Archive · 2026 · November 2026

Daily digest — 2026-11-18

Wednesday, November 18, 2026

Today's digest has 30 items: 9 medchem, 7 pharmacology, 7 computational and 7 industry. Highlights: a Nature chimera that locks the androgen receptor into a dead-end complex with p300/CBP, Lilly's up-to-$5.9B Stipple pact, and Roche's giredestrant approval.

Medicinal Chemistry

9 articles

chemRxiv

Robots on resin: solid-phase chemistry feeds direct-to-biology

  • Direct-to-biology libraries usually rely on a few assay-tolerant solution reactions; this work instead automates multistep resin-bound synthesis so crude products can go straight into assays.
  • Tested on several chemotypes as allosteric CCR2 ligands, crude-sample binding trends tracked those of purified compounds, supporting analogue ranking.
  • On a little-explored CCR2 hit, 1,700+ building-block combinations were accessible, and two DMTA cycles gave modest affinity gains.
PointsBased on the abstract

Crude-versus-pure agreement is the claim to stress-test first. Known CCR2 allosteric sites are intracellular and lipophilic, so truncated by-products, cleaved linker fragments or leftover reagent could mimic or mask binding. An orthogonal readout and a few deliberate outliers should come before anyone trusts crude ranking. Several-fold gains after two cycles are modest; the bigger win is reaching scaffolds that amide-coupling-style D2B never touches.

bioRxiv

Genome mining turns up burkstatins, two-ring protease-blocking peptides

  • Genome mining found several bicyclic microviridin-type RiPP lineages, among them a conserved Burkholderiaceae family with two core peptides per precursor.
  • In a Paraburkholderia cluster, the ATP-grasp ligase BksD closes an ester-linked ring in each core; heterologous expression and MS/MS gave burkstatin 1 and 2.
  • Same scaffold, different profiles: burkstatin 2 strongly inhibits subtilisin and elastase; burkstatin 1 is weaker and spares elastase, implicating core sequence.
PointsBased on the abstract

One scaffold with two selectivity profiles looks like a ready-made SAR series, assuming the ring-closing ligase tolerates core variation. But the abstract gives no potency or kinetic data, so the elastase gap could come from the P1 residue or from ring conformation. P1 swaps, an alanine scan and an elastase co-crystal would separate the two. Stability and permeability remain untested.

bioRxiv

CHAMP assay ranks 40 CPPs for reaching bacterial cytosol

  • A HaloTag-based flow cytometry assay (CHAMP) reads true cytosolic arrival, not mere cell association, for 40 cell-penetrating peptides in mammalian cells and E. coli.
  • Within each peptide class, mammalian cytosolic uptake varied widely, so sequence features matter more than class.
  • Where both systems gave signal, none accumulated more in mammalian cells than in E. coli; some entered bacteria with intact membranes, others only with damage.
PointsBased on the abstract

Cytosolic arrival is the readout that matters, since a CPP stuck on the surface or trapped in endosomes is a false hit. The mammalian-versus-bacterial comparison is a useful selectivity window, but the HaloTag chloroalkane tag is a bulky, artificial cargo. Before shortlisting carriers, test whether the ranking holds with real antibiotic payloads, and in a Gram-negative setting with outer membrane and efflux in play.

bioRxiv

An ester prodrug finally lets SUGCT inhibitors work in cells

  • Earlier crystal-structure-guided SUGCT inhibitors worked in vitro but not in cells, a gap for a target proposed for glutaric aciduria type I.
  • The team ran a further biochemical screen and built a better cell-based readout.
  • Masking valsartan as its ethyl ester prodrug improved cell entry and inhibited SUGCT in GA-1 cells; meclofenamic acid emerged as another cell-active inhibitor.
PointsBased on the abstract

The open question is whether the ester is actually cleaved to the free acid inside cells. If not, intact ester versus released valsartan is ambiguous, and valsartan's AT1 pharmacology could muddy a GA-1 readout. Meclofenamic acid, an NSAID with known COX activity, looks like a tool at best. An inactive matched analog and a SUGCT knockout as controls, then a real SAR campaign, would be the way forward.

bioRxiv

A live-cell BRET thermal shift assay scales target engagement to 500+ proteins

  • BRETSA pairs luciferase-tagged targets with probes whose BRET signal changes as proteins unfold, tracking melting in living cells in plate format.
  • Unfolding curves came from 500+ proteins in 24 families; ligand engagement was confirmed for over 200, including understudied kinases.
  • Orthosteric, allosteric and covalent ligands gave potency readouts, reportedly more sensitive and scalable than other thermal shift formats; molecular glues work too.
PointsBased on the abstract

Plate-format thermal shift in live cells could triage hits on understudied kinases without a tracer built for each target, which is the appeal. But a shift is not occupancy: binding outside the active site can stabilize too, and ligands that bind without stabilizing read as misses. Orthogonal confirmation and a rank-order check against known SAR should precede any use in lead optimization.

Nature

Chimera jams AR and p300/CBP into a dead-end complex

  • DALTACs push two domains of endogenous proteins together in a non-native orientation instead of degrading them; DALTAC-1 joins the AR ligand-binding domain to the p300/CBP bromodomain.
  • The mismatched interface traps the complex in an unproductive state, shifting p300/CBP acetylation patterns and dismantling AR enhancer assemblies.
  • Reportedly it beats combined AR and p300/CBP inhibition, selectively in dual-dependent prostate cancer cells and with in vivo tumour responses.
PointsBased on the abstract

Is this a degrader without the degradation? With no E3 ligase involved, SAR turns on linker geometry and the orientation it enforces, arguably harder to tune than ternary cooperativity in a PROTAC. Open questions are whether the AR-LBD binder is an agonist or antagonist and how selective the bromodomain binder is (BRD4 and other family members). Matched-linker inactive controls and structures of the forced interface would help.

Nature Reviews Drug Discovery

Kinase drug discovery, mapped across 25 years and every platform

  • This review surveys the kinase landscape, tracing a shift from ATP-competitive small molecules to allosteric and covalent inhibitors, activators, proximity modulators, degraders and antibodies.
  • Clinical-stage agents now touch nearly half of canonical human kinases, yet development clusters on a few blockbuster targets such as EGFR; over 150 approvals are reported.
  • It also examines traits of successful agents and reasons for discontinuation and withdrawal.
PointsBased on the abstract

Covering half the kinome sounds impressive, but clinical-stage is not validated: many of those kinases still lack a drug with proven efficacy. For a team choosing a target, the useful question is why the crowded ones stay crowded; my guess is tractable ATP pockets plus ready biomarkers. Sorting the discontinuations into selectivity-driven toxicity versus missing target dependency would show whether an allosteric or degrader route beats another ATP-competitive me-too.

Journal of Medicinal Chemistry

Macrocyclic sulfonamides switch on SIRT3 to protect failing hearts

  • A modular, biomimetic design built macrocyclic sulfonamides, and SAR work identified lead 2b (Y0855).
  • 2b activates SIRT3 while sparing SIRT1, SIRT2 and SIRT6, addressing the weak potency and poor selectivity of earlier activators.
  • In cardiomyocytes it eased oxidative stress, injury and fibrosis via SIRT3; in doxorubicin-induced mice it restored cardiac function.
  • Better metabolic stability makes it a useful lead starting point.
PointsBased on the abstract

Sirtuin activators have a history of assay artifacts, so how 2b was validated matters: is activation substrate-dependent, and does it hold with native, label-free peptides? Without a co-crystal or HDX map the binding site is a guess and the macrocycle SAR can't be rationalized. Direct engagement (CETSA, SIRT3 knockout cells) and a check for off-target mitochondrial effects would firm it up.

Journal of Medicinal Chemistry

Skipping the E3 ligase: degraders that dock on the proteasome

  • This Perspective reviews degraders that tether targets straight to the proteasome, bypassing E3 ligases and ubiquitination and possibly widening the degradable proteome.
  • It surveys ligands for the 19S regulatory particle and 20S core, and how different subunits and pockets have been used.
  • Challenges split into ligand issues (unclear sites, immature SAR, uncertain linker exits) and degrader issues (exposure, selectivity, off-targets, proteostasis risk).
PointsBased on the abstract

Swapping a CRBN or VHL handle for a proteasome binder only makes sense if the ligand is not an inhibitor in disguise. Many known 20S/19S ligands perturb catalytic activity, so a degrader needs a readout separating target loss from general proteostasis stress. And with so abundant a partner, it is unclear whether ternary-complex cooperativity matters as it does with E3s. Start with a non-binding-handle control and global proteomics.

Pharmacology

7 articles

Biochemical Pharmacology

Cocaine Hydrolase Fc Fusion Blocks Lethal Cocaethylene Dosing in Rats

  • CocH5-Fc(M6), an engineered long-acting butyrylcholinesterase–Fc fusion, hydrolyses cocaethylene and its more resistant metabolite norcocaethylene far faster than wild-type enzyme in vitro.
  • In rats, a low intravenous dose given just before a lethal cocaethylene challenge prevented death.
  • Pharmacokinetics showed parent drug cleared within minutes, shunted toward a benign cleavage product and away from oxidative formation of the toxic metabolite.
PointsBased on the abstract

The shunting result is the useful part: removing parent drug by hydrolysis also starves the oxidative route that makes norcocaethylene, so exposure to the toxic species falls with it. But this is prophylaxis one minute before challenge in male rats; a patient arrives after the combined use, so I'd want a rescue-dosing arm and some female data before calling it a treatment concept.

Biochemical Pharmacology

12-HETE and BLT2 Drive Pulmonary Hypertension Remodelling; 12/15-LOX Inhibitor Helps in SuHx Mice

  • Plasma 12-HETE was higher in patients with pulmonary arterial hypertension.
  • In the SU5416/hypoxia mouse model, the 12/15-LOX inhibitor ML-351 slowed disease progression.
  • Single-cell data suggested a subset of smooth muscle cells transdifferentiates into eosinophil-like cells that, with native eosinophils, secrete 12-HETE; the BLT2 antagonist LY255283 curbed hypoxic smooth muscle proliferation.
PointsBased on the abstract

Two druggable nodes, 12/15-LOX and BLT2, make this a tidy target story, but SuHx mice are a weak predictor of clinical PAH and ML-351 is a tool compound with limited exposure data. The eosinophil-like transdifferentiation is the novel claim and needs lineage tracing, not scRNA-seq inference alone, before I'd build on it.

Biochemical Pharmacology

Mendelian Randomisation Links Higher Plasma Asparagine to Lower Osteoarthritis Risk; Supplement Protects Cartilage in Mice

  • Mendelian randomisation tied higher plasma asparagine to reduced osteoarthritis risk.
  • In IL-1β-stimulated chondrocytes, asparagine limited ROS, lipid peroxidation and matrix loss while raising SLC7A11 and GPX4; the p53 activator Nutlin-3 partly reversed this.
  • In the DMM mouse model it reduced cartilage degeneration and subchondral bone erosion.
PointsBased on the abstract

Pairing human genetic evidence with a mechanistic rescue experiment is the right shape of validation. But the causal MR link doesn't tell us whether supplemented asparagine reaches joint tissue at the concentrations used in vitro, and no dose-exposure data are reported; I'd check joint levels and a delayed-start regimen before taking it further.

Chemical Research in Toxicology

Herbicide 2,4-D Links to Prostate Cancer Risk and Cell Growth via PTGS2, but Causality Is Open

  • A cross-sectional analysis of about 1,200 participants found higher urinary 2,4-D associated with greater prostate cancer odds after covariate adjustment.
  • Network toxicology and docking nominated PTGS2 as a hub target, with enrichment in HIF-1, NF-κB and ROS pathways.
  • In prostate cancer cells, nanomolar 2,4-D exposure boosted proliferation and migration.
PointsBased on the abstract

A single spot-urine measure against prevalent cancer can't separate exposure from reverse causation, and the association is modest, so the epidemiology is hypothesis-generating at best. The more useful check is whether a PTGS2 inhibitor or knockdown blunts the effect at realistic exposure levels; docking-predicted binding alone doesn't make PTGS2 a mechanism.

Chemical Research in Toxicology

Gadolinium Contrast Agents Trigger Danger Signals in Neuronal Cells, Linear Chelate and Free Salt Included

  • Neuroblastoma SH-SY5Y cells were treated with a macrocyclic chelate (gadoterate), a linear chelate (gadopentetic acid) and free GdCl3.
  • The work asked whether they release danger-associated molecular patterns, a step toward immunogenic cell death, and tracked intracellular gadolinium accumulation and localisation.
  • It points to an immune-activating component of gadolinium toxicity beyond direct cytotoxicity.
PointsBased on the abstract

The comparison I care about is macrocyclic versus linear chelate versus free ion, since stability is what separates them clinically. But the exposures are in vitro on a neuroblastoma line, and I'd want to know how they relate to the brain gadolinium levels actually measured after repeated dosing before reading any clinical risk into it.

ACS Pharmacology & Translational Science

PTP4A3 Flagged as AML Prognostic Marker; JMS-053 Inhibitor Synergises with Cytarabine In Vitro

  • High PTP4A3 expression tracked with poor outcomes, genomic instability, oncogenic signalling and poorly differentiated myeloid populations in AML datasets.
  • The inhibitor JMS-053 reduced leukaemic cell viability and colony formation across several AML models, with apoptotic signalling and lower ERK1/2 phosphorylation.
  • It was synergistic with cytarabine.
PointsBased on the abstract

Target validation here rests on expression correlations and a tool compound, and JMS-053 is a known promiscuous-looking phosphatase inhibitor, so on-target proof needs PTP4A3 knockout or rescue. The cytarabine synergy is the translational hook, but I'd want normal haematopoietic cells and an in vivo model to see whether any therapeutic window exists.

Drug Metabolism and Disposition

Two Mechanisms Explain Vismodegib's Dose-Independent Exposure in a PBPK Model

  • Vismodegib shows less accumulation than linear PK predicts, a faster approach to steady state and similar steady-state levels across a wide dose range.
  • New in vitro binding work to alpha-1-acid glycoprotein and albumin was built into a PBPK model that already included nonsink permeation.
  • Simulations suggest saturable AAG binding raises total clearance as concentrations rise, while nonsink permeation limits absorption; both contribute.
PointsBased on the abstract

This is a clean example of nonlinearity that isn't enzyme saturation: saturable binding to a minor, variable plasma protein plus a solubility-limited absorption step. Since AAG rises in inflammation and cancer, I'd ask how the model behaves with patient-level AAG variability, and whether unbound exposure, the part that matters for efficacy, is flat or still rising with dose.

Computational & AI

7 articles

chemRxiv

Open-source docking workbench pairs consensus rescoring with pose-persistence triage

  • AnCore is an open-source desktop workbench linking a 3D viewer to a six-stage pipeline: receptor preparation, redocking checks, ligand preparation, consensus rescoring of three docking scores, interaction fingerprints and post-docking triage.
  • It was tested only on a small five-target set of literature actives and presumed decoys.
  • Consensus rescoring modestly beat a single scoring function; the triage heuristic looked stronger but was tuned in-sample and rejected many actives, so prospective validation is still needed.
PointsBased on the abstract

Both ideas are sensible, but a small set with presumed decoys flatters almost any pipeline, and a triage rule tuned in-sample that discards many actives is a further worry. Enrichment shouldn't be trusted until it is shown on independent targets with experimentally confirmed inactives, LIT-PCBA-style.

Journal of Chemical Information and Modeling (ACS)

ML docking and co-folding tools find enzyme active sites but rarely pose substrates correctly

  • Eight machine-learning docking and co-folding tools were compared with a physics-based docking program on a curated set of diverse enzyme–substrate complexes, split by metal-ion presence.
  • The evaluation was retrospective, scoring pose accuracy and physical plausibility against known structures.
  • No ML tool beat the physics baseline on plausible poses, though co-folding models localized active sites well and a small older model matched them; metal ions affected performance.
PointsBased on the abstract

Before anyone drops a physics-based docker, note that the baseline is the news: if an older program still wins on plausible poses, the PoseBusters-style validity filter is doing real work. Co-folding recall on known enzymes is also hard to separate from memorization, so a breakdown by similarity to the training set is the missing analysis. For metalloenzyme projects, keep the docking step.

chemRxiv

Affinity model derives JAK-family selectivity from predicted potency differences

  • panJAK combines molecular representations with a ligand-conditioned encoding of aligned binding pockets to predict potency at four JAK paralogs, deriving selectivity from the differences without a dedicated selectivity objective.
  • It was trained on a curated set of roughly ten thousand compounds and tested retrospectively on scaffold-disjoint splits.
  • It beat baseline models at quantitative selectivity; correlated paralog errors partly cancelled, and calibrated intervals flagged a subset of comparisons with reliable direction.
PointsBased on the abstract

Deriving selectivity from the difference of two noisy potency predictions is risky, so the cancellation of correlated errors is the interesting part. Scaffold-disjoint splits on mined ChEMBL-like data still leave assay heterogeneity and series bias, though. Open questions are how it handles a new chemotype's TYK2 versus JAK1 call and whether the intervals hold on in-house measured compounds.

Journal of Chemical Information and Modeling (ACS)

New Martini 3 coarse-grained models extend simulation to twenty steroid hormones

  • Twenty endogenous and synthetic steroid hormones were parameterized for Martini 3, building on the cholesterol model and using deposited experimental structures rather than atomistic simulations.
  • Validation compared molecular volumes and hydrophobicity trends, ran free protein–hormone simulations to see whether known binding poses reappear, and checked membrane behaviour against atomistic reference data.
  • Volumes and hydrophobicity trends matched reasonably, with high hydrophobicity overestimated; most known poses were recovered, and more flexible protein models worked best.
PointsBased on the abstract

Martini 3 steroid parameters fill a real gap, since membrane-embedded steroid binding to nuclear receptors and GPCRs is where all-atom sampling is costly. Pose recovery in free simulations says little about affinity, though, and overestimated hydrophobicity could bias partitioning. Relative free energies against a known SAR series are the check needed before it is used for ranking.

Journal of Chemical Information and Modeling (ACS)

Fingerprint-similarity prior makes molecule-language embeddings more chemically faithful

  • SAGA is a training-time regularizer that aligns a model's molecular embeddings with fingerprint-based similarity using kernel alignment, adding no parameters or fingerprint computation at inference.
  • It was added to two existing molecule-text models and tested retrospectively on retrieval, analogue search, scaffold-split property prediction, captioning and generation.
  • Retrieval, analogue quality and average property prediction improved without clear loss elsewhere; control experiments tied the gains to the chemical content of the prior.
PointsBased on the abstract

Because it costs nothing at inference, SAGA is easy to try, and the control experiments are the most convincing part. Still, Tanimoto-aligned embeddings will naturally score well on analogue search and scaffold-split averages, which reward similarity. Activity-cliff pairs, where fingerprint similarity is exactly the wrong prior, are the test that matters for SAR use.

chemRxiv

Deleting a second bound ligand copy can derail redocking, notably for P-glycoprotein

  • Standard redocking removes every copy of a ligand, so structures with two copies in one cavity are docked into a pocket never seen empty.
  • Retrospective tests on deposited P-glycoprotein structures, with the search box fixed on the known pose, compared deleting versus retaining the second copy, plus a carbon-only ghost and a volume-matched box control, and a second docking program.
  • Retaining the copy markedly improved poses, driven mainly by apolar occlusion; across eight other proteins no consistent direction emerged. Authors advise reporting both ways.
PointsBased on the abstract

This is a nice reminder that a redocking success rate is partly a property of how we prepared the structure. The ghost-atom and volume-matched controls are the right ones. For P-gp or other multi-copy cavities on your project, run both preparations and treat any benchmark that silently deleted the second copy with some suspicion.

Journal of Cheminformatics (Springer Nature)

Explicit hydrogens help a neural network recover molecular bonds from noisy 3D coordinates

  • BondNet, an equivariant network, predicts connectivity and heavy-atom bond types from 3D coordinates, and was used to test how much explicit hydrogens contribute.
  • It was evaluated retrospectively on a conformer test set and a separate PubChem3D cohort, with and without added coordinate noise.
  • Full-graph recovery was near-perfect on clean input and fell with noise, but stayed far above default RDKit; hydrogen-free inputs did worse. Limits: charged molecules and generated geometries untested.
PointsBased on the abstract

Bond orders recovered from docked or generated poses are a familiar headache, so a learned alternative to RDKit's distance-based perception is welcome. But the untested cases, charged molecules and generated geometries, are where it is most needed, so it can't yet fix that workflow. ML-generated ligands, tautomers and conjugated systems would be the real test.

Industry News

7 articles

AriBio / Fosun PharmaBioSpace

AriBio's PDE5 inhibitor AR1001 wins Phase 3 only in the monotherapy subgroup

  • POLARIS-AD randomized 1,535 early Alzheimer's patients to oral AR1001 or placebo for 52 weeks, with two primary analyses, either of which could count as success.
  • The monotherapy subgroup (604 patients, no cholinesterase inhibitor or memantine) slowed decline by 31% and met its endpoint; the full population showed only 7%, not significant.
  • Diarrhea and treatment discontinuations were higher on drug. AriBio will talk to the FDA; Fosun, holding a $60M option, is deciding whether to exercise it.
PointsBased on public reports

A win in a prespecified subgroup of a missed full-population analysis is a thin basis for filing; the multiplicity design lets it count, but regulators will ask why symptomatic therapy would mask a disease-modifying effect. The comparison with anti-amyloid antibodies is cross-trial and should not be trusted. Fosun's decision on the option is the sharper commercial read than the press release.

Genetix Biotherapeutics (formerly bluebird bio)BioSpace

Ex-bluebird Genetix says its gene therapies are profitable, guiding to about $500M in 2026

  • Genetix, bluebird bio renamed after Carlyle and SK Capital bought it in February 2025, reported $327M of U.S. net sales year to date and guides to about $500M for 2026.
  • About 700 patients have been treated since 2022, 326 of them this year; blood-disease sales (Lyfgenia, Zynteglo) are up 95% year on year.
  • It says it has been profitable since Q4 2025, and is studying Lyfgenia in children as young as two, with a filing planned to extend use to age 12.
PointsBased on public reports

This is the best public proof that an ultra-expensive gene therapy franchise can work once the overhead of a listed biotech is stripped out. Two caveats: figures are self-reported by a private company no longer obliged to disclose, and the contingent value right for old shareholders needs $600M in 12 months, which guidance of $500M does not yet reach. Rivals with sickle cell programs should read it as evidence that demand is real but throughput-limited.

Eli Lilly / Stipple BioBioSpace

Lilly pays for tumor-selective binders: up to $5.9B Stipple oncology pact

  • Lilly gets exclusive worldwide rights to oncology products from Stipple Bio's Pointillist platform, which finds binders to tumor-specific epitopes on cell-surface proteins.
  • Terms: an undisclosed upfront plus an equity investment in Stipple, and up to $5.9B in development and commercial milestones.
  • Lilly announced it alongside the Calytrix buyout, funding pipeline-building from incretin cash.
PointsBased on public reports

The headline figure is almost all milestones, so the real signal is the equity stake and exclusivity, not the $5.9B. The platform targets on-target, off-tumor toxicity, which is where many bispecifics and ADCs stumble; no target is named, so Lilly is buying a method before a molecule. Compare how Lilly pairs small early-stage buyouts with option-style platform deals: it is spreading oncology bets cheaply while the incretin franchise pays.

Marrow TherapeuticsBioPharma Dive

Marrow debuts with $173M for an in vivo siRNA rival to sickle cell gene therapy

  • Marrow emerged from stealth with $173M across Series A and B from eleven investors, including founder Scion Life Sciences, OrbiMed and venBio.
  • Lead program MAR-101 is a transferrin-receptor antibody conjugated to an siRNA, aimed at reactivating fetal hemoglobin in bone marrow cells for sickle cell disease and beta thalassemia.
  • It is preclinical, with no clinical start date given; the pitch is a repeatable, titratable injection versus Casgevy and Lyfgenia.
PointsBased on public reports

A $173M round for a preclinical asset shows investors still want an alternative to conditioning-based cures that few patients actually receive. The open question is delivery: transferrin-receptor shuttles have moved siRNA into marrow cells only in limited settings, so hematopoietic knockdown depth and duration are what to watch. A reversible, dose-titratable drug would also be easier to price for low-income markets, but it must show fetal hemoglobin levels that approach a one-time cure.

Sellas Life SciencesBioSpace

Sellas' GPS vaccine clears Phase 3 REGAL in AML, with an accelerated-approval talk next

  • Phase 3 REGAL randomized 126 AML patients in second or later remission to galinpepimut-S maintenance or investigator's-choice best available therapy.
  • Median overall survival was 24 months versus 14.7 months, a 41% lower risk of death, described as statistically significant; hazard ratio and p-value were not given.
  • Safety was mostly mild injection-site reactions. Sellas will meet the FDA about accelerated approval, with a filing planned for the first half of next year.
PointsBased on public reports

Read this against the company's own history: this trial was long expected to miss, so the first thing to test is how the final analysis was triggered and whether the open-label design and a small event count flatter the result. A 9-month median gain against best available therapy is large for a vaccine, but without a hazard ratio, confidence interval or secondary endpoints, regulators will be the real judge. The comparator, a mixed best-available arm, is the weak spot to watch at the conference data.

Novo NordiskBioSpace

Novo's CagriSema loses about 21% of body weight at lower doses in REDEFINE 9

  • Phase 3 REDEFINE 9, shown at ObesityWeek, tested weekly 1 mg and 1.7 mg doses; at week 68 weight loss was about 21% and 20.7%, versus 2% on placebo, in the adherent-patient analysis.
  • Counting discontinuation and rescue use, loss was 18.6% and 19.4%. Nausea and vomiting were similar to REDEFINE 1.
  • The FDA obesity decision is expected in 4Q26; an 80-week REDEFINE 11 trial will probe further loss.
PointsBased on public reports

The lower doses roughly match the higher dose in REDEFINE 1, which suggests the top dose was not the efficiency story and could ease the tolerability burden, though nausea at 62-66% is hardly lighter. Cross-trial comparison with Lilly's tirzepatide is risky, since the earlier head-to-head failed. The next thing to check is dropout by dose, because the label dose decides how Novo prices against Zepbound.

Roche / GenentechSTAT News

FDA approves Roche's oral SERD Wilserdly (giredestrant) for ESR1-mutated breast cancer

  • FDA approved Wilserdly (giredestrant), an oral selective estrogen receptor degrader, on 16 November for advanced HR-positive, HER2-negative breast cancer with an ESR1 mutation.
  • Approval rests on the evERA trial, in which the drug extended time to progression.
  • Roche calls it a potential top seller and enters a SERD class where it had lagged oncology rivals.
PointsBased on public reports

The approval is for the narrower ESR1-mutant setting, but the larger prize is the early-breast-cancer filing that the company has described as pending, which would move the drug from a niche second-line option into adjuvant use. Roche arrives after other oral SERDs, so differentiation will rest on efficacy in the early setting and combination use rather than being first. Watch how it is positioned against existing ESR1-mutation options and CDK4/6 sequencing.