5 Science News Stories — October 26, 2026

Cancer Drugs Advance, the Early Universe Comes Alive, and CO₂ Goes Underground

Five verified research announcements dated October 25–26, 2026. Original study dates may be earlier; precise publication hours are not always available.

From precision oncology to cosmology and climate technology: five carefully sourced science developments. AI-generated editorial concept art, not documentary images or clinical evidence.

01   /   MEDICINE  /  ONCOLOGY   /   OCTOBER 25, 2026

A New KRAS-Targeted Drug Extends Survival in a Phase 3 Lung Cancer Trial

For years, KRAS was described as a cancer target that medicine could barely touch. A new Phase 3 trial suggests that the next generation of drugs aimed at one KRAS mutation can deliver a measurable survival benefit—not just a promising laboratory result. On October 25, researchers at UT MD Anderson presented the Krascendo 1 study at ESMO 2026, with results also published in The New England Journal of Medicine.

The study enrolled 338 people with previously treated advanced or metastatic non-small cell lung cancer carrying the KRAS G12C mutation. Participants received the investigational drug divarasib or one of two existing targeted treatments, sotorasib and adagrasib. KRAS G12C acts like a faulty growth switch; these medicines aim to keep the abnormal protein in its inactive state.

Median overall survival was 17.7 months with divarasib versus 12.0 months in the comparison group. Median progression-free survival—the time before the cancer worsened or the patient died—was 9.8 versus 5.7 months. The reported 44% reduction in the risk of death is a relative hazard reduction over follow-up, not a claim that 44% more patients were cured. Tumors responded in 58.4% versus 32.3% of patients, and severe treatment-related side effects occurred in 25% versus 35%.

These are clinically meaningful signals because the trial compared a new drug directly with approved drugs targeting the same mutation. Yet the findings apply only to this molecular subtype, not to every lung cancer. The study was open-label, relatively few control patients received adagrasib, and people with untreated brain metastases were excluded. It was sponsored by Roche and Genentech. Longer follow-up, regulatory assessment and independent clinical experience still matter.

The broader lesson is that precision oncology advances through carefully defined patient groups and direct comparisons—not through headlines about a universal cancer cure. For patients whose tumors carry KRAS G12C, the results could expand treatment choices if authorities approve the medicine.

PRIMARY SOURCE  UT MD Anderson Cancer Center — Krascendo 1, ESMO 2026

02   /   SPACE  /  COSMOLOGY   /   OCTOBER 25, 2026

Lumina Rebuilds the Universe’s First Two Billion Years on a Vast Digital Canvas

The early universe was not a quiet, empty place waiting for galaxies to appear. Matter clumped under gravity, the first stars switched on, and their radiation changed the gas between galaxies. A new simulation called Lumina, announced by the Smithsonian on October 25, attempts to follow these connected processes across a remarkably large volume of space.

Built by researchers at the Center for Astrophysics | Harvard & Smithsonian and collaborators, Lumina models the interaction of ordinary matter, dark matter and radiation during roughly the first two billion years after the Big Bang. It is a radiation-hydrodynamics simulation: gravity moves matter, gas flows and heats, and light changes the state of the gas as galaxies and black holes grow.

The modeled region spans a cube about 1.5 billion light-years on each side, more than 100 times the volume of earlier comparable simulations, according to the research team. Running it required more than 15,000 graphics processing units on the Frontier supercomputer at Oak Ridge National Laboratory. One unusual feature is its treatment of helium reionization: the simulation lets the changing ionization of helium feed back into the evolution of cosmic gas rather than adding the effect afterward.

Why build such a huge digital universe? Telescopes such as James Webb show astronomers luminous objects from a very early epoch, but a telescope captures only what the light reveals. A simulation connects those observations to possible physical histories. Scientists can compare predicted galaxies, radiation and intergalactic gas with actual measurements and ask which assumptions survive.

There is an important limit. Lumina is not a recording of what actually happened, and impressive scale does not prove its physics is complete. Its predictions depend on the input cosmology, how star formation and black holes are modeled, and numerical resolution. The underlying work also predates the October 25 public announcement. Future comparisons with Webb and the planned Roman telescope will provide sharper tests.

For Next Horizon, the most interesting possibility is not that computers have recreated the Big Bang perfectly. It is that observations and simulations can now challenge each other on scales that were previously beyond reach.

PRIMARY SOURCE  Monthly Notices of the Royal Astronomical Society — Lumina research paper

03   /   CLIMATE  /  TECHNOLOGY   /   OCTOBER 26, 2026

Underground CO₂ Storage Sets a New Record—but the Climate Gap Remains Huge

Carbon dioxide storage is often described as a future climate tool. A newly released international register shows that the technology is already operating at industrial scale, although it remains small relative to the size of the emissions problem. On October 26, Imperial College London announced the latest annual report from the London Register of Subsurface CO₂ Storage.

The register records 43.6 million metric tonnes of CO₂ geologically stored in 2024, the most recent year for which reporting is complete. It estimates 421.7 million tonnes stored cumulatively over roughly three decades. Its coverage has expanded to about 70 projects in 17 countries across six continents. Annual reported storage has more than doubled since 2016, when the figure was 21.0 million tonnes.

Geological storage works by capturing CO₂ from an industrial process or another source, compressing it and injecting it into suitable rock formations deep underground. Impermeable geological layers are intended to keep the gas trapped over very long periods. That is different from simply planting trees, and it is not automatically the same as removing carbon directly from the atmosphere. Much of today's stored CO₂ comes from industrial capture, and projects can differ in purpose and accounting.

The largest contribution by country came from the United States: 23 projects reported a combined 20.2 million tonnes for 2024. Brazil's Santos Basin pre-salt oilfield project was the largest individual operation, at 14.2 million tonnes. The report also documents growth in China and Australia, alongside newer participating countries and projects.

A crucial editorial detail is that these numbers can change. Researchers revised earlier totals to correct double counting, recycling of injected CO₂ and national reporting differences. This makes the register valuable as an accounting effort, not merely a record-setting headline.

The limitations are equally important. CO₂ must be captured, transported, monitored and kept underground safely; these steps cost money and energy. Storage capacity is geographically uneven, and even tens of millions of tonnes per year remain far below global annual emissions. Carbon storage may help tackle hard-to-abate sectors, but it cannot replace rapid emissions reductions. The future test is whether projects can expand with transparent measurement and credible long-term monitoring.

PRIMARY SOURCE  Imperial College London — London Register of Subsurface CO₂ Storage 2026

04   /   MEDICINE  /  PERSONALIZED CARE   /   OCTOBER 25, 2026

A 50-Gene Tumor Test Could Spare Many Breast Cancer Patients Chemotherapy

One of the hardest questions after breast cancer surgery is whether chemotherapy will help enough to justify its side effects. A genomic test can sometimes provide a more individualized answer. On October 25, University College London reported a new economic analysis from the large OPTIMA trial, presented at the ESMO Congress in Madrid.

The trial enrolled 4,429 participants with hormone-sensitive, HER2-negative early breast cancer who met specific clinical criteria. The Prosigna test measures the activity of 50 genes in a sample of the tumor and estimates the risk of recurrence. This is tumor gene-expression profiling, not a test of a person's inherited DNA and not, by itself, an artificial-intelligence diagnosis.

About 67% of participants received a low test score, indicating that hormone therapy without chemotherapy was a suitable option within the study protocol. Earlier OPTIMA findings supported the clinical strategy; the new presentation asked what the strategy costs. Researchers estimated that using the test could save the UK National Health Service about £654 per eligible patient during the first two years after diagnosis, or around £4 million annually if it were used across the relevant UK population.

The savings calculation is more subtle than simply subtracting the cost of chemotherapy. Testing itself costs money. The analysis included chemotherapy drugs, nursing and treatment time, management of side effects and other care costs. Avoiding unnecessary chemotherapy may also spare patients fatigue, nausea, hair loss and infection risk—benefits that matter beyond the budget.

The limitations should travel with the headline. These findings do not mean two-thirds of all people with breast cancer can safely skip chemotherapy. They concern the trial's specified tumor subtype and eligibility criteria. The £4 million figure is a modeled national estimate, not money already saved, and longer-term economic and clinical outcomes still need assessment. Access rules also vary between UK health systems.

This is a practical example of precision medicine: measuring a tumor's biology to avoid overtreatment, rather than treating every cancer with the same intensity. Readers interested in the distinction between genetic information and medical prediction can explore our Next Horizon guide to interpreting DNA and disease risk.

PRIMARY SOURCE  University College London — OPTIMA economic analysis at ESMO 2026

05   /   MEDICINE  /  CANCER PREVENTION   /   OCTOBER 25, 2026

An Experimental Vaccine Targets Cancer Risk Before a Tumor Appears

Most cancer vaccines in development aim to help the immune system fight an existing tumor or prevent a recurrence. A different idea is to train immune defenses before cancer has appeared in people at especially high inherited risk. At ESMO 2026 on October 25, researchers from UT MD Anderson presented early booster-dose findings for an investigational vaccine called Nous-209 in people with Lynch syndrome.

Lynch syndrome is an inherited condition linked to faults in DNA mismatch repair. It substantially increases the lifetime risk of colorectal cancer and several other cancers. The proposed vaccine is designed to recognize shared abnormal protein fragments—neoantigens—that can arise when tumors with mismatch-repair defects accumulate mutations. In principle, the immune system might then spot dangerous cells earlier.

The newly reported results came from an expanded Phase 1 study that examined two ways to give a booster one year after the initial vaccine series. Researchers reported that both strategies were tolerated without severe side effects and increased T-cell immune responses, with the strongest boost seen after a single neoantigen booster dose. They also reported no advanced colorectal cancers detected in follow-up colonoscopies among study participants.

That last observation needs exceptional care. A small early-phase study without the appropriate comparison and long-term follow-up cannot establish that the vaccine prevented cancer. Lynch syndrome cancers do not arise at a predictable moment, and colonoscopy surveillance remains an essential part of care. The strongest conclusion today is about tolerability and immune activity, not demonstrated cancer prevention.

If future randomized studies show fewer cancers or delayed onset, the implications could be substantial. People with known hereditary risk might one day receive an additional preventive option alongside surveillance and established risk-reduction strategies. But there are open questions: which neoantigens provide durable protection, how often boosters are needed, whether immune responses translate into meaningful clinical benefit, and who would be eligible.

The broader theme connects to the science of detecting disease before symptoms: modern prevention increasingly combines genetics, biomarkers and careful follow-up. For now, Nous-209 remains an experimental research program, not a licensed vaccine against hereditary cancer.

PRIMARY SOURCE  UT MD Anderson Cancer Center — ESMO 2026, Nous-209 Lynch syndrome booster study