Archive · 2026 · November 2026

Daily digest — 2026-11-17

Tuesday, November 17, 2026

Today's digest has 29 items: eight each in medchem and industry, seven in computational and six in pharmacology. Highlights are the optimization of KRAS G12D inhibitor INCB161734, J&J's up-to-$3.3B Ollin buy, Lilly's Calytrix deal and a benchmark showing cofolding models misjudge receptor state.

Medicinal Chemistry

8 articles

Bioorganic & Medicinal Chemistry Letters

Widening the SAR around ML221 for apelin receptor antagonists

  • Two ML221-derived series were built by varying the arylthio and benzoate regions, giving 32 analogues profiled in an APJ cAMP functional assay.
  • Lead Y9 beats ML221 in potency and shifts the Apelin-13 concentration-response curve further to the right, consistent with antagonism.
  • In HUVECs, Y9 had low cytotoxicity yet blocked tube formation and DNA synthesis, with G0/G1 arrest.
  • In zebrafish it curbed developmental and hypoxia-driven retinal angiogenesis without overt toxicity.
PointsBased on the abstract

The benzoate ester is the old liability: ML221's ester is a known soft spot for plasma hydrolysis, and Y9's stability is not reported in the abstract, which tempers the zebrafish result. Selectivity against the related AT1 receptor and other GPCRs is likewise absent, and a rightward curve shift alone does not prove competitive binding without Schild-type analysis.

Bioorganic & Medicinal Chemistry Letters

Lysine and tyrosine ligandability points to mitochondrial disease targets

  • Cross-referencing proteins with chemoproteomically ligandable lysine or tyrosine residues against human genetics shows enrichment for genes behind mitochondrial DNA depletion syndromes.
  • SUCLG1 carries a ligandable active-site lysine, a possible handle for covalent inhibition.
  • AGK has a ligandable tyrosine that could support an allosteric approach to modulating mitochondrial and immune signaling.
  • Offered as a framework for prioritizing genetically validated targets for covalent probes.
PointsBased on the abstract

A ligandable lysine is not necessarily an inhibitable one. SUCLG1's active-site lysine may sit where a CoA or nucleotide substrate binds, so whether a sulfonyl fluoride or activated-ester probe actually abolishes activity, and spares the heterodimer partner, remains open. For AGK, binding at a tyrosine only becomes an allosteric story once a functional readout backs it.

Journal of Medicinal Chemistry

KRAS G12D inhibitor INCB161734: how the clinical candidate was optimized

  • Optimization started from an earlier orally bioavailable KRAS G12D tool compound, with the explicit goal of lowering the predicted human dose by tuning potency and pharmacokinetics together.
  • Metabolic soft-spot analysis guided the design of low-clearance analogs; molecular size and Fsp3 were held in check to improve permeability and crystalline solubility.
  • The resulting candidate shows strong oral exposure, antiproliferative activity in xenograft models and selectivity over wild-type K/H/NRAS; it is now in patient trials.
PointsBased on the abstract

Selectivity over wild-type RAS is the headline, but its origin is unclear: engagement of the mutant aspartate, as in other non-covalent G12D series, or something distinct. Keeping size and Fsp3 down while pushing clearance lower is a genuinely hard balance in this class. Cellular selectivity against G12V and G12C backgrounds would show how far the mutant preference extends.

Journal of Medicinal Chemistry

Tricyclic pteridinone yields an oral HPK1 inhibitor that pairs with anti-PD-1

  • Structure-guided scaffold hopping and ring optimization from a known diaminopyrimidine carboxamide HPK1 chemotype gave a new tricyclic pyrido[2,1-h]pteridin-6-one core.
  • Lead 29 is highly potent with strong kinase selectivity, blocks SLP-76 phosphorylation and boosts T-cell activation in Jurkat cells and human PBMCs.
  • It shows good oral exposure and, combined with anti-PD-1 in the CT26.WT syngeneic model, clearly improves tumor control.
PointsBased on the abstract

Tricyclisation plausibly pre-organises the hinge-binding conformation and cuts the entropic cost, so the design logic is easy to follow. The harder test is selectivity within the MAP4K family, notably GLK, and against T-cell kinases such as LCK, where off-target activity could undercut the immune benefit. The combination result also lacks a monotherapy arm and an inactive analogue to show it is on-target.

Journal of Medicinal Chemistry

N-terminal amine drives NTS2 analgesia in selective neurotensin macrocycles

  • Focused library of macrocyclic neurotensin analogs merging substitutions known to help NTS2 affinity, subtype selectivity or antinociception; several kept strong NTS2 binding, one with exceptional selectivity over NTS1.
  • Stacking favorable substitutions was not additive, yet two leads were potent in spinal rat tail-flick and formalin pain models.
  • Deleting the N-terminal amine kept NTS2 affinity but erased analgesia; MD comparing a lead with its desamino analog showed lost extracellular-loop (ECL2/ECL3) contacts.
PointsBased on the abstract

This is affinity without efficacy: the desamino analogue still binds yet loses analgesia, which casts the N-terminal amine as a functional handle, perhaps a charged anchor to the extracellular loops, rather than a mere affinity contributor. G protein versus arrestin signalling for both compounds, and a stable amine isostere, would test that reading directly.

Journal of Medicinal Chemistry

Hydrazide-based degraders hit HDAC3 selectively without killing cells

  • Built from the HDAC3-preferential inhibitor FL-007, comparing two linker exit vectors; the best series keeps only modest class I enzyme inhibition yet degrades HDAC3 selectively.
  • Degradation was confirmed by immunoblot and HiBiT/LgBiT reporter assays in several cancer lines, with high maximal depletion.
  • No cytotoxicity at high doses, even in non-cancerous cells; HDAC3 loss only partly re-sensitized cisplatin-resistant cells, tied to increased H3 acetylation and p21.
PointsBased on the abstract

The interesting puzzle is where the selectivity originates. Alkyl hydrazides bind HDAC1, 2 and 3 alike, so the discrimination probably arises in the ternary complex, through how well ubiquitination geometry fits each isoform. A non-degrading hydrazide control, a ligase-dead analogue and HDAC1/2 proteomics would separate these explanations and show how wide the window really is.

Journal of Medicinal Chemistry

Electrophile-first screen yields covalent aminothiazole inhibitors of E2 enzyme UBE2D2

  • Screening the electrophile side first gave a covalent aminothiazole hit that crystallography showed labelling two cysteines of the E2 enzyme UBE2D2, the active-site Cys85 and a second site, Cys111.
  • Structure-guided changes to the amide and ester groups removed Cys111 engagement while strengthening binding at Cys85, giving a more site-selective analogue.
  • That analogue blocked UBE2D2 activity in vitro and curbed A549 lung cancer cell proliferation, offering a structural starting point for selective E2 probes.
PointsBased on the abstract

Losing Cys111 is a sensible refinement, yet E2 enzymes share a conserved active-site cysteine, so selectivity across related E2s and the wider proteome is still unproven. Without a reversible control or kinetic parameters, it is hard to tell genuine recognition of the scaffold from a warhead whose reactivity has simply been tuned.

Journal of Medicinal Chemistry

A fumarate handle turns a TTK inhibitor into a monovalent degrader

  • Appending a monomethyl fumarate electrophile to a TTK inhibitor gives a single-warhead degrader, with no bifunctional PROTAC linker needed.
  • The lead compound depletes TTK and outperforms the parent inhibitor in antiproliferative assays across several cancer lines.
  • Mechanistic work points to the E3 ligase RNF126, which the compound recruits and holds in contact with TTK, driving ubiquitination and proteasomal turnover.
  • Authors present it as a molecular glue-like TTK degrader and the fumarate as a handle with less nonspecific cytotoxicity.
PointsBased on the abstract

The central question is whether this is a true glue-like ternary complex or just a covalent adduct on an RNF126 cysteine. A saturated succinate analogue would show whether the Michael acceptor is essential, and proteome-wide reactivity profiling would be needed to back the selectivity claim. Tractability looks better than a PROTAC, though the exit vector from the kinase pocket probably dictates the SAR.

Pharmacology

6 articles

British Journal of Pharmacology

Redox and immune gene variants help explain who develops NSAID hypersensitivity, and which type

  • A Spanish case-control study compared people with confirmed NSAID hypersensitivity against tolerant controls, testing variants in antioxidant and immune genes with a replication cohort.
  • Cross-reactive reactions pointed to glutathione-handling genes, with a GSTM5 genotype tracking lower serum enzyme activity.
  • Reactions to a single NSAID instead implicated SOD1, AKR1C3 and IL4R, suggesting distinct biology for each phenotype.
PointsBased on the abstract

Serum GST activity tracking GSTM5 genotype is a nice functional hook, but serum is not the immune or skin tissue where the reaction happens, and no drug-challenge or ex vivo cell data link the enzyme to the phenotype. Before calling this causal, I'd want replication in a non-Spanish cohort and a test of whether SOD1 or IL4R variants shift T-cell or basophil responses to the culprit NSAID.

British Journal of Pharmacology

PAR2 and TRPV4 channel team up to break nasal barrier in dust mite allergy

  • In dust mite allergic rhinitis, nasal epithelium shows raised PAR2 and TRPV4; PAR2 tracked with barrier proteins in a small exploratory patient subgroup.
  • In a Der p 1 epithelial model, blocking TRPV4 prevented PAR2-driven barrier loss, and a PKCε-calcium feedback loop sustained by TRPV4 residue Y110 was implicated.
  • PAR2 and TRPV4 antagonists were evaluated in mouse allergy models.
PointsBased on the abstract

Weak synergy on co-activation undercuts the "team up" framing: if PAR2 and TRPV4 are merely additive, dual blockade may buy little over one arm. Epithelial culture with Der p 1 can't show that, and the claudin-1/ZO-1 link comes from a small exploratory subgroup, so it is correlation. Before choosing a target, I'd want single versus combined antagonists in the mouse models, with TRPV4 Y110 mutants as a specificity control.

British Journal of Pharmacology

Prenatal vitamin D eases autism-like behaviours in mice, with sex-specific neuroimmune shifts

  • A single vitamin D3 injection given to pregnant BTBR dams reduced repetitive, stereotyped behaviours in their offspring.
  • In males, treatment lowered TNF-α protein in prefrontal cortex and hippocampus, and raised hippocampal FABP4 and TGF-β2 gene expression.
  • TGF-β2 shifted in the opposite direction in females, suggesting sex-dependent neuroimmune responses in this preclinical model.
PointsBased on the abstract

Without maternal and offspring serum 25-OH vitamin D, this can't be read as a vitamin D effect with any exposure logic; a single GD9 injection is a pharmacological bolus, not a repletion model. TNF-α and TGF-β2 shifts are biomarkers, not mechanism. Next step: dose-response, female behaviour, and a second ASD-relevant model beyond BTBR.

British Journal of Pharmacology

Amylin-Type Receptor Joins CGRP Receptor in Driving Migraine Light Aversion, but Not Diarrhoea, in Mice

  • In mice, CGRP and amylin each triggered light aversion, a migraine-like behaviour, and diarrhoea.
  • The CGRP receptor blocker olcegepant failed to stop CGRP-driven light aversion, whereas the amylin-type antagonist AC187 helped partly; only the combination fully reversed it.
  • Diarrhoea tracked mainly with the canonical CGRP receptor, suggesting AMY1 is a distinct target for photophobia.
PointsBased on the abstract

Olcegepant alone failing to block CGRP-driven light aversion, while it fully blocks periorbital allodynia, is the useful dissociation: it implies photophobia runs through AMY1, not just the canonical receptor. But AC187 is a weakly selective peptide antagonist, so I'd confirm with a second AMY1-selective tool or knockout, plus exposure data, before calling AMY1 a standalone target.

ACS Pharmacology & Translational Science

Review: Adding Exercise to Incretin Drugs Remains Unproven in Obesity

  • A systematic review of 31 studies pooled roughly 10,000 adults with excess weight and compared incretin drugs, exercise, and the two together.
  • Drugs reliably beat placebo or usual care on weight and glycaemia, but few trials tested drug plus exercise head-to-head against drug alone.
  • Exercise regimens varied widely, often amounting to counselling, so any complementary benefit for fitness or weight maintenance is only tentative.
PointsBased on the abstract

Thirty-one studies and almost no head-to-head drug+exercise versus drug-alone arms: the question clinicians actually ask is barely tested. The pharmacologically interesting endpoint is lean-mass preservation during incretin weight loss, which needs DXA or MRI body composition, not scale weight. Until trials use standardized, supervised training with body-composition and fitness readouts, 'add exercise' stays a plausible hypothesis, not evidence.

bioRxiv

Not all nitazenes are equal: DFNZ spares brain oxygen where fentanyl and etonitazene do not

  • Oxygen sensors in freely moving rats compared three nitazenes with fentanyl, tracking tissue oxygen under the skin and in the nucleus accumbens.
  • All compounds lowered subcutaneous oxygen similarly, but the brain picture diverged: most drove hypoxia followed by rebound, and etonitazene caused the deepest, longest dip.
  • DFNZ barely affected brain oxygen, and only at far higher doses, suggesting nitazene analgesics need compound-by-compound safety assessment.
PointsBased on the abstract

DFNZ sparing brain oxygen while fentanyl and etonitazene do not is the interesting split, but without dose-matched analgesia (e.g. ED50-equivalent doses) it could simply mean DFNZ was underdosed centrally. Nucleus accumbens oxygen is also a regional vascular and metabolic readout, not ventilation. Next step: plethysmography or blood gases at matched analgesic exposure, with brain concentrations measured.

Computational & AI

7 articles

Journal of Cheminformatics

A movement-capped post-processing step fixes physically invalid docked poses without needing a reference ligand, while leaving pose accuracy essentially unchanged

  • CTMP takes an already selected pose and nudges the whole ligand and its rotatable acyclic bonds under strict displacement limits, keeping the molecular graph intact and emitting a geometric and identity audit.
  • It was tested on crystallographic self-docking cases from PoseX, an independent set of Vina-selected poses, and a multi-target LIT-PCBA screening panel, all retrospectively.
  • Physical validity improved markedly across all three sets, while RMSD and residue-contact agreement stayed essentially the same, so gains reflect usability rather than better localization; effects on target-level screening rankings were mixed.
PointsBased on the abstract

Clash and strain fixes are not new; most docking suites already run a final minimization, so the fair question is what this beats. A comparison against plain force-field relaxation, scored with PoseBusters-style checks, would show whether the movement cap adds anything. Since RMSD barely moved, I would not expect enrichment gains. Run it on your own top poses and see how many were actually unusable before rescoring.

Journal of Cheminformatics

A deep learning model links drug structure, cell-line gene signals and biological processes to predict synergistic anticancer combinations with interpretable attributions

  • SGP-Syn encodes each drug with a language-model and a graph-based representation fused by token-to-atom attention, then combines the drug pair with cell-line expression to infer gene-level signals.
  • Those signals pass through an ontology-structured neural network that yields hierarchical biological process representations, giving a structure-to-gene-to-process explanation chain.
  • Retrospective tests on public combination screens show gains over leading methods, especially for unseen drugs and drug pairs, with only comparable results for unseen cell lines; attribution analyses suggest testable hypotheses but are not experimentally validated.
PointsBased on the abstract

Synergy models usually look best under random splits, where the same drugs and cell lines leak across train and test, so the unseen-drug and unseen-pair gains are the part worth believing. But the weak result on unseen cell lines is what matters on a real project, since a new model system is where you most need a prediction. Before trusting the attributions, I would check whether the gene and process explanations survive randomized-ontology and shuffled-expression controls, then pick a few top-ranked combinations for a small dose-matrix test.

Journal of Chemical Information and Modeling

A graph-neural-network screening pipeline, backed by physics-based rescoring and lab assays, repurposes the beta-blocker carvedilol as a partial THRβ agonist

  • A tiered workflow first used a graph neural network to rank compounds from an internal library, then applied docking and MM/GBSA rescoring to narrow the list to carvedilol.
  • The network model beat conventional machine-learning baselines in retrospective comparison.
  • Biochemical TR-FRET and cell-based experiments showed partial agonism with coactivator preference similar to resmetirom, plus reduced lipid load and induction of metabolic genes.
  • Molecular dynamics suggested stabilization in the pocket by hydrophobic contacts and two key hydrogen bonds.
PointsBased on the abstract

Carvedilol is a promiscuous, well-characterized drug, so ask whether the GNN actually found it or whether docking and MM/GBSA did the selecting; an ablation without the network would tell you. "Beat baselines" retrospectively says little unless splits were scaffold-based. Partial agonism and coactivator bias are interesting, but what you'd check next is THRα versus THRβ selectivity and whether cell activity occurs at clinically reachable exposures.

Journal of Chemical Theory and Computation

Pairing induced fit docking with free energy calculations and a shared binding mode makes flexible, promiscuous anti-target proteins tractable for structure-based design

  • Anti-targets such as CYP3A4, CYP2D6, hERG and PXR have large, flexible pockets that bind many ligands, which has made structure-based design against them difficult.
  • The workflow combines induced fit docking with free energy perturbation, and requires a consensus binding mode across a congeneric series so that receptor conformations are sampled efficiently.
  • It was evaluated retrospectively on public datasets and applied prospectively in ongoing discovery programs, supporting its use for steering compounds away from these liabilities.
PointsBased on the abstract

The consensus-binding-mode requirement is both the trick and the weak spot. Promiscuous pockets like CYP3A4 and hERG routinely let close analogs flip poses, so FEP will look reliable exactly where the assumption holds and quietly fail where it doesn't. Before trusting it on my series, I'd check whether the method flags pose switches, and whether it beats a simple matched-pair trend on the assay we already run.

Journal of Chemical Theory and Computation

A restrained umbrella-sampling molecular dynamics protocol ranks molecular glue ternary complexes by stability and captures a key selectivity determinant that cheaper scoring methods miss

  • The authors built a molecular dynamics workflow that estimates ternary complex binding free energy by umbrella sampling along a geometric thermodynamic cycle, using conformational and orientational restraints to pull the complex apart.
  • It was tested retrospectively on interface point mutants of the CK1α–lenalidomide–CRBN complex, where predictions tracked experimental affinities closely, and on BRD4 degrader series engaging DCAF16, where the ranking agreed with biophysical assay trends.
  • The method explained BRD4 selectivity as dependent on one interfacial glycine, a result that faster MM/GBSA and HADDOCK scoring did not reproduce.
PointsBased on the abstract

Retrospective agreement on point mutants and two series is a low bar for a method this expensive: the question is whether it ranks within a congeneric glue series prospectively, where differences are small and cooperativity dominates. Umbrella sampling also hinges on the starting pose and restraint choices, so I'd want replicate runs and error bars. On my project I'd reserve it for a handful of shortlisted compounds, not triage.

Journal of Chemical Information and Modeling

A solvent-aware neural network potential corrects ligand strain estimates, especially for charged molecules that gas-phase references systematically misjudge

  • A neural potential trained on quantum-chemical data with an implicit solvent model handles neutral and charged ligands uniformly, and GPU batching makes solvated global-minimum searches feasible for whole libraries.
  • Applied retrospectively to a large set of crystal-structure ligands, strain grows with charge; the most highly charged anions are least reliable and are treated as upper bounds.
  • Gas-phase force fields match neutral cases but overstate charged-ligand strain by collapsing the unbound reference; adding implicit solvent largely fixes this, and strain proves pose-specific rather than a ligand property.
PointsBased on the abstract

Retrospective strain on crystal ligands has no experimental ground truth, so agreement between methods only shows consistency, not correctness. Before trusting it, I'd check the neural potential against DFT with the same implicit solvent on our own chemotypes, since training-set coverage of unusual charged groups is the usual weak spot. Then I'd rank-order a congeneric series by strain, and treat large values as a prompt to inspect the pose or the ligand's protonation state before blaming the ligand.

Journal of Chemical Information and Modeling

Four leading cofolding models often place ligands well yet get the receptor's functional conformational state wrong, limiting their use for state-selective design

  • Four current AI cofolding models were compared on kinase and Class A GPCR complexes, using structures deposited after every model's training cutoff so the test is genuinely out-of-sample.
  • Ligand poses frequently looked accurate while state-defining features, such as kinase activation-loop and DFG/DYG arrangements or GPCR intracellular loops, were predicted incorrectly.
  • Adding state-annotated templates or filtered alignments helped only in some systems, confidence scores tracked ligand placement but not pocket accuracy, and the one allosteric complex was missed by every model.
PointsBased on the abstract

A low ligand RMSD was never proof the pocket is right, and this study shows why that matters for kinase and GPCR work. If your design hypothesis hinges on DFG-out or an inactive-state loop, treat a cofolded pose as a hypothesis, not a structure. Before trusting it, I'd compare against docking into a state-matched experimental structure and check whether the model just recalled the nearest training-set state. The allosteric miss is a single case, so I wouldn't generalize it.

Industry News

8 articles

CG OncologyBioSpace

CG Oncology's Bladder Cancer Virus Cuts Recurrence Risk by Two-Thirds in Phase 3

  • Topline PIVOT-006 data: the oncolytic immunotherapy cretostimogene cut recurrence risk by 68% versus surveillance in intermediate-risk and high-grade Ta non-muscle-invasive bladder cancer.
  • The randomized study enrolled 367 patients; the company reports no grade 3 or higher treatment-related adverse events.
  • A second BLA is planned for Q2 2027; full data go to the Society of Urologic Oncology meeting on December 2.
PointsBased on public reports

This moves cretostimogene from the BCG-unresponsive niche into a far larger adjuvant population with no approved option, which is what the stock is really pricing. The caveats: the control arm is surveillance, not an active drug, and median follow-up is only 17 months, so durability is open. Compare it against intravesical rivals once full data show how many patients needed repeat dosing.

Flagship PioneeringEndpoints News

Flagship Puts £100M Into a London Lab to Spin Out AI-Native Companies

  • Flagship Pioneering opened the Flagship Origination Lab, U.K., in London with an initial £100M, its first company-creation unit outside the US.
  • It plans about 50 staff and expects partners to add £500M, with no partners or timetable named.
  • No first companies or founding teams have been announced; the focus is AI-native platforms in human and planetary health.
PointsBased on public reports

Flagship's model is serial company creation, so a UK lab is a bet that British academic talent and government goodwill can feed its assembly line cheaply. The £500M from partners is the figure to watch: until named, the lab is mostly a £100M promise. Compare Flagship's recent AI consolidation into Etcelera: scale is being pursued at home, and 'AI-native' is still a label to test against actual assets.

Helus PharmaFierce Biotech

Helus Psychedelic Wins Phase 3 in Depression, but Serious Adverse Events Cloud It

  • Phase 3 APPROACH tested HLP003, a deuterated psilocin, as add-on therapy for major depressive disorder in 222 adults.
  • It showed a 7.6-point placebo-adjusted MADRS improvement at six weeks, meeting primary and key secondary endpoints.
  • Three treatment-related serious adverse events occurred, all in the drug arm, one reportedly a suicide attempt; a second pivotal study reads out in Q4 2027.
PointsBased on public reports

A 7.6-point gap would be large for depression, but psychedelic trials struggle with unblinding, so functional unblinding and expectancy deserve scrutiny before the effect is believed. Safety is the bigger issue: three serious events versus none on placebo will draw FDA attention to monitoring and dosing-day supervision, whatever one case proves. Analysts split on that; EMBRACE is the real test.

IQVIA InstituteBioSpace

IQVIA: About 90% of 2016–2020 Drug Launches Won't Recoup Their Costs

  • IQVIA analyzed 825 novel US launches and projects that about 90% of medicines launched in 2016–2020 will not earn enough lifetime revenue to cover costs.
  • Positive-return launches fell from about half in the early 1990s to under one in ten.
  • Capitalized R&D cost per medicine rose from about $366M for 1991 launches to over $3.2B for 2020, with shorter product life cycles and lower success rates blamed.
PointsBased on public reports

The headline is a projection against a capitalized-cost model, so the 90% moves with its assumptions on cost of capital and failure allocation. Still, the direction explains the 2026 pattern: buy late, buy de-risked, and lean on blockbusters. A sharper reading: average economics hide a fat tail, so portfolio concentration, not average R&D efficiency, is what deal-makers should read this for.

Johnson & Johnson / Ollin BiosciencesEndpoints News

J&J Pays Up to $3.3B for Ollin to Fill Its Retina Gap

  • Johnson & Johnson will acquire Ollin Biosciences for roughly $2.2B upfront plus up to about $1.1B in regulatory and commercial milestones; closing is expected by year-end.
  • Ollin's lead, OLN324, is a bispecific antibody for diabetic macular edema and wet AMD; the company says an early head-to-head study beat Roche's Vabysmo on anatomy and vision.
  • A separate Blackstone Life Sciences agreement provides $750M of multi-year funding for late-stage development, in exchange for payments and royalties.
PointsBased on public reports

J&J has contact lenses and surgical vision but no retinal drug, so it is buying its way into the Vabysmo/Eylea market rather than building there. Note that the Blackstone money sits outside the headline price: the real net cost and the royalty drag depend on terms we can't see. The pitch rests on Ollin's own cross-study claim against Vabysmo, and the dosing interval, not just efficacy, will decide whether retinal specialists switch.

Eli Lilly / Calytrix BioBioPharma Dive

Lilly's 15th Buyout of 2026: $1.5B for Calytrix's CALR Antibody

  • Lilly will pay up to $1.5B for Calytrix Bio, with an undisclosed upfront and milestones tied to clinical and regulatory goals.
  • Calytrix is developing a biparatopic antibody against a cell-surface protein in CALR-mutant disease, aimed at myelofibrosis and essential thrombocythemia.
  • It is Lilly's 15th announced acquisition this year, over $30B in total, mostly of private startups.
PointsBased on public reports

Mutant-CALR targeting is the prize in myeloproliferative neoplasms: JAK inhibitors like Jakafi manage symptoms but don't go after the clone. Lilly is again buying early and cheaply, so the price is an option on unproven biology rather than on data. The undisclosed stage and upfront make it hard to judge; the key test is whether the antibody can spare normal cells while clearing mutant ones.

Seawolf Therapeutics / SanofiBioSpace

Seawolf Launches With $100M and a $1.2B Sanofi Deal for Redosable DNA Medicines

  • Seawolf Therapeutics raised a $100M Series A from Versant, Novartis Venture Fund, Lilly, CRISPR Therapeutics and others.
  • Its Trident platform delivers DNA payloads in redosable lipid nanoparticles, claimed to last longer than RNA-LNP therapies with similar potency and safety.
  • Sanofi signed a multi-target research and licensing deal worth up to $1.2B in milestones plus royalties, covering bleeding disorders and other indications.
PointsBased on public reports

Non-viral DNA delivery that can be redosed would sidestep the immunity and dosing limits of AAV, so a Sanofi deal plus Lilly and Novartis money is a notable vote. But the support is animal data only, and the lead program and targets are undisclosed. Compare it against durable RNA-LNP editing and Sanofi's own hemophilia pipeline before reading too much into the headline figure, since milestones dwarf the upfront.

Sofinnova PartnersBioPharma Dive

Sofinnova Closes €725M Crossover Fund to Bridge Europe's Late-Stage Financing Gap

  • Sofinnova Partners closed Sofinnova Crossover II at €725M (about $815M), its largest fund; the firm manages about €4B.
  • It plans around 15 investments in late-stage drug and device companies near commercialization, mostly European.
  • Partner Kinam Hong frames Europe's problem as scale-up financing, not innovation; BioPharma Dive data show over 70 private rounds this year went to Phase 1-or-later biotechs.
PointsBased on public reports

The signal is crossover money returning as the IPO window reopens. A €725M fund written into roughly 15 names implies cheques large enough to carry companies to a listing or a sale, the layer European biotech has lacked. Watch whether it truly stays in Europe: US-listing pull and US-heavy co-investors often drag the best companies westward. Abivax's rise is the pitch; the test is repeating it.