Archive · 2026 · November 2026

Daily digest — 2026-11-16

Monday, November 16, 2026

Today's digest has 27 items: eight each in medchem, pharmacology and computational discovery, plus three industry stories. Highlights include light-switchable and photocaged inhibitors, heat strain suppressing liver CYP activity, the FDA's rejection of Summit's ivonescimab and Novo's etavopivat filings.

Medicinal Chemistry

8 articles

Journal of Medicinal Chemistry (ACS)

Light-switchable GLUT1 inhibitors: azo analogues of BAY-876 block glucose transport only in the cis form

  • Starting from the GLUT1 inhibitor BAY-876, the authors simplified the scaffold and inserted an azo linker (azologization), giving nine photoswitchable analogues.
  • Eight switched cleanly with 365 nm light and resisted degradation over repeated cycling.
  • In cell models, the cis forms of three compounds (PWH203, PWH448, LLH003) inhibited GLUT1 strongly while trans forms were largely inactive, suggesting light-restricted dosing of tumours.
PointsBased on the abstract

Cis being the active form is the interesting twist, since most azologs are drawn up with trans as the binder. BAY-876 has a rigid, compact shape, so I'd want to know what the bent isomer mimics in the pocket. Also check residual cis-form activity after thermal relaxation, and whether trans retains any off-target GLUT1 binding in cells.

Journal of Medicinal Chemistry (ACS)

Coumarin-caged PD-L1 inhibitors stay silent in the dark and switch on with visible light

  • Two small-molecule PD-L1 inhibitors were masked with photolabile coumarin groups, at attachment points chosen from cocrystal binding modes.
  • In the dark the caged prodrugs were inactive; visible light released the parent inhibitors and restored T-cell function in PD-L1-dependent co-cultures.
  • The aim is local, on-demand checkpoint blockade that limits systemic immune toxicity, suited to superficial or light-accessible lesions.
PointsBased on the abstract

Dimerizing biaryl PD-L1 binders are already greasy and borderline for cell potency, so a bulky coumarin cage only helps if it truly kills binding in the dark. I'd want a residual-activity control on the uncleaved prodrug and on the photolysis byproduct. Also, visible light barely penetrates tissue, so red-shifted cages would be my next ask.

Angewandte Chemie International Edition (Wiley)

One-step photochemical Minisci reaction turns drug heterocycles into photoaffinity probes

  • Rather than redesigning a synthesis or relying on pre-installed handles, a photochemical Minisci reaction attaches photo-crosslinking and bioorthogonal groups directly to aromatic N-heterocycles of drug-like molecules in one step.
  • It works across a range of medicinally relevant heterocycles.
  • On pioglitazone, the resulting probes confirmed PPARγ as the target in crosslinking and pull-down experiments.
PointsBased on the abstract

Pioglitazone is a soft validation, since PPARγ was already known; I'd want to see this on an orphan hit. The pyridine in its tail is a sensible Minisci site, but I'd ask about regioselectivity and whether a bulky crosslinker plus handle still binds at the pocket. A competition pull-down with excess parent drug is the control that matters.

ChemMedChem (Wiley)

Halogens on the thiazolide core fix solubility and clearance in antiviral leads

  • Earlier thiazolide leads against HBV, HCV and influenza A were held back by poor solubility and fast metabolic clearance, so the authors made halogenated and 4'-alkylsulfone analogs.
  • A 3-position halogen improved solubility and metabolic stability together, with two chlorothiazolides keeping influenza A activity and acceptable safety margins.
  • In the sulfone series, short two- to three-carbon chains worked best, yielding three leads for in vivo work.
PointsBased on the abstract

A halogen improving solubility is the surprising part, since it usually adds lipophilicity. I'd want logD and melting point to see whether this is lower crystal packing or a phenol pKa shift, and whether the gain in stability is blocked glucuronidation. Also, if the nitrothiazole is retained, nitroreduction liability still needs an Ames check before in vivo.

Journal of Medicinal Chemistry (ACS)

Branched acyl-lysine peptides turn SIRT2 substrates into slow-turnover inhibitors that work in cells

  • Peptides bearing methyl-branched or cyclopropyl-modified acyl groups on lysine inhibit SIRT2 potently, lingering on the enzyme with almost no turnover, as expected for slow substrates.
  • Crystal structures with NAD+ show a competent but unproductive enzyme state, and comparison with SIRT1 and SIRT3 reveals isoform-specific acyl-chain paths behind selectivity.
  • Optimized peptides engage SIRT2 in cells and raise alpha-tubulin acetylation.
PointsBased on the abstract

Is this really inhibition or just a substrate the enzyme chews through slowly? The cell data suggest the former, but I'd want the thioacylation intermediate shown directly and a deacylase-dead SIRT2 control. Peptides still carry permeability and stability baggage, so the next step is a macrocyclic or peptidomimetic version, benchmarked against SirReal-type and thiomyristoyl inhibitors.

Journal of Medicinal Chemistry (ACS)

Indole CDK9 inhibitor W24 shrinks liver cancer xenografts

  • Virtual screening plus structure-guided design yielded a new indole chemotype against CDK9, a thinly explored scaffold space for solid tumors.
  • Lead W24 is a sub-nanomolar CDK9 inhibitor, with an encouraging first-pass kinase selectivity panel.
  • In liver cancer cells it caused DNA damage and lowered the survival protein Mcl-1, driving cell death.
  • It slowed HepG2 and Huh-7 xenograft growth at tolerated doses.
PointsBased on the abstract

Sub-nanomolar potency on CDK9 means little until I see how it sits against CDK2, CDK7 and the other transcriptional CDKs, since the ATP pocket is so conserved. Mcl-1 loss is the expected on-target readout, but a short half-life protein drops with any transcription hit, so I'd want a rescue or a matched inactive analog. Next step: a clean off-target panel.

Journal of Medicinal Chemistry (ACS)

First non-peptide neurolysin activators cut infarct size in a mouse stroke model

  • Neurolysin, a peptidase induced after ischemic stroke, is an emerging neuroprotective target; earlier activators were peptidomimetics.
  • A pyridine-piperazine scaffold gave the first non-peptide activators, with nanomolar potency and better stability and blood-brain barrier penetration than the peptidomimetics.
  • Lead 7g reduced infarct volume and brain water content in the MCAO mouse model, a first for a small-molecule activator.
PointsBased on the abstract

Activating an enzyme is a harder claim than inhibiting one: a fluorogenic-substrate readout can mislead if the compounds act in a substrate-dependent way. Before crediting the infarct benefit to neurolysin, I'd want a counter-screen against thimet oligopeptidase, its close homolog, and loss of efficacy in knockout animals. Still, a brain-penetrant non-peptide scaffold is a big step up from peptidomimetics.

Journal of Medicinal Chemistry (ACS)

Swapping tert-butyl for gem-dimethylcarbinol to cut lipophilicity

  • A perspective-style case for swapping tert-butyl for gem-dimethylcarbinol, which keeps similar bulk but is far less greasy than most tert-butyl mimics.
  • It may help balance solubility, permeability and metabolic clearance in multi-parameter optimization.
  • The main drawback is higher efflux, a poor fit for CNS programs unless the structure is tuned.
  • The group is stable, not a flagged toxicophore, easy to install, and present in some approved drugs.
PointsBased on the abstract

Before committing, I'd want matched pairs against the oxetane and CF3-cyclopropyl swaps. The new hydroxyl adds a hydrogen-bond donor, so a buried, greasy pocket may charge a desolvation penalty that eats the potency gain. The efflux liability is the same donor at work, so check permeability early.

Pharmacology

8 articles

Biochemical Pharmacology (Elsevier)

Gut TLR4 loss eases autism-like repetitive behavior in mice via serotonin

  • Prenatal lipopolysaccharide caused social deficits in offspring whether or not intestinal Tlr4 was deleted.
  • Repetitive behaviors and dorsal raphe serotonin suppression were milder when gut Tlr4 was absent.
  • Hypoxanthine, a colon metabolite able to cross into the brain, was identified as a candidate mediator; it was also higher in serum of children with ASD.
PointsBased on the abstract

Gut-restricted TLR4 knockout shows a mechanism, not what a drug could do: lifelong deletion is not antagonism in an adult, and social deficits persisted regardless, so only one behavioral domain moved. Hypoxanthine should not count as a target until it is given back to the knockouts and xanthine oxidase is blocked in wild-type pups, moving from correlation with human serum to causality.

British Journal of Pharmacology (Wiley)

Vitamin E directly blocks cIAP1 degradation and cell death signaling

  • A screen found that vitamin E blocks TNF-α and SMAC mimetic-induced cell death through a mechanism separate from its antioxidant role.
  • Binding assays indicate it competitively engages the BIR3 domain of cIAP1, preventing its degradation.
  • Preserved cIAP1 limits formation of the RIPK1-dependent death complex, and treatment eased colitis severity in mice.
PointsBased on the abstract

The colitis benefit is attributed to cIAP1 rather than redox, yet a lipophilic vitamin binding the BIR3 pocket needs a proper control to back that up. Non-antioxidant tocopherol analogs and cIAP1-knockout cells would test it. Also worth testing is whether vitamin E blunts SMAC mimetics in tumor models, the first clinically relevant consequence of this interaction.

bioRxiv

Sublingual buprenorphine film yields clinically relevant rat exposure

  • Buprenorphine/naloxone sublingual film was given to unanesthetized male rats, a route closer to clinical use than earlier rodent methods.
  • Plasma buprenorphine peaked within hours of a single dose and stayed measurable a day later, within a range described as therapeutic in patients.
  • After three weeks of daily dosing, liver weight fell while serum liver enzymes and histology were unchanged.
PointsBased on the abstract

Does matching plasma levels in a rat mean matching the clinic? Not by itself. Rats metabolize buprenorphine quickly, and the film's real value is bypassing first-pass loss, so active metabolite and naloxone exposure should be reported too. The lower liver weight without enzyme change also needs a vehicle-film control before anyone calls it benign.

bioRxiv

Blocking lymphocyte entry raises some HIV drug levels inside lymph node cells

  • In mice, blocking lymphocyte entry into lymph nodes raised intracellular levels of tenofovir and efavirenz there, but not raltegravir.
  • A computational model attributes this to a larger extracellular fluid volume in the node, which counterintuitively raises intracellular concentration.
  • The size of the effect depends on how quickly each drug equilibrates between extracellular and intracellular spaces.
PointsBased on the abstract

A higher whole-node number is not the same as more drug where the virus replicates. A raw intracellular concentration can move when the extracellular volume shifts, without any change in exposure inside infected cells. Cell-sorted intranodal measurements are needed before this is treated as a PK lever, along with a realistic view of how a trafficking blocker would ever sit alongside ART.

bioRxiv

Beta1 receptor location in heart cells shapes its signaling output

  • Beta1-adrenergic receptors in heart muscle cells were probed with localized activation and live imaging of cAMP signaling.
  • Receptors in the T-tubule membrane mainly influenced contraction, whereas those on the outer membrane engaged nuclear signaling linked to hypertrophy.
  • The split held in failing rodent and human hearts, and outer-membrane-selective beta-blockers eased remodeling without weakening contractility in experimental models.
PointsBased on the abstract

Compartment-specific pharmacology is appealing, but the benefit may come from lower effective occupancy of an ordinary beta-blocker rather than from location selectivity. Matched-occupancy comparators and myocardial exposure data showing the drug sits where claimed would settle it. Human tissue confirms the target geography, not the therapeutic window.

Journal of Pharmacology and Experimental Therapeutics (ASPET)

Bcl-xL inhibition curbs retinal fibrosis in a cell coculture model

  • A self-assembling retinal coculture modeled fibrosis; standard anti-VEGF agents showed no measurable antifibrotic effect.
  • A pilot screen flagged Bcl-xL as a promising target, and the clinical candidate inhibitor UBX-1325 was tested against it.
  • UBX-1325 induced apoptosis in activated Müller cells and suppressed fibrotic activation in RPE cells without apoptosis.
  • BCL2L1 expression was higher in patient proliferative vitreoretinopathy membranes than in epiretinal membranes.
PointsBased on the abstract

Anti-VEGF failing is a low bar to clear, so that comparator says little about Bcl-xL. On-target proof comes first: Bcl-xL knockdown or an inactive analog in the same coculture, because Bcl-2 family inhibitors can overlap in selectivity. After that, an intravitreal PVR animal model with measured retinal exposure, since a dish has no vitreous or clearance.

Drug Metabolism and Disposition (ASPET)

Heat strain in rats lowers CYP1A2 and CYP2E1 activity in liver

  • In a rat model of sustained, sublethal heat strain, animals showed lower body weight and higher water intake.
  • Liver microsome tests found reduced activity for CYP1A2 and CYP2E1, while several other isoforms were unchanged or trended lower.
  • Chlorzoxazone, a CYP2E1 probe, showed higher systemic exposure and slower elimination, with its metabolite forming later and persisting longer.
PointsBased on the abstract

The missing control is a pair-fed, thermoneutral group. Without it, reduced food intake alone could explain the CYP1A2 and CYP2E1 drop, so heat cannot be separated from lost weight. Chlorzoxazone is a fair CYP2E1 probe, but enzyme protein levels and hepatic blood flow data should come before any generalising. The sensible next step is a human probe-cocktail study in heat-exposed subjects, centred on narrow-margin CYP1A2 substrates.

Drug Metabolism and Disposition (ASPET)

Loxoprofen activation favors human intestine, reversed in rats

  • Loxoprofen is a prodrug activated in the liver. Rat studies suggested little GI injury, unlike clinical reports, prompting a human-rat comparison.
  • Intestinal conversion in humans exceeded hepatic conversion, while rats showed the opposite pattern.
  • Human intestinal activity is linked to high expression of the reductases AKR1C3 and CBR1.
  • Rat recombinant reductases were active toward loxoprofen, yet none fully explained tissue activity.
PointsBased on the abstract

Can a rat GI-tolerability study be trusted for a prodrug whose activation site flips between species? Probably not here. Human intestinal S9 or organoid data with selective AKR1C3 and CBR1 inhibitors, plus mucosal exposure of the active metabolite, should come before the rat signal is dismissed, since NSAID gut injury also reflects systemic COX inhibition and direct local effects.

Computational & AI

8 articles

Journal of Chemical Theory and Computation (ACS)

Equivariant graph network converts Martini 3 lipid membranes to atomistic detail

  • MemBack, an SE(3)-equivariant graph network, predicts heavy-atom CHARMM36 lipid coordinates from Martini 3 coarse-grained configurations in one pass, then cleans up automatically.
  • Per-lipid error stayed low and similar for lipid species and membranes held out of training.
  • In a held-out red-blood-cell membrane, thickness, area per lipid and chain order matched atomistic references after restrained minimization alone, with phase-separated domains preserved; very large systems also stayed consistent.
PointsBased on the abstract

Thickness, area per lipid and chain order after only restrained minimization is a decent test, but those are bulk averages that a model can hit while local packing is off. I'd want headgroup hydration, lipid-protein contacts and cholesterol orientation checked, plus a comparison against CG2AT-style fragment backmapping on cost and quality, before trusting it with a protein-embedded membrane.

Journal of Chemical Information and Modeling (ACS)

Clearance prediction is the main source of exposure error in ML–PBPK pipelines

  • Graph-based models predicted unbound fraction, clearance and volume of distribution, which fed fixed compound-specific PBPK models to simulate human exposure.
  • Swapping observed inputs for predicted ones one at a time, in a small compound set, showed clearance degraded exposure predictions most.
  • Graph-plus-descriptor models matched descriptor-only baselines, Cmax stayed underpredicted, and the authors expect optimism for new chemistry; retrospective only.
PointsBased on the abstract

Clearance dominating the error is unsurprising; it always has in human PK. What I'd want is the error budget in fold-error terms versus a simple in vitro–in vivo extrapolation baseline using measured microsomal or hepatocyte data. If the ML clearance doesn't beat that, the pipeline adds complexity without payoff. And with a small set and no scaffold-held-out test, I wouldn't trust it on a new series.

Journal of Computer-Aided Molecular Design (Springer Nature)

Machine-learning genome mining of Bacillus yields novel antimicrobial peptides, tested on a few synthetic candidates

  • Machine-learning mining of several thousand Bacillus genomes yielded putative antimicrobial peptides in several hundred families, sequence-distinct from human and animal AMPs yet similar in charge and hydrophobicity.
  • A MIC predictor ranked candidates; a handful were synthesized, with activity varying by species and strongest against Clostridium perfringens.
  • Predictions matched measurements for some species but poorly or inversely for others, so validation is partial.
PointsBased on the abstract

A MIC predictor that tracks measurements for some species and runs inverted for others, tested on a handful of peptides, tells me the model hasn't earned a place in triage yet. I'd want to know what it was trained on, how far these sequences sit from that training data, and whether a simple charge-and-hydrophobicity baseline ranks them as well. Then synthesize a larger, blinded set.

Journal of Computer-Aided Molecular Design (Springer Nature)

Generative and foundation-model pipeline narrows millions of molecules to tested 3CLpro inhibitors

  • Generative models, a biomedical foundation model and cheminformatics filters were chained to cut a huge, noisy pool of generated molecules to a short list.
  • Cut-offs came from an in-house 3CLpro dataset, and a subset of survivors was ranked by free energy perturbation.
  • All molecules tested experimentally kept target inhibition and one beat its parent lead; the foundation model also predicted activity well on a blind test.
PointsBased on the abstract

A handful of assayed compounds all keeping activity is encouraging but weak evidence for the pipeline: without a control arm, such as random picks or simple property-filtered analogs of the same lead, we can't tell whether FEP or the foundation model added anything. I'd also want to see how far these sit from the parent lead, and whether the cut-offs from a proprietary dataset transfer to our target.

Journal of Computer-Aided Molecular Design (Springer Nature)

Random forest screen yields GSK-3β hit with cellular anti-inflammatory activity

  • A random forest trained on literature GSK-3β inhibitors served as a first-pass filter for a commercial library, followed by ADMET filtering, docking and molecular dynamics.
  • Two selected compounds showed only moderate enzymatic inhibition.
  • The better one, a non-classical hindered aromatic ester, lowered TNF-α and IL-6 in cells with low cytotoxicity; the authors attribute this to lipophilicity and cellular accumulation, which was not directly tested.
PointsBased on the abstract

Two tested compounds with only moderate enzyme inhibition is a modest return for a screen, and I'd want the denominator: how many random picks from the same filtered library would hit at that level? Without a docking-only or similarity-search baseline, the forest's contribution is unproven. Since the cell effect is unexplained, a counter-screen against other kinases would be my next step.

Journal of Chemical Information and Modeling (ACS)

MD and network analysis map allosteric activation path in pneumococcal pyruvate kinase

  • Molecular dynamics and residue-network analysis traced how fructose-1,6-bisphosphate, bound far from the active site, activates S. pneumoniae pyruvate kinase.
  • Effector binding made motions linked to the catalytic site more coordinated and reorganized long-range correlations.
  • Mutating selected leucines on the predicted pathway altered activity, supporting the simulations; the authors suggest this refines the Rock and Lock model and could inform species-selective antimicrobials.
PointsBased on the abstract

Leucine mutants changing activity is a weak test of a network path: any perturbation near a conformational hinge can do that. I'd want a control set of off-path residues with matched burial and conservation, plus replicate or longer trajectories showing the correlations aren't sampling noise. For design, ask whether the path is pharmacologically exploitable or merely descriptive.

Journal of Chemical Information and Modeling (ACS)

EnzyArena benchmark finds zero-shot models most reliable for enzyme mutant prioritization

  • EnzyArena groups kcat and Km measurements by experimental campaign, so comparisons use matched assay conditions and avoid pooled-data confounders.
  • Across four model families, most correlated poorly or inconsistently with measured function; kinetic predictors faded on literature data, while zero-shot models held up and a consensus improved retrospective hit finding.
  • In a wet-lab round, ESM-1v beat random picks whereas UniKP fell below random.
PointsBased on the abstract

UniKP landing below random is the number I'd chase: a kcat predictor trained on pooled database values likely learned family and assay-condition signal rather than mutational effect, and the benchmark never tests that directly. Before trusting ESM-1v on our enzyme, I'd want to see how its gain over random varies by campaign and starting stability, plus a simple stability-predictor baseline combined with it.

Journal of Chemical Information and Modeling (ACS)

DockBind pairs quality-filtered docked poses with measured affinities for scoring-function training

  • Scarce training data limits machine-learning scoring functions, so the authors built a docking workflow aimed at realistic modeled poses.
  • It was tuned on an expanded benchmark comparing free and template-based docking with similarity and maximum-common-substructure pose filters, then applied to records from Binding MOAD, the PDB, ChEMBL and BindingDB.
  • The resulting DockBind set spans many targets and diverse chemistry for training and benchmarking scoring functions.
PointsBased on the abstract

My first question is what the labels actually are: affinities measured for the crystal ligand, now attached to a docked pose that may be wrong. The pose filters help, but a scoring function trained on this can learn the docking program's biases. I'd want a time- or target-split test, plus a check against crystal-pose training, before trusting any gain.

Industry News

3 articles

Summit Therapeutics / AkesoBioSpace

FDA rejects Summit's ivonescimab in EGFR-mutated lung cancer, citing missing survival benefit

  • FDA issued a complete response letter for ivonescimab plus chemotherapy in EGFR-mutated NSCLC after TKI therapy, which would have been the first US approval of a PD-1xVEGF bispecific.
  • The agency said progression-free survival gains without a statistically significant overall survival benefit do not show a positive benefit-risk profile; HARMONi missed its OS endpoint, and Summit's September follow-up analysis did not change that.
  • Summit says it will keep building evidence across several NSCLC settings and gave no next step for this indication; partner Akeso already markets the drug in China.
PointsBased on public reports

The rejection was widely expected, and the lesson is about trial design: a PFS-only win in a post-TKI setting will not clear today's FDA. The bigger pivot is the Western first-line HARMONi-3 readout, which now carries the whole US thesis, and the combination partners lined up with Summit (AstraZeneca, Pfizer, Revolution Medicines) are holding options, not commitments.

Novo NordiskBioPharma Dive

Novo files etavopivat for sickle cell disease in the US and Europe

  • Novo Nordisk has submitted US and European marketing applications for etavopivat, a once-daily oral treatment for sickle cell disease.
  • The drug came from the roughly $1 billion Forma Therapeutics acquisition; Novo describes it as a potential new standard of care and a future blockbuster.
  • Etavopivat is also in mid-stage testing for thalassemia.
PointsBased on public reports

For a company defined by obesity, a rare-disease filing is a quiet portfolio hedge, and an oral once-daily option fits a market still dominated by infusions and gene therapies. The comparison that matters is with Agios's mitapivat, another PK activator moving toward sickle cell: no head-to-head exists, so label wording on crisis reduction versus haemoglobin response will decide positioning.

Abeona Therapeutics / Ellodi PharmaceuticalsBioPharma Dive

Abeona buys Phase 3-ready EoE drug from Ellodi for stock and milestones

  • Abeona acquired APT-1011, now ABO-801, an oral fluticasone treatment for eosinophilic esophagitis, from Ellodi Pharmaceuticals, a startup formed by TPG Capital.
  • Consideration is $20 million in Abeona shares plus up to $60 million in milestones tied to FDA approval and sales; Abeona also raised $30 million in a private share sale.
  • Abeona says a new Phase 3 would start in mid-2027.
PointsBased on public reports

Abeona is a gene-therapy skin company, and buying a cheap, back-loaded asset in a different disease is a diversification play funded largely with paper. The risk is the Phase 3 itself: a steroid in EoE competes with approved Dupixent and budesonide products, so the new trial design has to show a clear symptom edge, and the dilution from stock plus the placement is the price shareholders pay for that option.