A new method reads ancient interbreeding out of the DNA of living people alone, with no archaic genome to compare against. It recovers the known Neanderthal and Denisovan signals, and finds two more — from lineages that have no fossil, no name and no sequenced DNA.
On 30 July a team led from the University of California, Berkeley published a technique called TRACE, which looks for the genetic residue of ancient interbreeding using only the genomes of people alive today — no archaic reference genome, and no population assumed to be free of archaic ancestry.1 Applied to the 1000 Genomes dataset, it first recovered the Neanderthal and Denisovan ancestry already established from fossil DNA, which is the validation step, and then turned up signals that matched neither.3
Those leftover signals resolved into two distinct sources. One is a “ghost” lineage that split from the modern human line around 800,000 years ago and interbred with our ancestors in Africa before the main dispersal out of the continent about 50,000 years ago. The other is a far older “super-archaic” lineage, descended from a population that had already been separate for 1.8 million years, which appears to have mixed with Denisovans in Eurasia rather than with us directly.2
Your genome is the complete set of DNA you inherited, half from each parent, and theirs from their parents, and so on. Because DNA is shuffled and passed down in chunks rather than as a single block, a genome is a patchwork: different stretches of it have different ancestral histories, and some of those histories run much deeper than others.
Archaic hominins are extinct members of the human family — Neanderthals and Denisovans are the two we have bones and sequenced DNA for — who were close enough relatives that they and Homo sapiens could have children together. When that happens between populations long separated, and the offspring’s descendants carry the imported DNA forward, geneticists call it introgression. It is the technical word for interbreeding that left a lasting trace.
A ghost lineage is a population whose existence is inferred only from the DNA of living people. No skeleton has been matched to it and no genome from it has been sequenced; what exists is a pattern in modern genomes that is hard to explain without it.
The percentages need care. Saying Neanderthal ancestry runs at 1–2% in people outside Africa means that share of an individual’s DNA descends from Neanderthal ancestors.4 It is a bookkeeping figure about where segments of a genome came from, not a measure of how much any person resembles a Neanderthal.
Finally, the machinery. For any single stretch of DNA there is a family tree describing how the copies carried by living people are related, and where those branches meet. An ancestral recombination graph stitches millions of those local trees together across the whole genome. TRACE hunts for stretches whose tree reaches unusually far back before the branches join — the signature you would expect if that segment entered the human population from a long-separated relative.1
The study places four separate episodes of interbreeding on one timeline: Neanderthals and Denisovans into non-Africans after the dispersal from Africa, the ghost lineage into the ancestors of everyone, and the super-archaic lineage into Denisovans.2 Reported as a list, they read as four comparable facts. Reported with their uncertainty attached, they do not.
Neanderthal gene flow is the one episode pinned down by physical evidence. Sequencing of ancient Homo sapiens remains dates it to a window beginning about 50,500 years ago and running roughly 7,000 years, averaging around 47,000 years ago.4 The two new episodes carry no such anchor. Each is bounded at one end by when the lineage split off and at the other by the latest point the gene flow could have happened — leaving windows 107 and 229 times wider.
Archaic ancestry is unevenly distributed. Denisovan DNA is concentrated in Oceania, where it can reach about 4% of an individual’s genome, but runs at around 0.1% in East Asia.24 The ghost lineage behaves differently: the team reports similar amounts in African and non-African populations alike, which is what points to gene flow before the out-of-Africa dispersal rather than after it.2
One result cuts against a standing interpretation. Parts of the genome carry no detectable Neanderthal or Denisovan ancestry at all — so-called ancestry deserts, read as regions where archaic DNA was selected out because it did not work in a Homo sapiens body. The authors find ghost ancestry inside those deserts, which suggests the deserts may have been hostile to those two lineages specifically rather than to archaic ancestry in general.1
Everything established about Neanderthal and Denisovan ancestry rests on a physical sample: DNA extracted from bone, sequenced, and matched against segments in living genomes. For these two lineages there is nothing to match against. Only one Denisovan genome has been published at all.2
So the claim being made is narrower than “we found their DNA”. It is that certain segments have genealogies too deep to be explained by modern human ancestry alone, and that the pattern of those segments across populations fits admixture from long-separated relatives. The authors report that simulations give the method high precision and a low false-discovery rate.1
Analysis. That is a real result, and it is worth being clear about what it does not settle. A deep genealogy is direct evidence of deep divergence; admixture is one explanation for deep divergence, and long-standing structure within the ancestral population is another. Simulations test a method against data generated under models the authors specify, which is the right check but not the same as an independent dataset. The technique is a week old at the time of writing and has not yet been applied by another group.
The identities are open too. The inferred divergence dates happen to overlap with Middle Pleistocene Homo groups in Africa around 800,000 years ago and with Homo erectus in Eurasia around 1.8 million years ago.2 Overlapping in date is not the same as being identified, and the paper does not claim otherwise. A preprint version of the work is openly available for readers who want the methods in full.5
Who either lineage was. No fossil has been assigned to them and no DNA has been recovered from them, so the identifications remain open.2 Whether the super-archaic ancestry reached modern humans only through Denisovans, or also by direct interbreeding, is unresolved, as is where that lineage lived and how widespread it was.1 The exact number and timing of ghost introgression events are not fixed by this work; four is the number of episodes the analysis supports, not a ceiling, and the team expects more signals to emerge as genome databases become less skewed towards European and East Asian samples.2 What the introgressed segments actually do — the paper notes an enrichment in regions linked to immunity and metabolism — has not been characterised.1