DNA-encoded libraries
7 articles
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.
In silico building-block selection widens the diversity of a 403,200-member peptide-like DEL
- A three-cycle, peptide-like DEL of 403,200 compounds was built, with a chemical space model of 4,774 commercial amino acids used to find under-represented building blocks.
- Affinity selections were run against three targets to compare library performance.
- Selection against carbonic anhydrase II recovered the known sulfonamide pharmacophore, supporting the computational design approach.
Recovering the sulfonamide on CA II is the expected sanity check, not proof of lead-like coverage. Peptide-like DELs often return big, polar hits, so what I'd judge is how many selected compounds sit in usable property space off-DNA, and how the two other targets behaved.
DEL hit becomes brain-penetrant reversible MAGL inhibitor MAGLi 432
- A bicyclopiperazine hit from a DNA-encoded library was optimized by structure-guided design into compound 1.5 (MAGLi 432), a reversible MAGL inhibitor.
- The compound combines high potency, selectivity over other serine hydrolases and CNS drug-like properties, with a favorable oral half-life in rats.
- It inhibited MAGL in human brain homogenate and worked in mouse neuroinflammation models; the campaign was steered by a translational "Five-Star Matrix" framework.
Reversibility is the part I'd want here, since irreversible MAGL agents ran into CB1 desensitization on chronic dosing. A non-covalent DEL chemotype that already shows target engagement in human brain tissue is a good template, and I'd steal the habit of testing in human tissue early rather than at the end.
Two-Step Route Expands Warhead Diversity in Covalent Genetically-Encoded Macrocycle Libraries
- A two-step route builds covalent genetically-encoded macrocycle libraries (cGELs): cysteines form the macrocycle via a diketone linchpin, then a separate Knorr-pyrazole step adds the electrophilic warhead.
- Separating cyclization (basic pH) from warhead installation (acidic pH) avoids the regioisomer mixtures that limited earlier one-pot chemical modification methods.
- This lets more diverse, reactive electrophiles be used safely, broadening covalent chemotypes accessible for selections against protein targets in DEL-style screens.
Regioisomer scrambling in one-pot warhead installation is exactly what kills confidence when you're deconvoluting covalent DEL hits — you can't tell if a signal is real or an artifact of a messy electrophile mixture. Decoupling cyclization from Knorr-pyrazole warhead chemistry means finally screening exotic, more reactive electrophiles beyond the usual acrylamide/chloroacetamide workhorses without second-guessing every hit's identity.
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.
A SuFEx Click Reaction Streamlines DNA-Encoded Macrocyclic Peptide Libraries
- DNA-encoded libraries (DELs) let researchers screen ultralarge collections of macrocyclic peptides to find hit compounds for drug discovery, but efficient on-DNA cyclization chemistry has remained a bottleneck.
- A new on-DNA macrocyclization method uses sulfur(VI) fluoride exchange (SuFEx), a fast, click-type reaction, to close peptide rings directly on DNA tags.
- The design embeds a phenol and an aryl sulfonyl fluoride within a single DNA-tagged peptide, positioning the two groups to react with each other intramolecularly.
- Simply dissolving the construct in basic aqueous buffer triggers rapid, high-conversion cyclization without added catalysts or harsh conditions.
- The method enables efficient, high-yielding synthesis of ultralarge DNA-encoded macrocyclic peptide libraries, expanding the chemical space available for DEL-based hit discovery.
On-DNA chemistry is where good click reactions go to die — CuAAC needs copper that chews up your barcode, most coupling chemistry leaves truncation side-products you chase off HPLC for weeks. SuFEx actually respects the constraints: no metal, no coupling reagent, just a phenol and an ArSO2F sitting in basic water reacting with each other. That's "click" chemistry finally earning its name in DEL work instead of just borrowing the marketing. High conversion with zero catalyst also means boring QC, in the best sense — fewer truncated/uncyclized side-populations polluting the pool means the hits you pull downstream are more likely real signal, not artifacts of your worst-yielding step. For macrocycles — the space between small molecules and biologics everyone wants but few can screen efficiently — a clean cyclization step is what lets you scale diversity without sacrificing quality. This is the unglamorous methods advance that keeps unlocking new DEL generations.
SuFEx Cyclization Enables DNA-Encoded Macrocyclic Peptide Libraries for Drug Discovery ↗
Bicyclic Peptide Libraries Deliver First Noncarbohydrate Inhibitors of the Hard-to-Drug Receptor Siglec-7
- Researchers used bicyclic genetically encoded peptide libraries (BiGEL) — phage-displayed peptides chemically stapled into two rings — to search for inhibitors of Siglec-7, a glycan-binding immune receptor considered notoriously hard to target with small molecules.
- Sequencing-based panning against Siglec-7 nominated hundreds of candidate peptides; a subset were confirmed as genuine binders by surface plasmon resonance and as functional disruptors of the Siglec-7:GD3 interaction by competitive ELISA, with machine learning models identifying further hits of similar potency.
- Structure-activity work showed the bicyclic ring architecture itself is essential for potency, and NMR studies located one optimized lead's binding site away from the canonical glycan-binding domain, pointing to a possible allosteric mechanism.
- Cell-based multivalent binding assays confirmed engagement with Siglec-7 but also revealed considerable nonspecific binding to receptor-negative cells, highlighting a selectivity challenge still to be resolved.
The allosteric site is the real hook here — a bicyclic peptide binding away from the canonical glycan pocket on a target everyone assumed only the carbohydrate-recognition domain was druggable gives you a completely different SAR vector than trying to out-compete sialic acid for its own pocket. That's genuinely exciting for Siglec-7. But I'd want the nonspecific binding in the cell assays resolved before getting attached to any single lead — multivalent avidity effects can paper over a mediocre core affinity, and that's exactly the kind of artifact that kills a program downstream. Bicyclics themselves are the right chemotype to bet on: synthesizable, chemically stable enough for iterative SAR, and unlike DEL small molecules they're not fighting the flat, shallow topology that's stalled every prior GBP inhibitor effort.