Connect with us

NEWS

The Blood Test That Sees Breast Cancer’s Hidden Cells

A Weizmann breast cancer blood test read nucleosome tags, found early tumors, and tracked relapse even in patients called low risk.

Published

on

A blood test that reads chemical tags on DNA spools in plasma separated early breast cancer from healthy samples with an AUC of 0.95. The findings, from labs at the Weizmann Institute of Science and the Hebrew University of Jerusalem, appear in Nature dated Nov. 11, 2026.

The same H3K4 tags that marked a rare aggressive cell state inside otherwise milder tumors were higher in the blood of women whose cancer came back within five years. That is the part of the study that would change follow-up, if larger trials hold.

The Classifier Hit 0.95 on Tumors Still Confined to the Breast

Prof. Efrat Shema’s Weizmann lab and Prof. Guy Ron’s physics group at the Hebrew University built the classifier on EPINUC, a single-molecule read of nucleosomes shed into blood when cells die. The Nature paper on epigenetic heterogeneity in breast tumours used a 29-parameter panel on less than 1 ml of plasma.

The first blood set came from 77 women with luminal, hormone-receptor-positive breast cancer who had been followed for at least five years at the Mayo Clinic in Jacksonville, Florida, plus 30 healthy women. Among the patients were 42 with T1 tumors, the early group Weizmann described as still confined to the breast and no more than two centimeters across.

THE BLOOD CLASSIFIER IN NATURE

Comparison Samples AUC (95% CI)
All Mayo luminal vs healthy 77 vs 30 0.95 (0.91 to 0.98)
Stage T1 only vs healthy 42 vs 30 0.95 (0.90 to 0.99)
Stage T2 to T4 vs healthy 30 vs 30 0.96 (0.91 to 1.00)
Original 20 EPINUC marks same Mayo set 0.85
Relapse vs no relapse Mayo follow-up 0.81
Second relapse group 30 women 0.92

An AUC of 1 is a perfect test and 0.5 is a coin toss, Ron wrote. The jump from 0.85 to 0.95 came when the team added breast-related proteins and histone marks chosen from the tumor work, including H4K16ac and H4K20me3. Shema said performance held when the blood set grew to 188 samples with Sheba Medical Center and Kaplan Medical Center. A 25 percent test slice of that multi-center set (13 healthy, 34 patient samples) was scored at a specificity of 0.9.

The Weizmann Institute posted the announcement when the paper went live.

Genetic Traces Are Too Scarce Early On

Liquid biopsies already help some patients by hunting fragments of tumor DNA. Early disease is the hard case, because mutated DNA in blood is scarce and easy to confuse with noise. Nucleosomes are different cargo. They are short stretches of DNA still wrapped around histone proteins, and they keep the chemical tags that tell a cell which genes to use.

Shema’s group had already shown that those tags can be counted one molecule at a time. A 2022 earlier nucleosome assay in colorectal cancer ran on 63 colorectal cases, 10 pancreatic cases and 33 healthy samples, and Shema said the model reached 92 percent precision from less than 1 ml of plasma. The new breast panel kept that small-volume design and added marks tied to breast identity, including cytokeratin-5, L1CAM and MUC1.

Cancer plasma in the Mayo set carried more H3K9ac and H4K16ac nucleosomes after the team normalized for how many nucleosomes were present, so the shift was in state, not just tumor bulk. Bivalent nucleosomes marked with both H3K4me3 and H3K27me3 also rose, a pattern the 2022 work had seen in colorectal and pancreatic disease.

A Nuclear Physicist Took the Tumor Data Home

Ron is a nuclear physicist at the Racah Institute of Physics. He has never taken a biology class, including in high school. He and Shema had both been Adams Fellows, and in 2021 she asked him to look at noisy CyTOF data from DIPG, a childhood brain tumor. He recognized the shape of the problem, dozens of numbers per cell, and stayed.

Breast cancer came through Shema’s tie to Kaplan Medical Center in Rehovot. The tags on a freshly cut tumor start to fade within hours. Dr. Noa Furth, in Shema’s lab, recruited 12 women who were about to have tumors removed, most of them luminal and one triple-negative. “Once a tumor is removed, there’s only a short window of opportunity before its epigenetic tags begin to deteriorate,” Furth said. “So as each surgery began, the doctors would alert me. I would rush the tissue sample from the hospital to our lab on ice, and we’d start analyzing it that very day.”

I’m a nuclear physicist. My day job is measuring beta-decay spectra and the radius of the proton, and I have never taken a biology class. Not even in high school. So how did my group end up on a breast cancer paper in Nature? A colleague asked me to look at some data.

Guy Ron, professor of physics, Hebrew University of Jerusalem

Yishai Arieli, Ron’s Ph.D. student, built the CyTOF analysis. Standard tools gave each cell a score but not a sense of whether that score beat chance, so the group wrote PermCell, which compares a cell’s real marker set with many random sets of the same size. Physicists, Ron noted, estimate the background before claiming a signal. Across those 12 tumors the CyTOF set covered 1,157,765 cells, a mean of 96,480 per tumor. A model trained on one slice of each tumor called the rest of its cells as cycling, luminal, or basal-like with 94 percent accuracy.

The Cells That Would Not Stay Luminal

Luminal tumors are the ones that usually respond to hormone therapy. Inside every luminal tumor the team profiled sat a smaller group of basal-like cells: high in cytokeratin-5, low in GATA3 and oestrogen-receptor signalling, and loaded with a chromatin pattern that clustered with triple-negative lines. Monomethylation and trimethylation of histone 3 on lysine 4 were the marks most tied to that loss of luminal identity.

The obvious fear was a separate mutant clone hiding in the mass. The lab tests did not read that way.

WHY THE SWITCH LOOKS EPIGENETIC

  • Every luminal tumor: The basal-like group turned up in patient after patient, not in a single outlier.
  • Mice and dishes: Luminal tumors grown as patient-derived xenografts kept the same mix, and MCF7 cells showed it too.
  • Single-cell clones: Dr. Christina Riemenschneider grew a line back from single cells, and the basal-like cells reappeared.
  • The H3K4 tags: Those two lysine-4 marks tracked the change more tightly than a mutation story would predict.
  • Shema’s read: “The differences appear to be mainly epigenetic. When we grew breast cancer models in the lab starting with nonaggressive cells, an aggressive population always emerged eventually.”

That last point is the clinical sting. A treatment can clear the cells that look well behaved and still leave a state that other cells can enter. Shema put it as cancer cells using epigenetic flexibility to change behavior, become more aggressive, and evade treatment.

Who Relapses After a Low-Risk Score?

Even when breast cancer is found early and treatment looks as if it worked, the disease returns in 20 to 40 percent of cases. Ron’s group then asked whether the basal-like blood signature tracked that return in the Mayo women, whose outcomes were already known.

THE RELAPSE SIGNAL IN PLASMA

  • Follow-up window: At least five years at the Mayo Clinic in Jacksonville.
  • Main split: EPINUC separated women who relapsed from those who did not with an AUC of 0.81.
  • Second group: The same approach reached 0.92 in a much smaller set of 30 women.
  • Low-risk flags: Ron said the test flagged relapse risk in patients conventional measures had called low risk.

Weizmann said the blood read outperformed conventional tissue biopsies for recurrence risk. In a Mayo subset that had OncotypeDX scores, those scores did not significantly differ between 40 women who stayed disease-free and 9 who relapsed, a small comparison that still fits the paper’s wider claim: the plasma basal signature carried prognostic information that standard sorting missed. “When we looked for their epigenetic signature in blood samples, we found that the more prevalent it was, the greater the risk that the disease would come back,” Ron said.

Dense Tissue Hid Tumors That Blood Still Flagged

Shema’s screening point was narrower than a replacement claim. “Breast cancer screening saves lives,” she said. “Regular mammograms reduce the risk of dying from the disease by 22 percent, but the denser the breast tissue, the easier it is for a tumor to ‘hide’ in the image.” The Weizmann write-up of how the blood test reads nucleosomes added that EPINUC’s diagnostic performance did not move with breast density, including BI-RADS scores for heterogeneous and extremely dense tissue, and that it held across the breast-cancer types in the expanded blood set.

FIVE-YEAR SURVIVAL BY HOW FAR IT HAS SPREAD

Extent at diagnosis Approx. five-year survival
Confined to the breast 99.6 percent
Nearby tissue or lymph nodes 86.6 percent
Distant sites 32.9 percent

Those survival figures, and the note that about one-third of U.S. breast cancers diagnosed from 2019 to 2023 were already past the earliest stage, sat in Weizmann’s science box beside the paper. They are why a blood mark that does not care about density is being watched. They are also why a 0.95 on stored samples is not a screening program. The classifier was trained on women who already had luminal disease, not on a random screening population, and the paper itself is still a research assay.

Stored Samples, a Patent and a Treatment Hunt

Public posts after the paper mostly repeated the 0.95 figure as if a clinic could draw the tube tomorrow. The corresponding physicist was blunter.

HOW EPINUC GOT HERE

  1. 2021: Shema hands Ron DIPG CyTOF data at an Adams Fellows seminar; the physics group never really leaves.
  2. Sept. 8, 2022: EPINUC is published on colorectal and pancreatic plasma, including early colorectal disease.
  3. During the breast study: Furth runs 12 fresh Kaplan tumors the day of surgery; Mayo supplies the five-year blood set.
  4. May 2025: The breast paper is submitted; reviewers demand more EPINUC cohorts.
  5. September 2026: Acceptance, after 16 months and, Ron said, more than twice the data of the first draft.
  6. Nov. 11, 2026: Nature publishes, with collaborators at Kaplan, Sheba, Mayo Jacksonville, Johns Hopkins, the Max Planck Institute for Molecular Genetics, the European Institute of Oncology, PSL University and Epivision Genetics in Boston.

Yeda, Weizmann’s technology-transfer arm, has filed a patent on EPINUC and licensed it exclusively to Epivision Genetics, with Shema among the named inventors, Ron wrote in his note on stored samples. Lead authors on the paper are Furth, Riemenschneider and Arieli. Ron was plain about the gap that remains: everything in the blood analysis comes from stored samples, from women whose outcomes the team already knew, and turning that into a diagnostic means bigger studies that follow patients forward in time. EPINUC is a research tool. Nobody can order it.

Shema’s next sentence points past prediction. The methods, she said, let the group track aggressive cell populations in more than one cancer, and the lab is now looking for treatments that could take away that epigenetic flexibility, and perhaps even coax cells that have already turned aggressive into calmer traits.

Disclaimer: This article is news reporting on a peer-reviewed research study and is for information only. It is not medical advice, a diagnosis, or a recommendation to start, stop, or change any breast screening, biopsy, or cancer treatment. Readers should talk with a qualified physician or oncologist about personal screening and follow-up before acting on any of these findings. Figures and statuses reflect the published Nature paper and the institutes’ statements and may change as larger trials are done.

Harry is the editor and lead writer of STUDIO ONE NETWORKS, an independent title he owns and runs himself. Ten years in journalism, reporting first and editing later, taught him that entertainment and business are one beat seen from two sides: a box office figure is a company number, a streaming deal is a contract, a casting rumour is not a story until someone puts their name to it. He works from the record, whether that is a distributor's statement, a licensing agreement, an interview transcript or a set of published ratings, and checks every number against it before publication. The same rule holds for the rest of the site, which covers news, technology, science, sports, lifestyle, travel, auto and gaming for an audience spread across the world. When he gets something wrong, the article is corrected and the change is noted and dated, under a corrections policy anyone can read. Reader mail is answered by him at support@studioonenetworks.com.

Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Trending