. Both figures came from the same source: fossils. Scientists extracted and sequenced actual Neanderthal and Denisovan genomes from ancient bones, then combed living genomes for a match.

But when researchers went looking, they kept finding genetic material that matched neither template. It was old. It was clearly inherited from something that had split off the human lineage a very long time ago. And it corresponded to no fossil anyone had ever sequenced. Multiple research groups over the years had flagged pieces of this pattern and given it a placeholder name: ghost ancestry. Nobody could say whether it showed up only in African populations or in everyone. Nobody could say when the interbreeding happened. Nobody, in the most basic sense, could say who it came from, because there was no bone to check it against.

That is the shape of the problem this team set out to solve. Not “did other extinct humans exist,” which was already assumed, but something narrower and harder: could you prove a specific extinct population contributed to your DNA, name roughly when it happened, and map exactly where in the genome it lives, using nothing but the DNA of people alive right now?

The Method That Needed Bones It Didn't Have

Every prior technique for identifying archaic ancestry worked the same way: get a sequenced genome from the extinct population, then search living genomes for matching stretches. It is a template-matching problem, and it only works if you have the template. That single requirement quietly determined which parts of human prehistory scientists could even ask questions about.

Ancient DNA survives best in cold, dry places. Siberian permafrost. Alpine caves. That is not a coincidence of what got studied; it's a limit on what could be studied at all. Africa, the tropics, and the deepest stretches of time are exactly where DNA degrades fastest and disappears soonest. The result is a strange asymmetry that the field has lived with for years: some of the best-documented ancient interbreeding comes from Eurasia, while Africa, the continent where our species actually originated and spent most of its history, has been comparatively silent. Not because less happened there. Because the physical evidence rotted.

So the ghost ancestry signal sat there for years, real and repeatedly observed, but scientifically stuck. You cannot publish a discovery of a population you cannot describe. You need a when, a where, and ideally a candidate identity. Without a fossil, none of those seemed reachable, until the Berkeley team stopped trying to match a template and started reading family trees instead.

How You Find a Ghost With No Body

The team, led by Priya Moorjani with computational biologist Yulin Zhang and Johns Hopkins postdoctoral researcher Arjun Biddanda as co-first authors, built a method called TRACE: TRacking Archaic Contributions via ARG Estimation. The idea starts from a simple fact about how genomes actually work. Your DNA is not one continuous inheritance from two parents. It's stitched together from thousands of separate fragments, each with its own history, because chromosomes shuffle and recombine every generation. Any two people share a common ancestor for most of their genome within the last few thousand years. But some fragments have a much older story. Their most recent shared ancestor with everyone else's DNA lived a hundred thousand, five hundred thousand, or more than a million years ago.

TRACE builds what geneticists call an ancestral recombination graph, a reconstruction of exactly how every fragment of DNA across the genome traces back through generations of ancestors, branching and merging as populations mixed. The team ran this reconstruction across more than 500 complete genomes from people around the world. Wherever a fragment's reconstructed ancestry reached back further than known human, Neanderthal, or Denisovan history could explain, that fragment became a candidate: DNA from something else.

As a check, the method first had to pass a test it wasn't designed to fail: could it find what scientists already knew was there? It did. TRACE correctly flagged Neanderthal-derived DNA across the genome and correctly identified Denisovan ancestry in Asian and Oceanian genomes, matching decades of fossil-based research without ever touching a fossil. Once the team trusted the method, they pointed it at the unexplained fragments and found two distinct populations hiding in the data.

The first population, which the team calls the ghost lineage, left its DNA in every living human population, African and non-African alike, at roughly 0.5 to 1 percent of the genome, close to the share contributed by Neanderthals. That even spread across Africans and non-Africans is itself the key clue: it means this lineage interbred with the ancestors of all modern humans while everyone was still in Africa, more than 50,000 years ago, before the migration that later met Neanderthals and Denisovans in Eurasia. The lineage itself appears to have split from the ancestors of modern humans around 800,000 years ago, roughly the same moment Neanderthals and Denisovans split from each other.

The second population is older and stranger. Its DNA shows up almost exclusively in people from Oceania, and only inside stretches of the genome that are already identifiably Denisovan, an average of about 0.002 percent of Oceanian genomes overall. That pattern points to a specific chain of events: this “super-archaic” lineage, which split off roughly 1.8 million years ago, before the common ancestor of modern humans, Neanderthals, and Denisovans even existed, interbred with Denisovans first. Denisovan genomes themselves carry an estimated 3 to 5 percent super-archaic ancestry. Only the fraction of Denisovan DNA that later entered modern humans carried a sliver of that even older inheritance along for the ride, a genetic legacy passed through one extinct population into another, and only then into us.

Who Were They, and Who Else Is Still Missing?

TRACE can tell you that a population existed, roughly when it split off, and roughly when it interbred. It cannot, on its own, tell you its name. For that, the team had to reach for the existing fossil record and ask which known hominins fit the profile.

For the ghost lineage, the timing points toward Homo heidelbergensis, a species known from fossils across Africa and Europe that persisted in Africa until roughly 300,000 years ago, comfortably late enough to have been there for an interbreeding event more than 50,000 years ago. Chris Stringer, a paleoanthropologist at the Natural History Museum in London who had no role in the study, told Live Science he found the identification plausible.

The super-archaic lineage points somewhere even more remarkable: Homo erectus, the longest-surviving human species on record, which persisted in parts of Asia for close to two million years. Fernando Villanea, a population geneticist at the University of Colorado Boulder who was not involved in the research, noted that H. erectus skulls recovered at Yunxian in China share features with Denisovan fossils, a physical echo of the same genetic connection TRACE found computationally. Neither identification is confirmed. Both are the best current explanation for a signal that, a year ago, had no explanation at all.

What the finding actually overturns is not a single fact but a shape: the mental image of human evolution as a tree with clean, separate branches, one leading to us and others dying out in isolation. The real picture, four independent lines of evidence now agree, looked more like a wide, tangled network of related populations across Africa and Eurasia, repeatedly meeting, interbreeding, and separating again over more than a million years.

There's a practical postscript worth sitting with too. The archaic DNA fragments TRACE found are not scattered randomly through the genome. They cluster disproportionately in regions tied to immune defense and metabolism, exactly the kind of traits where borrowed genes from a population that had already adapted to a place or a pathogen would be worth keeping. Interbreeding, on this reading, wasn't just something that happened to early humans. It may have been part of how they survived moving into new environments at all.

Moorjani's team is already turning TRACE toward the parts of the genetic record that remain thinnest: genomes from Africa and South Asia, regions still underrepresented in the reference datasets that make discoveries like this possible in the first place. If the ghost and super-archaic lineages were findable once someone stopped waiting for a fossil, there is no obvious reason they're the last two. The method that found them works on any species with enough sequenced genomes, which means the next ghost in the data might not even be human.