The Hidden Center of the Human Genome: Japan’s Scientists Make 3,332 Genomes Speak

A blind spot in the genetic map becomes searchable

A human genome can be sequenced, stored and compared with remarkable speed. Yet some of its most consequential stretches have remained difficult to examine across large numbers of people. At the center of that problem is the centromere, a chromosome region essential to the machinery that distributes genetic material as cells divide.

An international team led by researchers at Japan’s National Cancer Center, working with Kyoto University, Nagoya University and other institutions, has devised a way to infer centromere types from short-read whole-genome sequences already collected. The approach, implemented in software called ascairn, was announced in Japan on October 14 and published in Cell on October 13 U.S. time.

Why the genome’s center was hard to read

Centromeres support the assembly of the kinetochore, the structure through which the cell’s spindle machinery attaches to chromosomes during division. Their role is critical, but much of their DNA consists of extensive arrays of highly similar repeats.

Conventional short-read sequencing works by reading many small fragments and mapping them against a reference. In a repetitive region, countless fragments look alike. Even a technically excellent dataset may be ambiguous about exactly where a short piece belongs. Researchers have therefore had a peculiar imbalance: enormous libraries of genomes, but weak population-scale visibility into some of their most structurally intricate parts.

The trick: distinctive fragments inside repetitive DNA

The Japanese-led team sought small pieces of sequence that are unusual enough to distinguish one centromere lineage from another. Such sequence markers are known as rare k-mers: short stretches of k DNA bases whose patterns can be informative even when the surrounding region is highly repetitive.

Using high-quality reference centromere sequences, the researchers constructed a reference panel and trained the ascairn method to infer centromere haplogroups from existing short-read whole-genome sequencing data. It is a classification strategy—not a claim to reconstruct every repeat in every person’s centromere from short reads alone.

What 3,332 genomes reveal

The team analyzed genome datasets from 3,332 people across different parts of the world and found extensive diversity among centromere haplogroups, including patterns of geographical and population structure.

The result extends population genetics into regions typically underrepresented in large studies. It does not support simplistic biological categories or claims about the abilities of individuals. Rather, it indicates that human evolutionary history has left a record in difficult-to-sequence chromosome structures as well as more familiar parts of the genome.

A finding in a particular kind of brain tumor

The most immediately consequential disease-related analysis concerns oligodendroglioma, a type of brain tumor in which loss of chromosomal arms 1p and 19q together—known as 1p/19q co-deletion—is a defining molecular feature.

The Cell paper reports analysis of 142 co-deleted oligodendroglioma genomes. Researchers mapped breakpoints in relation to centromere regions on chromosomes 1 and 19 and identified an association between centromere haplotype structure and susceptibility to these rearrangements.

An association in tumor data is not proof that a given centromere type causes cancer. Nor does this constitute a new screening test, personalized risk score or treatment. Understanding why certain chromosome rearrangements occur will require additional mechanistic and clinical studies.

From the Human Genome Project to an era of complete references

The achievement makes more sense against the history of the genome itself. The Human Genome Project, launched in 1990, reached its landmark completion in 2003. Its sequence covered roughly 92% of the genome, according to the U.S. National Human Genome Research Institute. Highly repetitive regions, including centromeres, remained formidable gaps.

In 2022, the Telomere-to-Telomere consortium published the first gapless sequence of a human genome, overcoming many of those longstanding technical barriers. That was a triumph of long-read technology, assembly algorithms and international collaboration. But producing a highly detailed reference genome is not the same as cheaply characterizing a population of thousands or millions.

The new work helps bridge those achievements: knowledge gained from high-quality long-read references can now illuminate the huge reservoir of existing short-read datasets.

Why reusing old sequencing data matters

Short-read sequencing has generated extensive genomic archives because of its relatively accessible cost and established laboratory workflows. Re-sequencing every sample with long-read technologies would be expensive, impractical or impossible when original biological material is unavailable.

A method that extracts new biological information from those datasets can change the economics of discovery. It may allow researchers to revisit cohorts assembled years ago with questions those studies could not originally address. Nevertheless, inferred haplogroups are not full centromere assemblies; independent validation remains important when precise structural claims are at stake.

The scientists behind the software

The paper’s corresponding author is Yuichi Shiraishi, head of the Division of Genome Analysis Platform Development at Japan’s National Cancer Center Research Institute. Kyoto University collaborators include Seishi Ogawa, Yotaro Ochi, Taro Tsujimura and Takuya Yamamoto. Ryuta Saito of Nagoya University and numerous international collaborators also contributed.

That institutional breadth matters. The advance brings together computational genomics, basic chromosome biology and cancer research. It is better described as an international collaboration with substantial Japanese leadership than as a Kyoto-only invention.

Important unanswered questions

How accurately will ascairn distinguish rarer or previously unseen centromere lineages in more diverse cohorts? Will its associations with tumor rearrangements replicate in independent datasets? Could researchers move beyond statistical association to a direct cellular mechanism?

These are open scientific questions. Large genomic studies also raise governance questions about consent, privacy, equitable representation and appropriate interpretations of genetic ancestry. No one should confuse population-level patterns with a definitive description of an individual.

Discovery without another sequencing run

The most surprising aspect of this discovery is that its raw material was already there. Many of the genetic datasets needed to examine centromere variation had been collected for other purposes. What was missing was a reliable way to ask the right question of them.

For genomic medicine, the immediate gain is analytical access, not clinical treatment. For biology, it is a chance to investigate how chromosome structures differ and how those differences may influence genomic instability. Sometimes the next frontier in science lies not in gathering more data, but in learning to read the data humanity already possesses.