A third ancestor was hiding in Japanese DNA, breaking the two-origin story

The people of Japan do not descend from two ancestral groups, as the textbooks have it, but from three. A study that read the complete genomes of 3,256 people across the country found a distinct third lineage, strongest in the northeast, that earlier models had folded away. The finding rewrites a population’s origin story and ties strands of that ancient mixing to diseases people carry now.

The old picture was tidy. It held that modern Japanese ancestry came from the Jomon, the hunter-gatherers who lived on the islands for thousands of years, and from later migrants out of continental East Asia who brought rice farming. Two sources, cleanly drawn.

The genomes do not fit two boxes. The Jomon signal is still there and runs strongest in Okinawa, where it accounts for roughly a third of local ancestry. The continental signal dominates in western Japan and carries clear links to Han Chinese populations. But a third component sits apart in the northeast, where the historical Emishi lived, and it does not reduce to either of the others.

Reading whole genomes rather than scattered markers is what made the difference. The team sequenced every letter of the 3,256 genomes drawn from a national medical biobank and built them into a database meant to map both ancestry and health. At that resolution the third strand stops looking like noise and starts looking like a separate population.

The deep past also turns out to be medical. The analysis flagged 44 stretches of archaic DNA inherited from Neanderthals and Denisovans. A Denisovan-derived segment near the gene NKX6-1 is associated with type 2 diabetes and may even shape how patients respond to the diabetes drug semaglutide, while eleven Neanderthal-derived stretches track with coronary artery disease, prostate cancer and rheumatoid arthritis.

The three-way split is a strong statistical inference, not a pedigree. Ancestry proportions are reconstructed from patterns across many genomes, and the northeastern strand is described as Emishi-related rather than proven to be the Emishi themselves. The samples also come from a medical biobank, which skews toward the kind of people who enroll in one, so the exact balance of the three sources should be read as a best estimate rather than a final count.

The work, led by Chikashi Terao at RIKEN’s Center for Integrative Medical Sciences, appeared in Science Advances in 2024 and drew fresh attention this month as the genome database behind it, known as JEWEL, expanded. The same approach is now being turned on other populations long described in two-part terms, on the expectation that more of those tidy stories will turn out to have a third thread running through them.