Hanqun Cao, Mutian He, Ning Ma, Chang-Yu Hsieh, Xiaojun Yao, Cesar de la Fuente-Nunez, Chunbin Gu, Pheng Ann Heng

Cell Reports Physical Science

DOI: 10.1016/j.xcrp.2026.103565

Abstract

DNA-encoded library (DEL) screening links small molecules to unique DNA barcodes, enabling massively parallel measurements of compound-protein interactions in which sequencing read counts serve as surrogates for binding affinity. Converting these counts into reliable affinity estimates remains challenging because stochastic Poisson noise dominates low-abundance compounds, while synthesis, selection, and amplification biases can decouple observed enrichment from true thermodynamic binding strength. Here, we present DEL-Ranking, an integrated computational framework that learns stable relative orderings through complementary pairwise and listwise ranking objectives and incorporates a chemical-referenced correction (CRC) module to recalibrate count predictions using structure-activity priors. Across five carbonic anhydrase datasets, DEL-Ranking consistently improves agreement with experimental Ki values, including a 0.304-unit gain in Spearman correlation under high-noise overexpression conditions, and independently prioritizes pyrimidine sulfonamide—a known bioisostere of benzene sulfonamide—without explicit supervision for that scaffold. These results show that coupling ranking-aware denoising with chemically informed correction yields more reliable, interpretable, and validation-ready DEL hit lists.

Mosè Fabbri, Lorenzo Campari, Sara Puglioli, Laura Alvigini, Francesca Migliorini, Samuele Cazzamalli, Dario Neri, Nicholas Favalli

Bioconjugate Chemistry

DOI: 10.1021/acs.bioconjchem.6c00428

Abstract

DNA-encoded chemical libraries (DELs) accelerate ligand discovery by combining combinatorial synthesis with parallel affinity selection of DNA-tagged compounds. Conventional DEL selection protocols involve the immobilization of protein targets on a solid support such as magnetic beads. Several washes are performed on the beads after library incubation to remove non-binder library members and to enrich for binding compounds. However, excessive washing could also deplete weak or rapidly dissociating binders. Here, we systematically study the impact of the number of washing steps on DEL hit enrichment during screening of a newly synthesized library named THPP-DEL and based on a central 2-amino-3-(4-(1,2,3,6-tetrahydropyridin-4-yl)phenyl)propanoic acid scaffold and containing ∼8 million compounds. THPP-DEL was screened against Carbonic Anhydrase IX (CAIX), Prostate-Specific Membrane Antigen (PSMA), and Human Serum Albumin (HSA) by systematically varying the number of washing steps from 0 to 20. High-throughput sequencing revealed enriched chemotypes against the three targets, with affinities ranging from low nanomolar to low micromolar concentration values. For all hits, total counts and enrichment factors depended on washing conditions and stabilized after three washing steps. The results presented in this paper define practical selection conditions that preserve micromolar binders, control background levels, and reduce unnecessary hit resynthesis of false positives.

Paweł Kozyra, Maciej Masłyk

Chemistry - A European Journal

DOI: 10.1002/chem.71733

Abstract

Organic chemistry sits at a pivotal moment. Two decades after combinatorial chemistry promised to industrialize molecule making, the field has converged on a far more interesting reality: medicinal chemists now operate at the intersection of new catalytic paradigms, biological‐scale automation, machine learning, and an expanding set of therapeutic modalities that look very little like the flat aromatic scaffolds of the 1990s. This Perspective surveys the dominant trends shaping contemporary organic synthesis in the context of drug discovery—late‐stage C─H functionalization, photoredox and electrochemical catalysis, biocatalysis, continuous flow manufacturing, AI‐guided retrosynthesis, DNA‐encoded libraries (DELs), and proximity‐based modalities such as PROTACs and molecular glues—and discusses how their convergence is redefining how molecules are designed, made, and validated. The future of the discipline lies not in any single breakthrough, but in the controlled integration of these threads into closed‐loop, data‐rich, and increasingly autonomous discovery pipelines. We also identify open challenges: predictability of complex sp 3 ‐rich scaffolds, scalability of new modalities, environmental footprint, and the need for chemistry education and infrastructure to keep pace with computational advances.

Maria Heinke, Mosè Fabbri, Jörg Scheuermann, Samuele Cazzamalli, Nicholas Favalli, Dario Neri, Sara Puglioli

Chemistry - A European Journal

DOI: 10.1002/chem.71759

Abstract

DNA‐Encoded Library (DEL) screenings require sufficient copies per library member to ensure reliable hit identification and reproducibility across experiments. Here, we screen various primers and template DNA sequences for PCR amplification and next‐generation sequencing (NGS), aiming to lower the required library input while maintaining robust hit discovery rates. Analysis of selection fingerprints using an optimized primer design shows that the minimum input required to identify hits can be reduced to 10 3 copies per member, representing a 100‐fold decrease relative to prior reports. This optimization is particularly impactful for large libraries containing billions of compounds, where increased synthesis costs and larger absolute quantities required for selections necessitate minimizing the amount consumed per selection experiment.

Bingqi Tong, Sunny A. Tang, Yifan Deng, Zhihan Nan, Alison X. Gao, Gregory A. Michaud, Simone Bonazzi, Frédéric Berst, Frédéric J. Zécri, Shuang Liu, William J. Gibson, Stuart L. Schreiber

Advanced Science

DOI: 10.1002/advs.77819

Abstract

Chemical‐induced proximity, particularly via molecular glues, represents a rapidly advancing therapeutic paradigm. Expanding the scope of proximity‐based therapeutics requires generalizable discovery platforms that can be easily tailored to diverse presenter proteins. Here, we report a versatile DNA‐encoded library (DEL) strategy designed to identify highly cooperative chemical inducers of proximity (CIPs) for multiple presenter proteins. By leveraging a novel precursor library constructed via on‐DNA strained allene cycloadditions, we generated three distinct CIP‐DELs biased toward Cereblon (CRBN), VHL, and FKBP12. Screening the CRBN‐focused library with BRD9 identified B67b , a novel compound that mediates ternary complex formation between CRBN and BRD9 with nanomolar potency and strong molecular glue‐like cooperativity (α >300). B67b engages BRD9 in a CRBN‐dependent manner, a characteristic of classical molecular glues, and its high cooperativity is driven by the synergistic contribution of all its structural and stereochemical components. This work establishes a scalable, generalizable framework for repurposing generic DELs into presenter‐specific screening tools, offering a powerful approach to accelerate the discovery of CIP therapeutics for a broad range of targets.

Marissa Dolorfino, Daniel Santos Perez, Yao Fu, Shu-Hang Lin, Sean McCarty, Matthew J. O’Meara, Terra Sztain

Journal of Chemical Information and Modeling

DOI: 10.1021/acs.jcim.6c01611

Abstract

Predicting protein–ligand binding is a central challenge in computational drug discovery, and while machine learning (ML) and cofolding methods have advanced rapidly, their ability to generalize beyond training or parametrization regimes remains insufficiently understood. DNA-encoded libraries (DELs) enable ultralarge screening of billions of molecules simultaneously, providing a useful testbed for evaluating these approaches at scale. A recent NeurIPS competition revealed that even top-performing ML models trained on DEL data failed at generalizing to out-of-distribution (OOD) chemical space. We investigated whether integrating structural modeling could bridge this generalization gap. We systematically assessed state-of-the-art ML, docking, and cofolding methods, including Schrödinger Glide, Rosetta GALigandDock, and Boltz-2 with three biologically diverse protein targets screened against libraries containing multiple DEL synthesis formats. While ML excels in-distribution, OOD hit discrimination is dependent on both the target and ligand context, with no single method consistently dominating. These findings demonstrate that benchmark performance alone is insufficient to predict OOD performance, highlighting the need for system-dependent evaluation of binding prediction methods. We provide an open-source package for assessing protein–ligand prediction methods and analyzing high-throughput screening data: DEL-iver.