Science Breakthroughs This Week: 22-26 October 2026, Scored

A recurring, evidence-scored roundup of the newest science on the Frontier index.

Papers that entered the index after this issue are ranked on New this week, which updates every day.

24 new breakthroughs went live on Frontier between 22 and 26 October 2026, and biomedicine led them with 9.

Every entry in this issue is a primary research result traced to its source and scored from 0 to 100 on three axes: Evidence, Impact and Novelty. All 24 are peer-reviewed journal papers: 23 were published between 8 and 24 October and the 24th, a pancreatic cancer trial, on 31 May, and all of them entered the index this week. Thirteen appeared in three of the five general flagship journals: Nature (8), the New England Journal of Medicine (3) and Science (2). Scores ran from 59 to 71 (46 to 74 in the previous Dispatch), and two results reached 71.

The top two (Frontier Score 71 each) are both in Nature. In worms, cold raised muscle production of a metabolite, BH2, that extended their lifespan and curbed disease-related protein clumping in distant tissues; and in mice and human neurons, loss of the Angelman syndrome gene UBE3A left an excess of the enzyme asparagine synthetase, and correcting that excess genetically rescued the mice's small brains and many behavioural deficits. Next come a four-region map of the human insula from 87 people with brain electrodes (69) and blinks timed to head movements in everyday gaze (68).

Section 2 sets two Nature Genetics papers on the statistics of gene hunting side by side; section 3 takes the week's four clinical trials (59 to 66) and explains why the week's most-cited paper, a pancreatic cancer trial, scored joint lowest (59); section 4 gathers four papers on how coasts, forests, ice and power grids respond to storms, drought and a changing climate.

Biomedicine and health led with 9 entries, then neuroscience, climate and materials and energy (4 each), physics (2) and space (1). Scores as read on 27 October; the index re-scores entries as evidence moves, so a live page can differ.

1. The top of the table, in depth

The four results that scored 68 or above, in score order, each linked to its Frontier entry and to the paper.

Cold activates a BH2 metabolic network promoting longevity and proteostasis

Frontier Score 71 · Biomedicine & Health · Nature

Moderately cold temperatures extend lifespan across species, but how has been unclear. In the nematode worm Caenorhabditis elegans, cold switches up an enzyme (CAT-4, GCH1 in humans) that makes dihydrobiopterin, or BH2, in muscle; BH2 travels to the intestine and neurons, where it is turned into BH4, a cofactor for an enzyme that makes the amino acid tyrosine. Making more BH2 in muscle alone was enough to extend the worms' lifespan and suppress pathological protein aggregation in distant tissues, even at normal or raised temperatures. In human cells, adding BH2 or tyrosine reduced the aggregation of proteins associated with Huntington's disease and of mutant TDP-43 and FUS, which cause amyotrophic lateral sclerosis. The lifespan results are in worms and the human evidence is from cells, which is not the same as protecting a person's neurons, and the abstract gives no figures for how much lifespan was extended or how far aggregation fell. Its Novelty, 79.0, is the week's highest. Frontier entry · Paper

Persistent neuronal stress signalling underlies Angelman syndrome phenotypes

Frontier Score 71 · Neuroscience · Nature

Angelman syndrome, a severe neurodevelopmental disorder, is caused by loss of UBE3A, an enzyme that tags other proteins with ubiquitin, but which of its targets drive the symptoms had been uncertain. Using proteomics alongside gain- and loss-of-function mutations, the authors show that asparagine synthetase (ASNS) is a target of UBE3A and is more abundant in the brains of Ube3a-null mice and in human neurons. Genetically correcting that excess rescued microcephaly and many behavioural deficits in the mice. Excess ASNS activated stress signalling and globally reduced protein synthesis in neurons, which the authors read as UBE3A loss trapping neurons in persistent stress during sensitive periods of development. The rescue was a genetic correction, not a drug, and the abstract does not say which deficits were rescued and which were not; that ASNS and the stress response are therapeutic targets is the authors' suggestion. Frontier entry · Paper

Functional architecture of the human insula

Frontier Score 69 · Neuroscience · Nature

The insula has a central role in pain, interoception (the sense of the body's internal state) and feelings, but how its parts are organized and influence one another had been poorly understood. Combining electrical stimulation, recordings and connectivity mapping in 87 people with implanted brain electrodes, the authors describe four regions: a dorsal-posterior one that encodes pain-related (nociceptive), temperature and somatosensory signals by body part, a ventral-posterior one with access to sensation across several body parts, a mid-anterior one tied to visceral sensations and anxiety, and an anterior-polar one, largely silent when stimulated directly but active during salience detection. By the authors' connectivity measures, that anterior region exerts strong, direct influence on the others while receiving weaker, indirect input back, which they read as a pathway for the cognitive modulation of pain. "Direct" rests on responses evoked by single electrical pulses, which the open preprint of the paper calls "a proxy measure of causal interactions", and the abstract reports no test of whether changing the anterior region's activity changes how pain is felt. Frontier entry · Paper

Blinks are strategically coupled with head movements in unconstrained natural gaze behavior

Frontier Score 68 · Neuroscience · PNAS

Blinks interrupt vision for up to 10% of waking time, yet go largely unnoticed. In a large dataset of gaze behaviour during two natural tasks, with the head free to move, blink probability rose with head velocity, dropped before a head movement began and peaked during it, when visual input is unreliable anyway. People who blinked more at baseline tied their blinks to head movements more tightly, which the authors read as an individually tailored strategy for coordinating eyes and head that limits what blinks cost. The abstract gives no number of participants, does not name the two tasks and reports no effect sizes, and that the timing saves visual information is the authors' interpretation of an observed relationship. Frontier entry · Paper

2. The week's pair: gene hunting checks its own statistics

Two Nature Genetics methods papers joined the index this week, and both look at steps geneticists run routinely after a genome-wide association study (GWAS), the scans that link stretches of DNA to a trait. One finds that a popular way of reading a gene's effect often gets its direction wrong; the other builds a faster, more accurate way of separating independent signals when studies pool people of several ancestries. They score 66 and 63: the audit's higher Impact (50.1 against 40.0) and Novelty (65.0 against 62.1) outweigh the method paper's higher Evidence (90.7 against 86.2).

High false sign rates in transcriptome-wide association studies

Frontier Score 66 · Biomedicine & Health · Nature Genetics

Transcriptome-wide association studies (TWAS) are widely used to find genes involved in complex traits and to say whether more of a gene's expression raises or lowers a trait. The authors checked results for plasma proteins against the expectation that more expression of a gene generally means more of its protein, and results for complex traits against loss-of-function burden tests. TWAS had a false sign rate of 23% for plasma proteins and 33% for complex traits. Much of the discordance, they propose, comes from expression signals in tissues not relevant to the trait tagging trait-associated stretches of DNA through distinct, tightly linked variants, so TWAS can find the right gene with an unreliable direction. The benchmarks are expectations rather than measured truth, and the abstract does not say which TWAS methods or studies were assessed. Frontier entry · Paper

Multiancestry conditional and joint analysis applied to GWAS summary statistics

Frontier Score 63 · Biomedicine & Health · Nature Genetics

Conditional and joint analysis (COJO) is the step that sorts a GWAS hit into independent signals, and GWAS meta-analyses increasingly pool several ancestry groups. In simulations and real data, the new method, Manc-COJO, found more independent signals with fewer false positives than single-ancestry COJO and its ad hoc multiancestry adaptations; a companion algorithm finds ancestry-specific associations, and the software runs more than 120-fold faster than GCTA-COJO and works from precomputed linkage disequilibrium matrices when data cannot be shared. The abstract gives no figures for how many more signals it finds or how far it cuts false positives, and does not say which traits or ancestry groups the real-data analyses covered. Frontier entry · Paper

Neither paper is a discovery about a disease. Both bear on how much weight to put on the gene lists other studies publish: one says a TWAS direction of effect deserves caution, and the other gives studies of mixed ancestry a tool built for them.

3. Four trials, and why the week's most-cited paper scored joint lowest

Four clinical trials joined the index this week, three in the New England Journal of Medicine and one in Nature Medicine, and they score from 59 to 66. Physicians who read trials this closely are the reviewers AI labs now pay by the hour to check their models' medical answers, and the physician roles on AITraining.jobs, a jobs site built by Frontier's makers, list that work with its pay.

Neoadjuvant nivolumab with or without visugromab in muscle-invasive bladder cancer: the phase 2a GDFATHER-NEO trial

Frontier Score 66 · Biomedicine & Health · Nature Medicine

GDF-15 is overexpressed in muscle-invasive bladder cancer and has been named as a possible target for overcoming resistance to PD-1/PD-L1 immunotherapy. In this exploratory trial, 31 patients who were ineligible for or refused cisplatin received three cycles of nivolumab before surgery, with the GDF-15-neutralizing antibody visugromab or with placebo. The pathologic complete response rate was 33% (5 of 15) with the combination against 8% (1 of 13) with placebo; major pathologic response was 60% against 23%, with no treatment-related delays to surgery. Patients were allocated sequentially rather than randomized, the trial was single-blinded, the confidence intervals are wide and overlap (12 to 62% against 0 to 36%), and no survival results are reported. Frontier entry · Paper

Ianalumab for the Treatment of Patients with Sjögren's Disease

Frontier Score 63 · Biomedicine & Health · New England Journal of Medicine

Sjögren's disease has no approved targeted therapy. In two placebo-controlled trials, NEPTUNUS-1 and NEPTUNUS-2, monthly ianalumab, an antibody that depletes B cells and blocks the B-cell activating factor receptor, lowered a disease activity score running from 0 to 123 by 1.3 and 1.0 points more than placebo. Both trials met their primary end point, but both confidence intervals reach zero (-2.6 to 0.0 and -2.0 to 0.0), NEPTUNUS-1's P = 0.0496 sits just under the conventional threshold, and the patients' own symptom scores did not improve significantly more than with placebo at week 48. Serious adverse events were similar in both groups, injection reactions were more common with ianalumab, and the trials were funded by Novartis Pharma. Frontier entry · Paper

Mechanical Thrombectomy 24 to 72 Hours after Large-Vessel Ischemic Stroke

Frontier Score 62 · Biomedicine & Health · New England Journal of Medicine

Clot retrieval is established for acute ischemic stroke, but whether it helps patients who arrive more than 24 hours after onset had been uncertain. In 336 patients in China with a large-vessel stroke of the anterior circulation and salvageable brain tissue on imaging, randomized a median 38.5 hours after onset or last known well, thrombectomy shifted 90-day disability scores toward better outcomes (common odds ratio 1.74). It also brought more harm: symptomatic brain bleeding in 4.8% of the thrombectomy group against none of the controls, and death within 90 days in 6.0% against 1.8%. The trial was open-label with blinded outcome assessment, and the abstract does not say how salvageable tissue was defined. Frontier entry · Paper

Daraxonrasib or Chemotherapy in Previously Treated Metastatic Pancreatic Cancer

Frontier Score 59 · Biomedicine & Health · New England Journal of Medicine

Oncogenic RAS mutations are present in more than 90% of pancreatic ductal adenocarcinomas, and current therapies offer limited benefit once the disease has been treated. In this phase 3 trial of 500 patients, the oral drug daraxonrasib, which inhibits the active form of both mutant and normal RAS, gave a median overall survival of 13.2 months against 6.7 with chemotherapy; in the 91.8% with RAS G12 mutations, where the primary end points were measured, it was 13.2 against 6.6 months, and median progression-free survival 7.3 against 3.5 months. Fewer patients stopped treatment for treatment-related adverse events (1.2% against 11.2%), but events of grade 3 or higher were common in both groups (61.8% and 69.6%). The trial was open-label and funded by Revolution Medicines. Frontier entry · Paper

Where the points went

The Frontier Score weighs Evidence, Impact and Novelty at 25%, 25% and 50%, then applies a recency factor. The pancreatic cancer trial shows what that factor does. Published on 31 May, it has gathered 125 citations in its first five months, a field-weighted citation impact of 319.7 times its field's average, so its Impact, 84.0, is far the highest in this issue (the next is the microquasar model at 52.9). Weighted by the three pillars alone it reads 71.2, second only to the Angelman paper's 71.4. At five months old, though, the recency factor keeps 83.1% of that standing, which brings it to 59, the week's lowest with the microquasar model. The same age shows inside Novelty, where its recency reading is 75.9 against 97 to 99.6 for the October papers. The factor exists so that the index's top stays on the active edge; a result keeps 75% of its standing at eight months and never falls below half.

The other three trials are October papers. The bladder cancer trial has the highest Evidence of the three (86.2), and it leads the Sjögren's trial on Impact too (50.7 against 40.3) despite lower Novelty (64.6 against 67.3); the Sjögren's trial in turn edges the stroke trial on Evidence and Novelty despite lower Impact (40.3 against 50.8). Novelty's lead signal, conceptual novelty, reads how unusual a paper's median pairing of research topics is: the Sjögren's trial pairs salivary gland disorders with lupus research (13,019 prior works made that pairing), and the stroke trial pairs acute ischemic stroke with cerebrovascular disease (149,959), the most common pairing in the issue, which is why it trails the Sjögren's trial on Novelty (60.9 against 67.3). The bladder cancer trial is open access in Nature Medicine, and the other two are closed-access NEJM papers (Evidence 80.9 and 75.5): closed access costs a paper 11 of its 100 Evidence points against an openly licensed one, about 3 points of score, partly offset here by NEJM's flagship venue grade.

None of this says the older trial matters less: a gain in median overall survival from 6.7 to 13.2 months in a cancer with few options bears directly on patients, and its citation count says the field has already taken notice. The score asks a narrower question, how new and how well established a result is and how strongly it is being cited so far, and it deliberately favours what is new.

4. Coasts, forests, ice and power grids

Four papers this week measure or model how coasts, forests, ice and power grids respond to storms, drought and a changing climate, and they score 63 to 65.

The contribution of salt marshes to wave attenuation during extreme storms

Frontier Score 65 · Climate & Environment · Nature

Salt marshes protect coasts with their raised platform and their plants. In a giant wave flume, the soil of a transplanted stretch of marsh barely eroded under waves of up to 2 m significant height, while its plants were progressively damaged and damped waves less and less. Vegetation damped waves most on low-energy coasts, with waves up to 1 m, where attenuation exceeded 10% and reached up to 50%; on high-energy coasts it did not exceed 10%, a distinction the authors' global assessment supports. It is one flume experiment, and the abstract does not say which marsh or plants were tested or whether the vegetation recovers. Frontier entry · Paper

Amazon carbon emissions driven by human actions amplified by droughts

Frontier Score 65 · Climate & Environment · Nature

From 15 years of atmospheric measurements over the Amazon, the authors find that the region was a net carbon source of 0.32 ± 0.08 billion tonnes of carbon a year, with fires contributing 0.40 ± 0.07, and forest uptake increasingly unable to offset emissions. Emissions were highest in warm, dry years, at 0.49 ± 0.07; wetter years with little deforestation, burning and degradation came to 0.11 ± 0.08, and 2019 to 2022, years of low climate stress, reached 0.43 ± 0.13 from human activities, mainly deforestation and degradation. These are fluxes derived from CO2 measurements, and the grouping of years by human activity and climate stress is the authors' reading; the abstract does not split the warm, dry years' emissions between human causes and drought. Frontier entry · Paper

Late Holocene reexpansion of the Laurentide Ice Sheet

Frontier Score 63 · Climate & Environment · Science

The regrowth of the Laurentide Ice Sheet from its smallest extent would mark the start of a glaciation. On Baffin Island, Barnes Ice Cap, the ice sheet's last remnant, reached its smallest size after about 3,600 years ago, regrew from about 2,000 years ago and peaked around 1960; human-caused summer warming then turned its mass balance negative even though summer sunlight there is still declining, which the authors present as the influence of greenhouse gas forcing. The conclusion rests on one ice cap, the abstract does not say how its past extent was dated, and calling its regrowth the onset of the next glaciation is the authors' framing. Frontier entry · Paper

Disentangling the evolving effects of climate change and technology diffusion on building electricity demand

Frontier Score 63 · Climate & Environment · PNAS

Running 209,352 machine learning models of building electricity demand across 183 US counties, the authors find that climate change and the spread of heat pumps and insulation do not simply add up. In cold grid regions, the two together drive modelled peak demand growth roughly up to 10 times that of climate change alone by 2060; in hot regions the technologies counteract the peak growth warming drives, and warming reduces or removes the shift toward winter peaks the technologies bring. By 2060, technology diffusion accounts for 10 to 80% of the uncertainty in peak demand. These are projections that rest on the scenarios assumed, and "up to 10 times" is the high end of the range. Frontier entry · Paper

5. Also new this week

The other 10 entries, by field, with each one's score and journal; each title links to its Frontier entry, and each ends with a link to the paper.

Biomedicine & Health

• Wireless electrocatalytic closed-loop system for drug-free management of diabesity (66, Science). In type 2 diabetic mice, an implant that oxidizes blood glucose into gluconate, which the kidneys clear, restored glucose homeostasis within 1 hour without any drug, with no detectable hypoglycemia over 6 weeks of treatment, and, linked to a commercial glucose monitor, ran as a smartphone-controlled closed loop; the abstract gives no figures for how far weight or glucose fell and reports no test in people. Paper
• IL-2-producing TFH cells sustain Treg cell homeostasis and immune tolerance in the gut (62, Nature Immunology). In mice, IL-2 from follicular helper T cells sustained regulatory T cells in gut lymphoid tissue, and mice lacking these IL-2 makers developed colitis and fatal multiorgan inflammation, which antibiotics eased; whether the same niche exists in the human gut is not reported. Paper

Neuroscience

• CX3CL1 (fractalkine)–CX3CR1 interface drives CD8⁺ T cell–induced neurodegeneration in human brain organoids (65, PNAS). In lab-grown human brain organoids, amyloid-β raised the neurons' fractalkine, more CX3CR1⁺ CD8⁺ T cells migrated toward them, and T cells that got in injured neurons; neutralizing the T cells' cytokines and cytotoxic molecules rescued the damage, but blocking fractalkine or its receptor themselves is not reported in the abstract. Paper

Materials & Energy

• Gene-driven self-morphing microtubule-based active matter (63, Nature Materials). Cell-free gene expression inside a microtubule material made the motors and cross-linkers that drive it, and the gene circuit chosen set whether it ran steady, pulsed, arrested or oscillating; the abstract gives no figures for how long the behaviours last or how large the forces are. Paper
• Magnetically guided programmable thermomechanical protection enabled by biomimetic asymmetric design (63, PNAS). A two-faced magnetic elastomer modelled on shellfish and echinoderms cut impact force by 85% and gave a 40 °C temperature difference in lab tests, with stiffness tunable about 40-fold; the abstract does not give the impact energies or temperatures behind those figures. Paper
• Touch-sensing digital pixel fibre enables interactive garment GUI (62, Nature). Fibres with addressable pixels, touch sensing and circuitry turned the forearm of a sleeve into a 96 x 54 pixel touch display refreshing at 30 frames per second; the abstract gives no figures for how many washes the fibres survive or how much power they draw. Paper
• Additive manufacturing of ultralong and architected single crystals (62, Nature). Alternating, pulsed heat let a 3D printer grow metre-tall titanium alloy (Ti-6Al-4V) single crystals and single-crystal lattices, with an order-of-magnitude gain in creep resistance over standard grain-size scaling predictions, a comparison that is not against conventionally grown crystals. Paper

Physics

• The extraction of work from hidden degrees of freedom (63, Nature Physics). With a time-delayed double measurement on an optically trapped particle, an information engine used memory in the particle's surroundings to extract, in some regimes, more work than the energy stored in the observed degree of freedom alone; the abstract gives no figures for how much. Paper
• Octupole-driven spin torque switching of antiferromagnetic tunnel junctions (60, Nature). A current switched nanoscale tunnel junctions with antiferromagnetic electrodes on both sides of the barrier (PtMn3|MgO|PtMn3) by a spin-transfer torque long assumed not to exist in such devices, with 363% room-temperature magnetoresistance; no figures for speed or endurance are given. Paper

Space & Astronomy

• Microquasars as the major contributors to Galactic cosmic rays around the knee (59, Physical Review Letters). A population model finds that about 10 powerful microquasars active in the Galaxy at any time would be enough to account for the cosmic-ray proton flux around the knee of the spectrum; it is a model, prompted by gamma rays LHAASO detected from 4 sources associated with microquasars, not a detection of the protons themselves. Paper

6. What the week says

Mechanisms in model systems. The week's two top results both trace a whole-organism effect to one molecule, BH2 for cold-extended lifespan and asparagine synthetase for the symptoms of Angelman syndrome, and both find it in animals: worms in the first, mice in the second, each with supporting work in human cells or neurons. Other results further down the table are in mice too, from the IL-2-making T cells whose loss gave mice colitis to the glucose-oxidizing implant tested in diabetic mice for 6 weeks. Such results most often lead to the next experiment, and are the ones a headline most easily carries to patients too soon.

Effects that come with costs. Three results this week report a benefit with a price attached. Late clot retrieval improved stroke patients' function and raised their risk of brain bleeding and death; ianalumab met its primary end point in Sjögren's disease by about a point on a 123-point scale, with more injection reactions than placebo and no significantly greater improvement in the patients' own symptom scores; and in the flume, salt marsh plants damped waves substantially only on low-energy coasts and were progressively damaged with continued wave exposure.

Methods under audit. The TWAS paper is the clearest case of science checking its own tools this week, with a false sign rate of 23 to 33% in a widely used method. The insula map shows why such checks matter: its "direct" influence rests on responses to single electrical pulses, a proxy its own preprint names as such.

7. How these scores are built

The Frontier Score weighs Evidence (25%), Impact (25%) and Novelty (50%), then applies a recency factor under which a result keeps 75% of its standing at eight months old and never falls below half, so the active edge stays on top. A retraction collapses the score to zero, and a separate confidence label says how settled each result is: every entry in this issue but the pancreatic cancer trial reads Early signal, because their citation histories have barely begun, and the pancreatic cancer trial, five months old and already widely cited, reads Firming up.

The index takes in up to five new papers a day from leading journals, first the papers their own journal ran a commentary on, then the strongest of the rest. The full recipe, its sources and its limitations are on the methodology page, and our guide to the replication crisis explains how to judge a result this new. To see where any paper stands on citations, inside the index or not, our free paper check shows its field-weighted citation impact and its citations in five open indexes, and the list check does the same for up to 25 papers at once.