A ghost lineage human DNA signal appears to run through every modern population studied, according to a Berkeley-led research team reporting in Science. The finding points to interbreeding between early modern humans and an unknown African relative before the migration that carried some human groups out of Africa.
The group was not Neanderthal or Denisovan. Researchers have genomes from those archaic humans, which makes their surviving DNA easier to spot in living people. This newly detected contributor remains a “ghost” because no matching ancient genome has been recovered, so the team can infer its existence from patterns in modern DNA rather than from a fossil with usable sequence data.
What is a ghost lineage in human DNA?
A ghost lineage is a population whose genetic traces can be detected in descendants even though scientists do not yet have its own genome. In this case, the Berkeley-led team says the signal came from a lineage that split from modern humans more than 800,000 years ago.
The researchers used a software tool called TRACE, built to search for DNA segments that have an odd combination of age markers. Segments inherited from a separate lineage should look old because their ancestry branches away from modern humans far back in time. They should also show fewer signs of recombination than expected for DNA that has stayed in the modern human line the whole time, because the segments were added later through interbreeding.
TRACE relies on ancestral recombination graphs, a computational way to estimate how each point in a genome relates to others across generations. Those estimates are noisy at individual positions, because mutation, recombination and missing pieces make a mess of any neat family tree. Across billions of bases, the team argues, the pattern becomes useful.
The researchers first tested the method against known Neanderthal and Denisovan ancestry. Using the better-performing of two graph-building tools, they reported a false discovery rate below 0.25 percent and accuracy above 90 percent. In African populations, where Neanderthal and Denisovan ancestry mainly arrived through later back-migration from Eurasia, TRACE found about 0.1 percent ancestry from those archaic groups, which the team says fits the tool’s low error rate.
Applied to roughly 500 modern human genomes, TRACE found the expected Neanderthal and Denisovan segments, plus the unknown African signal. The ghost lineage accounted for about 0.5 to 1.1 percent of present-day human genomes. Combined across people, the detected segments covered nearly 1.5 billion DNA bases, compared with about three billion bases in a full human genome.
The signal showed up in all modern human populations examined, which is why the researchers place the interbreeding before the out-of-Africa expansion. African populations carried more varied ghost-lineage segments, including some not seen elsewhere. The team also identified about 100 genome regions where non-African populations lacked this ghost ancestry entirely.
The segments are shorter on average than known Neanderthal and Denisovan segments in modern genomes. That fits an older arrival into the modern human population, since recombination chops inherited blocks into smaller pieces over time.
The study also complicates an older idea about “deserts” in the genome, regions where Neanderthal and Denisovan DNA is absent. Because ghost-lineage DNA appears in some of those same regions, the researchers suggest the absence may reflect problems specific to Neanderthal and Denisovan variants rather than a blanket intolerance for archaic DNA there.
The team also searched for still older ancestry that earlier work had suggested entered Denisovans from a “super archaic” population, possibly Homo erectus or another ancient human relative. In people from Oceania, where Denisovan ancestry is relatively high, the researchers traced about 0.3 percent of Denisovan-derived DNA to that older lineage. They estimate it branched away from modern humans nearly 1.8 million years ago.
The paper does not settle what these ancient segments do. The researchers report that they are less common near genes, while some nearby genes are linked to metabolism and immune function. Whether any of the variants helped modern humans remains unresolved, because a statistical signal in a genome is not the same thing as proof of adaptation.
This story draws on original reporting from Ars Technica.
