AI & machine learning
24 articles
A solvated neural-network potential makes ligand strain estimates usable at library scale
- Conventional strain estimates reference a single gas-phase minimum, which ignores solvent and distorts charged ligands whose internal electrostatics collapse them into compact structures in vacuo.
- AIMNet2-CPCM, a neural network potential trained on DFT-level energies and forces in a continuum solvent, handles neutral and charged species on equal footing.
- GPU batch inference makes a solvated global-minimum search feasible across a full screening library.
- Across roughly 8,000 PDB ligands, solvated strain rises with net charge, with a trend that holds up to a charge of −2.
Strain filters are only as good as the unbound reference, and anyone who has seen a zwitterion fold into a salt bridge in vacuo knows why gas-phase numbers mislead. A solvated, charge-aware potential fast enough for library triage is something I would actually put in a docking post-filter. The caveat is the less reliable behavior at higher negative charge, which matters for acids and phosphates.
Cofolding models can place the ligand right while getting the protein conformation wrong
- AlphaFold3, RoseTTAFold3, Boltz-2 and Chai-1 were benchmarked on kinases and Class A GPCRs, families whose pharmacology depends on defined conformational states.
- Only complexes released after all four models' training cutoffs were used, plus one earlier control structure, to give a prospective test.
- Correct ligand placement turned out to be systematically decoupled from accurate prediction of the surrounding protein conformation.
For anyone docking into cofolded poses, this is the warning label: a plausible ligand pose does not mean the DFG state or GPCR activation state is right. That matters most for type II kinase inhibitors and biased GPCR ligands, where the conformation is the whole point. I'd treat cofolded structures as hypotheses to validate against known conformational markers before using them for SAR design.
Pairwise-difference QSAR with quantile aggregation extrapolates beyond known potency
- Standard QSAR struggles to predict analogs more potent than anything in its training set, which is exactly the lead-optimization question.
- Pairwise-Difference QSAR learns activity changes from descriptor differences between compound pairs, then aggregates the predictions.
- On ChEMBL-derived benchmarks across about a hundred targets with the most potent compounds held out, a low-quantile aggregation beat mean and kNN aggregation, direct QSAR and Chemprop.
Every optimization team has felt this: the model ranks the series fine, then fails at the top end where you actually need it. Holding out the most potent decile is the right stress test. My caution is that public ChEMBL tails differ from a live program's SAR cliffs, so I'd trial it prospectively on one of our own series before trusting it to prioritize synthesis.
Breaking the Potency Barrier: A Practical Recipe for QSAR Extrapolation ↗
De novo designed proteins bind MYC's disordered regions and slow MYC-driven tumors
- Designed proteins target conserved epitopes in MYC, including intrinsically disordered motifs needed for its oncogenic activity.
- They bind MYC tightly in vitro; in cells they engage MYC, disrupt MAX pairing and reshape MYC-associated complexes.
- Expressing the binders curbed proliferation across several MYC-dependent hematopoietic, liver, pancreatic and patient-derived models, and depleted tumor cells in an orthotopic pancreatic model.
MYC has no pocket for a small molecule to sit in, so going after disordered epitopes with designed binders is a sensible way around that. For us it's a target-validation tool first: expression inside the cell says nothing about delivery, and a protein binder won't become a drug without a delivery story. Still, it shows where a small-molecule PPI or degrader program should aim.
Benchmark asks whether structure-based methods rescue DEL-trained ML out of distribution
- A recent NeurIPS competition showed top ML models trained on DEL data failing to generalize to out-of-distribution chemical space; this study tests whether structural modeling can bridge the gap.
- It benchmarks ML, Glide docking, GALigandDock and Boltz-2 cofolding on three biologically diverse targets screened against DELs built in several synthesis formats.
- ML does well in-distribution; the paper's focus is how each approach holds up on new chemistry.
DEL-trained models that shine on random splits and collapse on new chemotypes is exactly what bites when you try to seed a different series from a DEL campaign. Three targets is a small panel, but a like-for-like comparison with physics-based methods on identical data is useful. I'd use it to calibrate how much to trust ML triage on scaffolds the library never sampled.
Interface graph neural network predicts binding energy for folded and disordered partners
- FlexENN is a message-passing graph neural network built on a local interface graph with geometry, flexibility descriptors and sequence per residue.
- Two models are trained independently: one for folded complexes and one for complexes with a disordered partner, using a newly built IDP dataset.
- Benchmarks include tubulin–tubulin interfaces in microtubules and microtubule–MAP complexes.
Affinity predictors trained on folded complexes usually fall apart on disordered partners, so a dedicated IDP set is the real contribution here. For med chem it is protein–protein interface work rather than small-molecule scoring, and I'd test whether it ranks mutations or peptide variants reliably.
Boltz-2 Co-folding Narrows an 867-Compound Library to 40 Picks, and Most Turn Out Active
- The authors tested whether a co-folding model can replace docking for triage, applying Boltz-2 to the cancer-linked enzyme AKR1B10.
- From a chemically diverse library of 867 candidates, they chose 40 by predicted affinity plus a diversity filter.
- About 70% of those tested showed inhibition at low-micromolar or better, and several submicromolar inhibitors with novel scaffolds emerged.
A 70% confirmation rate from 40 compounds is a striking number for virtual screening, though the starting library was already small and the target is an enzyme with a lipophilic, forgiving pocket. I'd want to see the same experiment on a harder target and against a docking baseline run on identical compounds. Still, a tractable assay count and novel chemotypes make this an easy workflow to copy for hit finding.
Efficient Discovery of Potent AKR1B10 Inhibitors Using Co-folding Model Boltz-2 ↗
AI Pocket-Based Design Yields a Potent STAT3 Inhibitor for Colorectal Cancer
- A pocket-conditioned, autoregressive generative AI model was used to design candidate molecules directly within the STAT3 binding pocket, rather than screening pre-existing compound libraries.
- This structure-based generative approach produced a potent small-molecule STAT3 inhibitor that advanced to biological evaluation.
- The inhibitor showed activity against colorectal cancer, where STAT3 hyperactivation is a recognized driver of tumor growth and survival.
STAT3's SH2 dimerization pocket is shallow and dynamic — the reason most inhibitor programs (niclosamide analogs, peptidomimetics) never reached clinic. Generating atoms directly inside that pocket geometry, rather than rescoring a static virtual library, is the real methodological jump here. Still, "potent" on paper means little until we see SH2 selectivity over STAT1/5 and a synthesizable route — that's where de novo hits usually stall.
Two-Stage Generation Dynamics Enable Training-Free AI Drug Design
- Bayesian flow networks design molecules for protein pockets by jointly generating atom types and 3D coordinates.
- The two channels behave differently: atom types shift abruptly at a defined point, while coordinates sharpen gradually, a pattern consistent across model variants.
- Timing refinement to these transition points yields a training-free method matching baseline docking performance on a standard benchmark, with no extra predictor needed.
What earns this a second look is the lack of per-target retraining — that's the step that kills most generative tools before they reach my bench, so a training-free trick timed to a phase transition is worth trying. But matching baseline docking, not beating it, is a floor, not a win — and CrossDocked2020 scores say nothing about synthesizability or real binding.
Hierarchical Phase Transitions in Bayesian Flow Networks Enable Training-Free Molecular Design ↗
Antibody-Inspired Peptides Take Aim at Flu's Fusion Machinery
- Applies a hotspot-centric computational design strategy, grafting key antibody-binding side chains onto small L-peptide scaffolds that mimic how neutralizing antibodies engage influenza hemagglutinin.
- Draws design cues from the antigen-binding interfaces of broadly neutralizing antibodies FI6v3 and CR9114, both directed at the hemagglutinin stem.
- Aims to yield peptides that block the hemagglutinin conformational change driving membrane fusion, offering a structure-guided route to antibody-inspired flu therapeutics.
FI6v3 and CR9114 are smart hotspots to steal from, and a compact L-peptide is far more tractable to make and dose than the antibody it's mimicking. But stem-directed HA binders have a rough history against escape mutants — durability of coverage across drift, not just initial fusion-blocking potency, is the question I'd want answered before getting excited about a therapeutic path.
Antibody Hotspot-Centric de Novo Design of Peptide Inhibitors of Influenza A Virus Membrane Fusion ↗
Docking Showdown: Physics-Based, Co-Folding, and Hybrid Models Face Off
- Compares physics-based docking, deep-learning co-folding, and hybrid rescoring methods on how well they generalize to unfamiliar protein targets and chemical scaffolds.
- Builds on the recent Runs-n-Poses benchmark, which found that classical physics-based docking held an edge over several co-folding tools on truly novel targets.
- Assesses pose accuracy and ranking performance to offer practical guidance on choosing a docking strategy based on target novelty.
Physics-based docking still beating co-folding tools on genuinely novel targets is the finding I keep coming back to — these deep-learning models are pattern-matching PDB coverage, not solving folding physics, and it shows the moment a target has no close analog. Until a co-folding tool proves itself on a zero-analog case, I'm running both and trusting the classical scorer as the tiebreaker.
Multiparadigm Benchmark of Molecular Docking: From Physics to Co-Folding and Hybrid Models ↗
AI-Guided Optimization Yields a Potent New HBV Capsid Assembly Modulator
- Starts from known capsid assembly modulator NVR 3-778 and applies a multi-objective optimization workflow combining structure-based fragment growing with medicinal chemistry.
- Machine-learning models are used alongside traditional design to jointly predict antiviral activity and drug-like/ADMET properties.
- The workflow generated and evaluated roughly four dozen novel sulfamoylbenzamide analogs.
- Lead compound 11B-6 emerged with superior antiviral potency and improved drug-like properties versus the starting scaffold.
Forty-some analogs to land one clearly superior lead is a tight design-make-test cycle, and coupling ML-based ADMET prediction to fragment growing off NVR 3-778 is exactly where these models earn their keep — multi-parameter optimization, not potency-chasing alone. Next question: does 11B-6 hold up against the CAM-resistant HBV variants that have dogged this chemotype before.
Machine Learning and Docking Methods Show Uneven Generalization on DNA-Encoded Library Screens
- The study benchmarks machine learning, physics-based docking, and cofolding methods for predicting protein-ligand binding using DNA-encoded library (DEL) screening data.
- Methods compared include state-of-the-art ML models alongside physics-based docking tools (Schrodinger Glide, Rosetta GALigandDock) and the cofolding method Boltz-2.
- Ground-truth binding data came from ultralarge DEL selection experiments across three biologically diverse protein targets and multiple DEL synthesis formats.
- The work follows a NeurIPS competition in which even top-performing ML models trained on DEL data failed to generalize to out-of-distribution (OOD) chemical space.
- ML models performed well in-distribution, but out-of-distribution hit discrimination depended on the specific target and ligand context, with no single method consistently dominating across cases.
- The authors conclude that benchmark performance alone cannot predict OOD generalization and release an open-source package, DEL-iver, for evaluating binding-prediction methods against DEL screening data.
This is the paper you send to anyone pushing to swap your docking rescore step for a single ML model. "No method dominates OOD" is the whole ballgame: your real triage work is always on a scaffold the model hasn't seen, and a benchmark AUC built on in-distribution DEL data tells you almost nothing about how a model will rank the novel matter you bring in next week. I don't read this as an argument against ML - Boltz-2 and friends are genuinely useful, fast filters - but as a case for running physics-based docking/cofolding in parallel, not as a fallback you only reach for when ML looks shaky. Target- and ligand-context dependence also means there's no one 'best' pipeline to standardize on company-wide; you validate per target, per chemotype, every time. DEL-iver going open-source is the practically useful bit - a shared OOD benchmark beats trusting a vendor's Tanimoto-flattered internal numbers.
AI-Driven Lead Optimization Delivers 11B-6, a More Potent HBV Capsid Modulator
- Investigators describe a multi-objective lead-optimization campaign built on the sulfamoylbenzamide (SBA) chemotype exemplified by NVR 3-778, an earlier clinical-stage HBV capsid assembly modulator (CAM).
- An AI-guided computational workflow was integrated with iterative synthesis to simultaneously optimize antiviral potency alongside physicochemical and ADME/drug-like properties across the series.
- The campaign identified 11B-6 as the optimized lead, reported to combine stronger anti-HBV potency with improved drug-like characteristics relative to earlier series analogues.
- Mechanistically, 11B-6 functions as a capsid assembly modulator, engaging the HBV core (capsid) protein to misdirect nucleocapsid assembly and interfere with viral replication.
- The work positions 11B-6 as a candidate for further preclinical evaluation within the sulfamoylbenzamide CAM class for chronic hepatitis B treatment.
What actually catches my eye isn't the potency bump — it's that AI-guided optimization was wired into the synthesis loop, not run as a pre-filter before chemists took over. That's the difference between a screening deck and a real DMTA cycle. Going back to the SBA/NVR 3-778 chemotype rather than hunting a new capsid-binding scaffold is the pragmatic call: the pocket and assembly-misdirection mechanism are already de-risked clinically, so this campaign is buying SAR knowledge instead of paying discovery tax twice. That said, 'improved drug-like properties' is a claim I want backed by real microsomal stability, plasma protein binding, and hERG numbers before I get excited — potency plus a nicer cLogP on paper doesn't guarantee 11B-6 clears whatever tolerability hurdle likely sank NVR 3-778. Worth watching, not yet worth celebrating.
All-Atom Receptor-Aware Deep Learning Model Advances De Novo Peptide Binder Design
- Researchers developed REAPS, a geometric graph neural network for designing peptide binder sequences that models the target receptor as a fully observable, all-atom structural context, rather than relying on backbone geometry alone as typical inverse-folding methods do.
- In head-to-head evaluations against the widely used inverse-folding model ProteinMPNN, REAPS showed improved sequence recovery, structural fidelity, and interface-level biophysical properties for both linear and macrocyclic peptide binders.
- The team paired REAPS with structural hallucination in a closed-loop de novo design pipeline to generate novel peptide candidates without requiring a known starting scaffold.
- Candidates produced through this pipeline were synthesized and tested against the neurokinin-3 receptor (NK3R), where they showed experimentally confirmed functional activity.
Sequence recovery and "improved structural fidelity" benchmarks against ProteinMPNN are nice, but what I actually care about is whether the closed-loop hallucination + REAPS pipeline produced binders without a template scaffold — that's the real test, and NK3R hits with confirmed functional activity clear that bar. All-atom receptor context should, in principle, capture side-chain packing and H-bond networks that backbone-only inverse folding misses, which matters a lot for macrocycle design where conformational preorganization is everything. Before I'd plug this into a campaign: what's the actual hit rate and binding affinity on those synthesized NK3R peptides, not just "functional activity"? And does performance hold on a GPCR-like target outside training distribution, or is this another tool that shines on its own benchmark set and stalls on my target. Worth piloting on a live target before I swap out MPNN wholesale.
Stress-Testing AI Co-Folding Screens: Lessons from a Keap1–Nrf2 Case Study
- Multi-stage virtual screening cascades that combine AI co-folding models with learned-affinity (ML) scoring are now standard practice in computational drug discovery.
- Each stage's output score is typically taken at face value and used to advance or discard candidates without independent validation of what it actually captures.
- This preprint systematically evaluates, stage by stage, whether cascade scores provide the information practitioners need to trust them, using the Keap1-Nrf2 protein-protein interaction as a case-study target.
- The authors characterize failure modes that arise at different stages of the cascade, offering a quality-control perspective on where and why these pipelines can mislead.
Every screening campaign now runs some flavor of co-fold-then-ML-score cascade, and honestly most of us take the ranked list at face value because checking each stage is a pain. That's the exact habit this preprint is poking at, using Keap1-Nrf2 — a notoriously flat, shallow PPI interface that's already brutal for docking and pharmacophore methods — as a stress test where scoring artifacts should show up fast. If the abstract's framing holds, the real message isn't "don't use these tools," it's "don't trust a single composite score without knowing what each stage is actually measuring" — pose confidence, affinity ranking, and physical plausibility are not the same thing, and conflating them is how false positives sneak to the top. Worth revisiting once the full methodology and failure-mode breakdown are out; details here are still thin.
Benchmark Study Reveals Blind Spots in AI-Based Virtual Screening for Keap1-Nrf2
- A new benchmarking study examines a virtual-screening pipeline that chains structure-based co-folding models with learned-affinity scoring, an increasingly common setup for computational hit-finding.
- The cascade was applied to the Keap1-Nrf2 protein-protein interaction, a notoriously hard-to-drug target due to its flat, polar interface.
- Instead of trusting only the final ranked output, the authors assessed each stage separately against experimentally meaningful signal.
- They identify specific points where a stage's score diverges from that signal, a failure mode that can mislead downstream compound prioritization.
- The paper proposes per-stage control checks practitioners can run before relying on a cascade's final hit list, framing the work as a methodology and reliability study rather than a report of a new inhibitor.
Keap1-Nrf2 is exactly the flat, polar PPI where a co-folding model will happily spit out a confident-looking pose and the affinity scorer will rank it #1 for the wrong reasons — so a paper that pries the cascade apart stage-by-stage is more useful to me than another "we found a hit" claim. What I actually want out of this: a checklist to run on my own vendor or in-house pipeline before I let its top 20 dictate what goes on the synthesis queue. If the co-folding confidence and the affinity score are quietly diverging, that's exactly the failure mode that burns weeks on compounds that never had real signal behind them. No new chemical matter here, no binding mode to chase — just a reminder that for a target this treacherous, the ranked list is not the QC step. Filed under "read before ordering," not "read for the inhibitor."
AI Screening Pipeline Flags Repositioned CASP4 Inhibitor Candidates for Alzheimer's
- CASP4, an inflammatory caspase linked to Alzheimer's disease, is the target of a new computational drug-repositioning pipeline built to screen DrugBank compounds for potential inhibitors.
- The workflow combined docking-based prioritization with ChemBERTa language-model embeddings and physicochemical descriptors to train machine-learning classifiers.
- A Random Forest classifier and regressor were used to identify likely active compounds and rank them by predicted potency.
- Downstream cheminformatics, PK-PD evaluation, and molecular docking narrowed the pool to five repositioning candidates.
- Molecular dynamics simulations and MM/PBSA free-energy analysis were then applied to assess binding stability, identifying two compounds (DB05316 and DB00519) with predicted binding compared favorably to the reference drug donepezil.
- The authors present the pipeline as a general strategy for prioritizing repurposed compounds against novel drug targets.
A five-stage funnel — docking, ChemBERTa embeddings, Random Forest, then MD/MM-PBSA — down to two compounds looks disciplined, but each stage has its own error rate, and stacking filters doesn't guarantee the errors cancel out. MM/PBSA rescoring of structurally unrelated ligands is notoriously unreliable for absolute ranking; chain enough filters and you can end up polishing false positives rather than eliminating them. Where this actually sits in the AI hit-finding wave: not novel chemistry, but faster triage of known-compound space, which is a real and growing niche. The genuine advantage is DrugBank sourcing — DB05316 and DB00519 already carry ADMET, safety, and formulation data, so there's no synthesis or fresh tox package standing between candidate and IND-enabling work. None of that substitutes for a binding assay. Until CASP4 activity is confirmed biochemically, this is a ranked hypothesis list, not a hit.
ChemBERTa-Guided Virtual Screening Identifies Repurposable CASP4 Inhibitors for Alzheimer's Disease
- Ligand-based virtual screening pipeline mined DrugBank for CASP4 inhibitor candidates, combining docking-score prioritization with SMILES-derived ChemBERTa embeddings and physicochemical descriptors.
- A Random Forest classifier trained on the hybrid ChemBERTa/physicochemical feature set separated active from inactive compounds with high accuracy and enriched actives among top-ranked hits.
- A companion Random Forest regressor, trained on experimental pIC50 values, ranked candidates by predicted potency.
- Cheminformatics filtering, PK-PD evaluation, and molecular docking narrowed the field to five DrugBank-listed candidates.
- Molecular dynamics simulations and MM/PBSA free-energy analysis confirmed stable binding poses for the top candidates.
- Positions the workflow as a drug-repositioning route toward CASP4-targeted Alzheimer's disease therapy, a caspase linked to neuroinflammation and pyroptosis.
What actually sells this to me is the DrugBank sourcing, not the screening funnel itself: the five candidates that survive docking, ChemBERTa/RF classification, and MM/PBSA already come with known ADMET and need no de novo synthesis to reach an animal model — that's a tractability shortcut a genuine hit-finding campaign rarely gets. CASP4's pocket is the usual caspase catalytic Cys-His dyad, so the real question the MD trajectories should answer is whether these scaffolds actually engage that dyad or just sit in a peripheral exosite; binding energy alone won't tell me that. Selectivity is the open flag — caspase-1/3/4 share fold homology, and off-target caspase hits in a neuroinflammation context could cut the wrong way. Methodologically, ChemBERTa embeddings layered onto classical descriptors for RF triage is a sane upgrade over pure docking-score ranking, though I'd want held-out chemotype validation before trusting the pIC50 regressor outside its training distribution.
Machine-Learning Screen of 5M Candidates Yields Light-Activated PARP1 Inhibitors
- Combined protein–ligand docking, ML force fields, quantum chemistry, and free-energy perturbation (FEP) to screen roughly 5 million hypothetical photoactive PARP1 ligands.
- Docking identified candidates with differential PARP1 binding between light-exposed and dark states; ML force fields and QM calculations predicted pKa, absorption spectra, and thermal half-lives, while graph-based surrogate models screened the remaining library.
- Excited-state nonadiabatic dynamics simulations (with ML force fields) estimated photoswitching quantum yields; FEP refined binding-affinity predictions for top hits.
- Prioritized a small set of synthetically accessible candidates combining red-shifted absorption (favoring deeper tissue penetration), therapeutically suitable thermal half-lives, and light/dark-selective PARP1 binding.
- Predictions were experimentally validated, demonstrating an end-to-end computational-to-experimental pipeline for designing light-activated (photopharmacological) cancer therapeutics.
Five million virtual candidates funneled down to a handful that actually work is the real story — that pipeline (docking → ML force fields → QM → nonadiabatic dynamics → FEP) is doing what a med chem team's SAR intuition used to do by hand, just faster and across chemical space no bench chemist would ever hand-explore. The binding-mode hypothesis is straightforward: a photoswitch appended to (presumably) a nicotinamide-mimetic PARP1 pharmacophore toggles pocket occupancy between light and dark isomers, so "selectivity" here means spatiotemporal control over tumor tissue, not isoform selectivity against PARP2/tankyrases — a different axis entirely from what olaparib or niraparib were ever optimized for. Red-shifted absorption for tissue penetration plus a therapeutically workable thermal half-life are the tractability bottlenecks that usually kill photopharmacology leads, so getting synthetically accessible hits past both, and into validated wet-lab data, is the genuinely notable part.
Computational Design and Experimental Validation of Photoactive PARP1 Inhibitors ↗
AI Co-Folding Model Boltz-2 Speeds Discovery of Novel AKR1B10 Inhibitors
- AKR1B10 is an aldo-keto reductase enzyme overexpressed in several cancer types, making it an attractive but challenging drug target.
- Researchers used Boltz-2, an AI-based protein-ligand co-folding model, to computationally screen a large, chemically diverse compound library and prioritize a small, diverse shortlist of candidates for lab testing.
- Experimental testing of that shortlist produced an unusually high hit rate, with most of the tested compounds showing real inhibitory activity against AKR1B10.
- Several confirmed hits reached submicromolar potency and carried chemical scaffolds not previously reported as AKR1B10 inhibitors.
- The results support co-folding AI models as a practical, resource-efficient way to convert computational predictions into experimentally validated drug leads, reducing reliance on brute-force screening.
• Most of a small, deliberately diverse shortlist came back with real inhibition. That hit rate isn't HTS-style luck; it's a model that has actually learned something about the pocket, not the kind of docking score I'd normally discount. • Novel scaffolds are the real prize, but I want SAR before I get excited: is potency riding on one privileged fragment, or is there a real vector to optimize? Submicromolar out of the gate on unprecedented chemotypes usually means headroom, not a dead-end analog of known reductase binders. • AKR1B10 sits close to AKR1B1 (aldose reductase) — selectivity is the whole game here; nobody needs another dirty aldo-keto reductase pan-inhibitor with off-target baggage. • This is one target, one campaign. I'd want to see it replicated on something without such a cooperative pocket before I trust co-folding over ensemble docking/FEP as a default triage step. Promising, not proven — but the direction hit-finding is heading.
Efficient Discovery of Potent AKR1B10 Inhibitors Using Co-folding Model Boltz-2 ↗
AI-Guided Path to VAV1-Targeting Molecular Glues
- VAV1 is a hematopoietic-restricted signaling protein of interest in blood cancers and autoimmune conditions, prompting efforts to target it via induced protein degradation.
- High-throughput, unbiased global proteomics screening identified phenyl-glutarimide degrader compounds that recruit CRBN to selectively degrade VAV1.
- VAV1's C-terminal SH3-2 domain proved essential for ternary complex formation, and structural work uncovered a novel "RT-loop" degron there—distinct from the "G-loop" degrons used by other CRBN glues—validated by mutagenesis.
- An AI-based protein structure-prediction workflow modeled the ternary complex to support this mechanistic finding.
- Free energy perturbation calculations on the predicted structures generated cooperativity scores that correlated with experimental potency, demonstrating a computational strategy to prioritize and optimize molecular glue analogs faster than empirical screening alone.
The real headline isn't the molecule, it's the RT-loop degron. G-loop CRBN neosubstrates have basically been the only playbook since the IMiD era, so a structurally distinct degron sitting in VAV1's SH3-2 domain is a genuine expansion of the addressable degradable proteome, not just one more target on the pile. The phenyl-glutarimide warhead itself is bog-standard IMiD-adjacent chemistry — nothing novel there; the real innovation is in the target-engagement biology, not the medchem. VAV1's hematopoietic-restricted expression is a nice built-in selectivity advantage most degrader programs would kill for. On the FEP/AI ternary angle: cooperativity scores tracking potency across one series is promising, but it's one dataset — I'd want to see it prospectively rank blind analogs before I trust it over empirical SAR.
Multimodal AI Framework Targets Drug-Resistant Bacteria While Screening for Toxicity
- Antimicrobial resistance is narrowing treatment options for pathogens like Staphylococcus aureus and Neisseria gonorrhoeae, motivating new computational discovery tools.
- MAPViS, a multimodal deep-learning framework, fuses molecular graph, fingerprint, and descriptor/3D geometric features via gated fusion, pretrained on ~10 million molecules using hierarchical pseudolabel classification and masked graph contrastive learning.
- After fine-tuning, it jointly predicts antibacterial activity against S. aureus and N. gonorrhoeae plus cytotoxicity risk in three human cell lines (HepG2, HSkMC, IMR-90).
- On benchmark and external validation data, MAPViS showed strong early-enrichment performance for S. aureus, with all ten top-ranked candidates confirmed experimentally active, while generalization to N. gonorrhoeae was more modest.
- The authors describe MAPViS as a scalable tool for joint activity-toxicity modeling to support antibiotic virtual screening.
Ligand-based, not structure-based — no target structure in the loop, so I'd worry about applicability domain once you push past its ~10M-molecule pretraining chemotypes into a new target class. The joint activity/cytotox head is the real differentiator: folding HepG2/HSkMC/IMR-90 flags into the same ranked list saves the counterscreen step I'd otherwise run manually — genuine time saved, not just convenience. But the S. aureus (10/10 confirmed) vs. N. gonorrhoeae ("modest") split is the tell: performance tracks training-data density per pathogen, not real antibacterial chemotype intuition. Before trusting a ranked hit list on an unfamiliar target, I'd want per-pathogen calibration and confirmed true negatives, not just top-10 hit rates. Good as a cheap pre-HTS triage filter; not a substitute for structure-based design or real counterscreening.
Reversible-Work Metadynamics Delivers Fast Hit Triage — and Exposes a Memorization Blind Spot in AI Co-Folding
- Virtual screening often yields false positives from fast, approximate scoring functions, while accurate free-energy calculations to filter them remain too costly for routine high-throughput use.
- The authors introduce CTMD (c(t)-based metadynamics), which ranks binding stability using the nonequilibrium reversible-work estimator c(t) from short, independent metadynamics runs, without requiring converged binding free energies.
- Across diverse targets and chemotypes, CTMD shows robust early enrichment while staying fast, transferable, and largely tuning-free.
- Benchmarking AI co-folding, the authors find Boltz-2's enrichment tracks similarity to its training data and persists even after active-site modifications, pointing to memorization rather than true binding physics.
- CTMD is offered as an open-source middle ground between quick docking/co-folding and costly free-energy methods, cutting wasted discovery-campaign resources.
A triage tier between docking and full FEP is exactly what's been missing from my workflow. Standard docking scores are too noisy to prioritize synthesis, and running FEP on every virtual-screening hit is a non-starter for budget and turnaround — CTMD's short, independent metadynamics runs with the c(t) estimator could slot right into that gap and actually change which compounds we send to synthesis, without a cluster's worth of compute. Open-source and largely tuning-free means it's adoptable without a specialist babysitting it. The Boltz-2 finding is the one I'd flag to my team before anyone leans on AI co-folding for real decisions: enrichment tracking training-set similarity even after active-site mutation means it may be pattern-matching known chemotypes, not scoring genuine binding physics. That's a trap for exactly the novel scaffolds we're chasing.
Early-Enrichment Hit Discovery via Reversible-Work c(t) Estimation in Metadynamics (CTMD) ↗