The Current Toolbox for Covalent Inhibitors: From Hit Identification to Drug Discovery

Abstract

Covalent modification of therapeutic targets has emerged as a powerful platform for creating clinical drugs and chemical probes. Covalent drugs have evolved from serendipitous discoveries to rationally designed therapeutics, driven by advances in electrophile-first screening technologies. This perspective takes stock of alternative technologies currently available in laboratories and industry that collectively enable targeted covalent inhibitor development across historically “undruggable” targets. We highlight five such technologies: activity-based protein profiling (ABPP), provides functional proteomic mapping to identify ligandable residues; covalent tethering, exploits dynamic chemistry to capture transient pockets; covalent DNA-encoded libraries, leverages trillion-member libraries for multiresidue targeting; phage/mRNA display, which facilitates evolution of covalent macrocyclic peptides; and sulfur­(VI) fluoride exchange (SuFEx), engages residues beyond cysteine. Integration of these approaches with chemoproteomics and artificial intelligence accelerates the discovery of covalent inhibitors with enhanced selectivity and reduced off-target risks. This technological convergence establishes a new paradigm for precision covalent therapeutics, offering innovative solutions to overcome drug resistance and target challenging protein interfaces.

1. Introduction

Covalent drugs are pharmacologically active compounds that form covalent bonds with their biological targets. In most cases, a covalent drug consists of a noncovalent backbone and a warhead. , The noncovalent backbone binds to the target protein via reversible interactions, such as hydrogen bonding and hydrophobic effects, providing initial affinity and selectivity. Warheads are chemically reactive functionalities (e.g., acrylamides, α-halo ketones) that form reversible or irreversible covalent bonds with nucleophilic residues (e.g., cysteine, lysine) in enzymes or receptors. , Precisely tuning the features of both the warhead and noncovalent backbone depending on the structural specificity requirements of the target protein, maximal inhibition can be achieved while minimizing toxicity. Consequently, the discovery of covalent inhibitors has evolved from serendipitous findings to rational design. Pharmaceutical researchers rationally develop covalent inhibitors through strategic installation of warheads onto optimized noncovalent scaffolds with demonstrated target binding affinity. Such compounds are termed “targeted covalent inhibitors” (TCIs), exhibiting exceptional target selectivity. −

The superior potency of covalent inhibitors compared to their noncovalent counterparts stems from their unique mechanism of covalent bond formation. Covalent inhibitors exhibit prolonged target residence time due to their capacity for sustained target engagement with enzymatic/receptor binding sites. By forming nonequilibrium covalent bonds with target proteins, covalent drugs effectively circumvent competition from endogenous substrates. This unique mechanism establishes them as a pivotal technological platform for: (i) overcoming drug resistance through irreversible target engagement, (ii) accessing previously undruggable sites via warhead-mediated covalent trapping, and (iii) achieving prolonged pharmacodynamic activity. − Recent advances in understanding covalent inhibitor mechanisms and exploring electrophilic warheads for nucleophilic amino acids have led to significant progress in covalent drug development. At least 50 covalent inhibitors have been approved for clinical use across diverse diseases , with prominent examples including the BTK inhibitor Ibrutinibwhich covalently targets Cys481 in Bruton’s tyrosine kinase for B-cell malignanciesthe third-generation EGFR inhibitor Osimertinib that irreversibly binds Cys797 in EGFR T790 M mutant nonsmall cell lung cancer, , and the KRAS ^G12C inhibitor Sotorasib forming a covalent bond with mutant Cys12 in KRAS-driven solid tumors. The clinical success of these agents validates covalent strategies for achieving sustained target inhibition, enhanced selectivity, and therapeutic efficacy against resistance mechanisms.

Covalent inhibitors have emerged as a transformative therapeutic class, offering unique solutions to longstanding challenges in drug discovery: overcoming acquired resistance, targeting “undruggable” sites, and extending pharmacodynamic activity. Despite these advantages, significant limitations impede their clinical translation. Off-target effectsprimarily from unintended reactions between electrophilic warheads and nucleophilic residues (e.g., glutathione)compromise therapeutic efficacy through drug depletion, hepatotoxicity risks, and dosage prediction complexities. The irreversible nature of traditional covalent binding further introduces safety concerns, as permanent target modification may cause irreversible cellular damage without clinical recourse, narrowing therapeutic windows. Additionally, the inherent complexity of co-optimizing warhead reactivity, selectivity, and drug-like properties is a fundamental reason why the development of covalent drugs remains a lengthy process, often spanning over ten years. Under such circumstances, the application of novel covalent ligand discovery technologies to expedite drug development has emerged as a key scientific challenge.

Numerous strategies for linking desired covalent probes with target proteins have been developed in the past decades (Figure ), with most exploiting chemical biology and/or bio-orthogonal chemistry that utilizes covalent ligand discovery. This proposal synthesizes current knowledge while providing novel perspectives on mechanistic innovations in covalent warhead chemistry, technological convergence of structural biology and chemoproteomics, translational frameworks for toxicity minimization, and emergent discovery platforms. Critically, we provide a comprehensive analysis of integrated covalent drug design methodologiesfrom warhead optimization to off-target risk mitigationthat collectively address the historical challenges of covalent inhibitor development, offering actionable strategies for accelerating the discovery of safer covalent therapeutics.

1.

Brief timeline showing the coevolution of covalent drug discovery and electrophile-first screening technologies.

2. Structure-Based Design of Covalent Inhibitors

The early discovery of covalent drugs was largely serendipitous, whereas contemporary development relies on structure-based rational design. Co-crystal structures determined by X-ray crystallography are central to this strategy: by revealing the atomic binding mode between inhibitor and target protein, they enable focused iterative optimization to improve potency while minimizing off-target liabilities.

With the continuous advancement in targeted therapy for EGFR-mutated nonsmall cell lung cancer (NSCLC), the design of covalent inhibitors has become a critical strategy to overcome acquired resistance. Early reversible EGFR inhibitors, such as Gefitinib, competitively and reversibly bind to the ATP-binding site, effectively inhibiting EGFR kinase activity. These agents were used to treat advanced NSCLC patients with sensitive mutations (e.g., exon 19 deletions or L858R mutation). However, the T790 M resistance mutation frequently emerges during treatment. This mutation induces steric hindrance within the ATP-binding pocket, reduces the binding affinity of first-generation drugs like Gefitinib, and enhances ATP binding affinity, thereby leading to drug resistance and reactivation of downstream signaling pathways.

To address this challenge, researchers employed rational drug design based on the core structure of Gefitinib by introducing covalent reactive functional groups, leading to the development of irreversible inhibitors. Afatinib, as the first second-generation covalent EGFR inhibitor approved by the U.S. FDA, was developed based on this strategy. Building on the quinazoline scaffold of Gefitinib, an acrylamide group was introduced at the terminal end of the C6 side chain as an electrophilic warhead. This warhead specifically targets the conserved Cys797 residue in the kinase domain of EGFR, forming an irreversible covalent bond via Michael addition, resulting in sustained inhibition of T790 M mutant EGFR (Figure A). The successful development of second-generation covalent inhibitors like Afatinib provided a clinical solution to T790M-mediated resistance. More importantly, it demonstrated the feasibility of rational structure-based design for covalent inhibitors, laying a solid foundation for subsequent drug development. Driven directly by this concept, the third-generation EGFR inhibitor Osimertinib was successfully developed. Through optimization of the warhead structure (using an acrylamide-based warhead) and scaffold design, Osimertinib efficiently and selectively forms a covalent bond with Cys797, effectively inhibits T790 M mutant EGFR, and exhibits reduced activity against wild-type EGFR, resulting in improved safety and efficacy. The development of Afatinib established a structure-based paradigm for covalent drug design, profoundly influencing drug development strategies in the field of targeted therapy. , Inspired by the success of covalent inhibition strategies in overcoming EGFR resistanceas exemplified by agents such as Osimertinib and Afatinibour team has applied this rational drug design principle to develop novel covalent FGFR inhibitors. In 2019, our group identified the FGFR lead compound 9 through virtual screening, and X-ray cocrystal structure analysis (PDB: 6ITJ) revealed that compound 9 acts as a reversible FGFR inhibitor, effectively binding to FGFR1 by forming hydrogen bonds with hinge region residues Ala564 and Glu562 (Figure B), while the benzofuran moiety extends into the hydrophobic pocket. To further enhance kinase inhibitory potency and antiproliferative activity, we employed a structure-based design strategy by introducing an acrylamide warhead at the phenyl group in compound 9, targeting the conserved cysteine Cys488 in the FGFR1 P-loop. Subsequent optimization of the warhead and the 5-position substituent on the benzofuran moiety led to the development of compound 15, which exhibits a more balanced profile of potency and pharmacokinetic properties (FGFR1 IC₅₀ = 20.4 nM; KG1 cell IC₅₀ = 3.7 nM). To improve the solubility and oral PK profiles, a series of novel covalent FGFR inhibitors were developed through molecular docking-guided covalent warhead screening and structural optimization. , Among them, compound 2e demonstrated potent inhibitory activity, with IC₅₀ values of 24.1, 13.6 and 52.6 nM against FGFR1, FGFR2, and FGFR3 wild-type kinases, respectively, and maintained high efficacy against the FGFR2-N549H resistant mutant (IC₅₀ = 16 nM). Structurally, the replacement of the phenyl linker in compound 15 with a flexible piperidine moiety in 2e enhanced conformational flexibility, which likely reduced crystal packing density and contributed to improved aqueous solubility (50 μg/mL in PBS pH 7.4). This increased solubility is expected to facilitate intestinal absorption and may also reduce first-pass metabolism, together with improved PK profiles. In FGFR1-amplified H1581, FGFR2-amplified NCI-H716, and SNU-16 tumor xenograft models, oral administration of 2e exhibited significant antitumor efficacy, inducing tumor stasis or even regression.

2.

Structure-based covalent drug design: (A) Crystal structure of the mutated EGFR kinase domain shows Afatinib (yellow) (PDB: 3UG2) covalently bound to Cys797, whereas Gefitinib (pink) (PDB: 4G5P) binds via a noncovalent manner; (B) Crystal structure of FGFR1 with hit compound 9 (green) (PDB: 6ITJ) identifies Cys488 as the key residue for covalent bond formation; further optimization yielded compounds 15 and 2e with markedly improved FGFR1 enzymatic and cellular potency; (C) Crystal structure of SARS-CoV-2 3CLpro with Bofutrelvir (PDB: 6LZE) shows the inhibitor covalently bound to catalytic Cys145 and hydrogen-bonded to Phe140, His164, His165 and Glu166; Depicted peptidomimetics (14a, 28f) and novel warheads serve as templates for covalent antiviral drug development.

Meanwhile, covalent inhibitors play essential roles not only in cancer therapy but also in almost all major disease areas, such as viral infections. Our research team has long been dedicated to the development of antiviral drugs targeting the 3-chymotrypsin-like protease (3CL^pro). During our research, we discovered that the monomer of SARS-CoV 3CL^pro exhibits high structural similarity to the 3CL^pro of viruses such as CoxB3, HRV, and EV71. Based on this finding, we utilized AG7088a classic 3C^pro inhibitor of human rhinovirus (HRV)as the core scaffold to develop broad-spectrum inhibitors targeting both enteroviruses and coronaviruses. A specific strategy involved replacing the hydrolytically labile unsaturated ethyl ester moiety of AG7088 with the α-ketoamide warhead of Telaprevir (Figure C). A series of ketoamide derivatives was designed, and among which compound 11r demonstrated broad-spectrum antiviral activity, exhibiting inhibitory effects against multiple viruses (SARS-CoV EC₅₀ = 2.1 μM; MERS-CoV EC₅₀ = 5.0 μM; EV71 EC₅₀ = 3.7 μM). In 2020, we further optimized the core skeleton of 11r through structure-based drug design, taking into account the characteristics of the substrate-binding pocket of SARS-CoV 3CL^pro, and successfully obtained a powerful aldehyde-based SARS-CoV-2 3CL^pro inhibitor Bofutrelvir (FB2001). Bofutrelvir achieves highly efficient inhibition by forming a covalent bond between its aldehyde warhead and the catalytic residue Cys145. It features a classic and stable glutamine mimican (S)-γ-lactam ringat the P1 position to effectively occupy the S1 site of 3CL^pro. At the P2 position, Bofutrelvir incorporates a cyclohexyl group that fills the typically spacious S2 site. Additionally, Bofutrelvir forms a critical hydrogen bond interaction with the S4 site via its indole moiety (Figure C). This rational design not only enhanced inhibitory potency but was also expected to improve its drug-like properties. Experimental results demonstrated that Bofutrelvir is a potent 3CL^pro inhibitor with an IC₅₀ of 53 nM and exhibited good anti-SARS-CoV-2 activity in Vero E6 cells (EC₅₀ = 0.53 μM). Now, Bofutrelvir is currently in a Phase II/III clinical trials for COVID-19. Based on the cocrystal structure of Bofutrelvir in complex with SARS-CoV-2 3CL^pro (PDB: 6LZE), our team has carried out a series of structure-guided rational drug design efforts for the development of more useful covalent ligands. By introducing a rigid octahydroindole structure to optimize the pharmacokinetic properties of the molecule, we successfully developed the peptidomimetic inhibitor 28f (SARS-CoV-2 3CL^pro IC₅₀ = 160 nM). Structurally, this modification bridges the cyclohexyl group and the adjacent amide nitrogen via a carbon linkage, forming a conformationally constrained bicyclic system. This redesign eliminates one labile amide bond, thereby reducing susceptibility to proteolytic cleavage and contributing to improved oral exposure and an extended half-life. Meanwhile, further modification of the peptidomimetic scaffold led to the discovery of a class of multitarget compounds capable of simultaneously and potently inhibiting host Cathepsin L (CTSL), Calpain-1 (CAPN1), and viral SARS-CoV-2 3CL^pro. The representative molecule 14a demonstrates exceptional broad-spectrum antiviral activity against SARS-CoV-2 and various Variants of Concern (VOCs), with EC₅₀ values ranging from 0.80 to 161.7 nM across multiple cell-based assays. In parallel, we advanced the development of nonpeptidomimetic small molecule covalent inhibitors: based on the structural features of the target, we discovered and validated the thiocyano group as a novel covalent warhead that specifically targets the catalytic residue Cys145 of 3CL^pro. Structure-based optimization yielded the active compound 3c (IC₅₀ = 322 nM). Meanwhile, we identified [1,2,4]-thiadiazole as a novel scaffold which covalently binds to the catalytic cysteine 145 via a ring-opening metathesis reaction mechanism. The compound 6g, designed based on this scaffold, displays excellent inhibitory activity (IC₅₀ = 118 nM).

In brief, structure-based covalent design has served as the central engine driving the advancement of this field. However, the efficiency of this strategy is constrained by the availability of high-quality crystal structures. To overcome this bottleneck, the discovery process has been moved upstream, leading to the emergence of a series of “electrophile-first” high-throughput screening technologies that do not rely on prior structural information. These approaches have significantly expanded the scope of targetable protein space for covalent drugs.

3. Covalent Drug Screening Strategies Based on “Electrophilic Preference”

In recent years, there are increasing interests in the development of new technologies for covalent ligands discovery in various diseases beyond structure-based covalent design. Target discovery and lead generation technologiescovalent DEL, tethering, and mRNA displaycollectively address the challenges of validating undruggable targets and designing effective molecules. Covalent DEL enables high-throughput hit screening through DNA-encoded libraries of billions of compounds (e.g., incorporating acrylamide warheads), exemplified by the discovery of EGFR T790M/L858R inhibitors. Covalent tethering employs dynamic covalent bonds (e.g., imine bonds targeting lysine, disulfide bonds targeting cysteine) to resolve structural modes of weakly bound fragments, guiding rational optimization. mRNA display introduces electrophilic warheads into gene-encoded peptide libraries to identify high-affinity covalent peptide binders, providing scaffolds for small-molecule design. Selectivity assessment and off-target analysis technologiesABPP and SuFExaddress the safety challenges of covalent drugs. ABPP employs broad-spectrum probes (e.g., acyl-phosphates) to globally label cysteine/serine hydrolases, quantifying kinase-wide off-target risks. SuFEx utilizes arylfluorosulfate warheads to specifically modify tyrosine/lysine residues, covering noncysteine covalent modification blind spots. Taken together, they enable comprehensive safety profiling. With an eye toward expanding chemistry used for covalent drug design, this short perspective highlights recent advancements in the discovery technologies of novel covalent ligands with limited the discussion to key aspects.

3.1. Covalent Tethering Approaches

Conventional high-throughput screening (HTS) critically suffers from millimolar-affinity fragment detection limits, hindering drug discovery for intractable targets. Addressing this gap, Erlanson’s team pioneered disulfide tethering technology through thymidylate synthase (TS) studies. , This innovative technique leverages a thermodynamically balanced reversible capture system: researchers incubate a library of disulfide-containing fragments (average molecular weight ≈ 250 Da) with target proteins featuring natural or engineered surface cysteines in the presence of 1 μM β-mercaptoethanol (BME). Within this dynamic equilibrium, fragments specifically binding near the cysteine residue undergo disulfide exchange to form stable covalent complexes, while nonspecifically bound fragments are efficiently displaced by the excess BMEeffectively eliminating false positives. Following mass spectrometry identification of these complexes, structure-guided optimization enables dramatic affinity enhancements. The TS case exemplifies this workflow: starting from the weakly binding fragment N-tosyl-d-proline (K _i = 1.1 mM), researchers achieved a 50-fold affinity improvement through glutamic acid moiety incorporation, ultimately reaching nanomolar potency (K _i = 330 nM; 3000-fold gain) by further introducing a β-alanine linker. The approach establishes a versatile platform for evolving covalent inhibitors.

Covalent tethering technology has transformed covalent drug development from “accidental discovery” to “rational design”. By enabling targeted covalent modification of noncatalytic cysteines, this approach provides a universal solution for traditionally undruggable targets that lack classical binding pockets.

The technology of covalent tethering is a major breakthrough in the development of nondruggable targets. The key breakthrough is to realize covalent inhibition of metastable sites for the first time, breaking the bottleneck of KRAS “non-druggable”. KRAS has long been held undruggable due to two major reasons. First, its nucleotide-binding pocket demonstrates picomolar affinity for GTP/GDP approximately 1000-fold stronger than typical kinase-inhibitor interactions. Second, the protein’s globular structure lacks conventional druggable pockets beyond the high-occupancy nucleotide site. These properties, combined with millimolar intracellular GTP concentrations, made traditional competitive inhibition strategies ineffective. Researchers discovered that the KRAS^G12C mutant adopts a unique GDP-bound conformation, exposing a transient cryptic pocket (S-IIP) in the switch II region. Using a disulfide-based covalent tethering strategy, they screened a fragment library via thiol-disulfide exchange with Cys12, identifying 6H05 as the lead compound (94% modification efficiency). Co-crystallography confirmed 6H05 targets the S-IIP through Switch-II helix rearrangement. Subsequently, the disulfide tether was replaced with irreversible electrophiles (e.g., acrylamide or vinyl sulfone) to optimize covalent engagement. The resulting inhibitor 12 reached 100% KRAS^G12C modification at 10 μM (vs 6H05: 94% at 100 μM) and exhibited submicromolar potency in H1792 cells (EC₅₀ = 0.32 μM in H1792 cells). This proved that mutant KRAS was not only druggable but also could be targeted with mutant selectivity, thereby minimizing potential toxicity to healthy cells expressing wild-type KRAS. Subsequent chemical optimization of 12 yielded ARS-853, a compound with significantly improved cellular potency in vitro. It selectively binds to KRAS^G12C-GDP (Figure ), where it exerts dual suppression mechanisms: disrupting Switch-I and Switch-II conformations to reduce GTP affinity relative to GDP, trapping Ras in its inactive state, and displacing Gly60 to block the binding interfaces for Raf effectors and the SOS regulatory factor. This foundational research directly enabled Amgen’s development of Sotorasib (AMG 510), which was the first FDA-approved KRAS^G12C covalent inhibitor for NSCLC in 2021. It was followed by Adagrasib (MRTX849), which received accelerated approval for the same indication in 2022. However, this landmark success, also highlights a characteristic aspect of the tethering approach: the initial hits (e.g., 6H05) are often identified primarily for their efficient covalent engagement of the target residue, which does not necessarily equate to high noncovalent binding affinity. The journey from 6H05 to ARS-853 and ultimately to the clinical drugs involved extensive optimization to build upon the initial covalent linkage and develop high affinity, underscoring that tethering provides a powerful starting point rather than a final product.

3.

Disulfide-tethering platform for covalent fragment screening. (A) Disulfide-fragment library architecture: a variable fragment motif is appended to a two-carbon disulfide linker (R = H or Me) and terminated with a soluble cap. (B) Disulfide-tethering workflow: the target protein is incubated with disulfide-bearing fragments; reversible disulfide exchange allows covalent but reversible attachment. Structural modification is quantified by intact-protein mass spectrometry; the extent of modification guides fragment prioritization. (C) Application to KRAS^G12C. A focused library of 480 disulfide fragments was screened against KRAS^G12C. Primary hit 6H05 was optimized to compound 6 and further elaborated to ARS-853, the first direct KRAS^G12C covalent inhibitor shown to abolish p-ERK signaling in cells. The cocrystal structure of ARS-853–KRAS^G12C (PDB: 5F2E) reveals the covalent bond to Cys12 and key interactions within the S-IIP pocket. (D) Chemical structures of JS30, 2A2, IF8 and Fragment 28.

As covalent tethering technology evolves, innovative screening methods continue to emerge. In the BFL-1 case study, researchers designed a peptide chain (derived from BimBH3) that binds with high affinity to BFL-1’s hydrophobic groove via its BH3 domain. The sulfonium center within the peptide acts as a warhead, undergoing nucleophilic substitution with Cys55 upon binding to form an irreversible covalent bond. Crucially, the sulfonium warhead activates only near the target site, mitigating off-target effects. This approach fuses a covalent warhead with peptide stabilization to target the protein–protein interface, addressing BFL-1’s historically undruggable nature.

Covalent tethering serves not only as a screening tool for covalent lead compounds but also demonstrates unique value in elucidating protein dynamic functions. Walensky’s team employed this technology to identify CBI1, a small molecule that reversibly targets BAX C126 residue. CBI1 inhibits BAX via competitive blockade of lipidation and stabilization of the α1-α2 loop conformation, while mitochondrial lipid trans-2-hexadecenal sensitizes BAX activation through C126 lipidation. Wells’ team pioneered a disulfide tethering strategy by engineering an L159C mutation at the allosteric site adjacent to PDK1’s PIF pocket, establishing a covalent molecular anchor for targeted screening. Systematic evaluation of 480 disulfide-containing fragments identified two functionally opposed modulators: activator 2A2 (and its optimized analog JS30) and inhibitor 1F8 (Figure ). Structural analyses revealed that JS30 stabilizes the active α C-helix in conformation, enhancing kinase activity by facilitating substrate binding, whereas 1F8 induces α C-helix displacement, disrupting substrate recognition and suppressing catalytic function. This work demonstrated that diametrically opposed functional outcomesactivation versus inhibitioncan be achieved through distinct conformational changes at a single allosteric locus. The authors proposed that these covalently tethered fragments could serve as starting points for chemical optimization, where elaboration into adjacent subpockets would enable the development of highly potent and selective modulators. Collectively, disulfide tethering transcends traditional covalent inhibitor design by shifting protein research from static structural analysis to dynamic functional control, establishing a paradigm for precision conformation-targeted therapeutics.

Traditional disulfide tethering enables reversible control of single-protein conformations but suffers from limitations including genetic engineering requirements (e.g., cysteine mutations like PDK1-L159C) and residue selectivity constraints. Ottmann’s team pioneered imine tethering targeting natural lysine residues to overcome these barriers. This approach leverages dynamic covalent chemistryaldehyde fragments spontaneously form physiologically reversible imines with interfacial lysines (e.g., Lys122 in 14-3-3). The complex microenvironment enables site-specific binding through a “templating effect” (illustrated by the p65-Ile46 hydrophobic pocket), directly stabilizing PPI networks.

This technology demonstrates exceptional potential for selective hub protein stabilization. In the 14-3-3/Pin1 system, fragment 28 (a 2,4-difluorophenyl derivative) was identified via aldehyde library screening. It triggers a conformational flip in Pin1’s Trp⁺¹, forming π–π stacking and activating an Asn42-Gln⁺³ hydrogen-bond network. This achieves 96.8-fold stabilization gain with ultraselectivity (no effect on ERα/p53 complexes). The strategy provides a universal platform for historically undruggable PPI targets like Keap1/Nrf2 and MDM2/p53. Ultimately, imine tethering shifts covalent drug design from “covalent inhibition” to “covalent construction”, reshaping development paradigms for molecular glues and allosteric therapeutics.

Currently, covalent tethering technology significantly accelerates lead compound discovery and the identification of allosteric pocket functions through weak-binding interaction screening. This provides strong support for the subsequent optimization of covalent drugs. Future development will focus on two main objectives: developing new, more biocompatible dynamic covalent chemistry to enhance the technology’s versatility and safety; and overcoming the limitations of targetable residues by expanding applications to more types of natural amino acids (such as tyrosine, lysine, and histidine), thereby covering a broader proteome. It is important to note, however, that the initial hits from tethering screens are often prioritized for their efficient covalent reactivity, which may not inherently confer high noncovalent binding affinity, thus necessitating significant subsequent optimization. Furthermore, by integrating omics-driven virtual screening and data analysis, this technology holds promise for offering comprehensive references and powerful design guidance to address the challenge of “undruggable targets.”

3.2. ABPP Approaches

To address off-target toxicity in covalent drugs, chemical biology techniques continuously evolve to reveal the functional mechanisms of “undruggable” targets. Among these innovations, activity-based protein profiling (ABPP) emerged from the postgenomic era’s urgent need to dynamically analyze protein function. Traditional proteomics methods (like 2D electrophoresis and mass spectrometry) provide only static abundance information. Crucially, numerous studies demonstrate a frequent decoupling between enzymatic activity and mRNA/protein expression levels (e.g., cysteine proteases require zymogen activation; serine hydrolases often complex with endogenous inhibitors). To directly capture the functional, active-state conformation of proteins under physiological conditions, the teams of Cravatt and Bogyo independently pioneered the novel concept of “activity-directed chemical probes” around 1999–2000. Cravatt’s team designed the first broad-spectrum serine hydrolase probe FP-biotin (Figure ). The fluorophosphonate warhead of FP-biotin covalently modifies the catalytic serine residue, locking the active site. An eleven-carbon alkyl linker connects to a biotin tag, enabling ultrasensitive detection (sensitivity down to 0.35 pmol). This probe’s activity-dependent nature was validated through three key experiments: labeling significantly decreased with heat-denatured proteins, the probe failed to label a catalytic serine mutant (e.g., FAAH-S241A), and labeling intensity drastically dropped for a trypsin-inhibitor complex (STI-trypsin). Kinetic analysis further revealed stratified real-time reaction rates among hydrolases in testis extracts. This work established ABPP as a transformative tool for functional proteomics and provided a critical methodological foundation for covalent inhibitor screening. Concurrently, Bogyo’s team developed the first cysteine proteases covalent probe DCG-04 based on the natural inhibitor E-64, and its epoxysuccinate warhead specifically targets the nucleophilic cysteine residue in the active site (Figure ). A dual-mode detection system (radioiodination/biotin) enabled synergistic gel visualization and affinity purification. Its breakthrough was the construction of a P2 positional scanning library containing 19 natural amino acid variants via solid-phase synthesis, creating a pioneering “programmable selectivity” probe platform. These two foundational studies jointly established the core principle of ABPP: activity-dependent labelingprobes bind only targets in their functional conformation. This principle fundamentally shifted the paradigm for covalent drug discovery: At the target discovery level, ABPP reveals allosteric sites invisible to traditional methods. For safety optimization, competitive ABPP quantifies inhibitor off-target effects. Ultimately, this guides the rational design of allosteric covalent drugs, opening new pathways to tackle “undruggable” targets.

4.

(A) Composition of ABPP Probes. (B) General approach for ABPP. (C) Strategies for discovering covalent inhibitors with ABPP. Inhibitor-based probe: Converts an inhibitor into a covalent probe to label and identify its direct binding targets in cells or lysates. Residue-based probes: Uses a broad-spectrum cysteine probe (e.g., IAA-alkyne) to prelabel the proteome; inhibitor occupancy blocks labeling, enabling residue-specific target identification and high-throughput screening.

While ABPP technology has established a revolutionary pathway for revealing protein function, translating this discovery power into an efficient drug discovery platform necessitates overcoming key challenges: target identification and validation, compound selectivity optimization, and breakthroughs in drugging challenging targets, particularly via allosteric site identification. This perspective focuses on applying ABPP to address these critical challenges, demonstrating its utility through specific case studies to advance drug discovery strategies.

Many targets in the human proteome (e.g., transcription factors, scaffold proteins) are considered “undruggable” due to the lack of traditional ligand-binding sites. Conventional fragment screening, relying on purified proteins, struggles to address targets within complex biological systems. To overcome this, Cravatt’s team constructed a covalent fragment library containing chloroacetamide or acrylamide warheads. They screened this library directly in lysates from breast cancer cells (MDA-MB-231) and Ramos cells using isoTOP-ABPP (isotopic Tandem Orthogonal Proteolysis–Activity-Based Protein Profiling). Quantitative analysis of >6150 cysteine sites identified 758 ligandable cysteines across 637 proteins, with a large number of inactive sites. This approach identified lead compound 7, which selectively targeted the catalytic cysteine C360 of procaspase-8 but did not inhibit active CASP-8. This selective targeting suggests a mechanism of allosteric inhibition by stabilizing the zymogen form. Validation using probe 61 confirmed specific labeling of CASP-8 C360. Subsequent optimization yielded a selective CASP-8 ligand 63-R. This work demonstrated that covalent chemistry targeting functional cysteines can address traditionally “undruggable” targets and established a high-throughput method for screening covalent fragments in complex systems. Nomura’s team employed competitive isoTOP-ABPP to map the targets of Withaferin A (WFA) in breast cancer cell lysates, using a broad-spectrum cysteine-reactive probe to compete with the natural product. They identified C377 on PPP2R1A (a regulatory subunit of the PP2A complex) as the specific target of WFA (competition ratio >5). WFA binding activated PP2A phosphatase activity, inhibiting the AKT pathway, validated by knockout, enzymatic assays, and phenotype correlation. Screening of a fragment library against PPP2R1A-C377 yielded hit compound DKM 2-90 (IC₅₀ = 10 μM), and subsequent structural optimization lead to a highly selective ligand JNS 1-40 (IC₅₀ = 630 nM), which significantly inhibited breast cancer growth in vivo. This study exemplifies the use of competitive isoTOP-ABPP for covalent inhibitor discovery. ABPP technology enables in situ, quantitative, and scalable target identification and validation, addressing key bottlenecks such as target blind spots, inefficient validation, and undruggable targets in traditional drug discovery. It provides a powerful approach for drug design targeting allosteric sites and challenging targets like PPP2R1A-C377 and CASP-8.

Meanwhile, ABPP significantly contributes to optimizing covalent compound selectivity. Cravatt’s team addressed limitations of traditional cysteine-directed ABPP (e.g., incomplete coverage of low-abundance or insoluble peptides) by integrating protein-directed ABPP and cysteine-directed ABPP. They used alkyne-functionalized tryptoline acrylamide stereoprobes to treat cancer cells and quantitatively identify the interactome. Subsequently, they employed iodoacetamide-desthiobiotin (IA-DTB) probes to label remaining cysteine residues, enabling competitive experiments to pinpoint specific sites occupied by the stereoprobes. This approach systematically mapped interactions of the tryptoline acrylamide stereoprobes within cancer cells. By comparing the interactions of the four stereoisomers with 271 protein targets and 238 cysteine sites, they demonstrated that the stereochemical environment of the warhead critically influences covalent modification selectivity. For example, specific engagement of Cys186 on the MAD2L1BP protein disrupted its interaction with the spindle assembly checkpoint complex, delaying mitotic exit in cancer cells. This work provides a reference for “stereochemical optimization” in anticancer covalent drug design, aiding future efforts to reduce off-target risks.

The Cravatt team developed an innovative “Paralog Hopping” strategy. First, they used ABPP to discover a unique, covalently sensitive Cys109 in the paralog protein CCNE2. Next, they introduced a corresponding mutation (N112C) into CCNE1, which naturally lacks this site. This mutant CCNE1-N112C selectively bound the covalent tryptamine acrylamide probe (e.g., WX-02-520). Guided by this strategy, they created a NanoBRET assay to screen for inhibitors that reversibly bind the CCNE1-N112C:CDK2 complex (e.g., I-125A). X-ray crystallography ultimately revealed that these inhibitors bind to a hidden allosteric pocket near N112 at the CCNE1:CDK2 interface. This strategy successfully transformed a covalent probe-discovered cysteine site into an entry point for developing reversible allosteric inhibitors, overcoming limitations in characterizing allosteric pockets with traditional methods. The broader value of this approach was demonstrated in another independent study on the WRN helicase allosteric covalent inhibitor VVD-133214, which is being evaluated in a Phase 1 clinical trial (NCT06004245). Facing challenges with WRN’s lack of a traditional catalytic pocket and difficulty competing with ATP, the research team used ABPP for covalent inhibitor discovery. First, they globally profiled solvent-accessible cysteines in over 300 cancer-related proteins in tumor cell lysates using iodoacetamide (IAA) probes. This identified the highly reactive residue Cys727 in WRN’s RecA domain hinge region. Second, they established a dual “lysate-live cell” screening platform. This identified the hit compound VVD-109063 (lysate TE₅₀ ≈ 10 μM) and crucially identified VVD-133214 as an ATP-synergistic inhibitor. VVD-133214’s binding efficiency increased in live cells due to ATP-induced conformational changes (TE₅₀ improved from 0.58 to 0.065 μM). Preincubation with ATP also boosted its biochemical inhibition potency 17.7-fold (IC₅₀ decreased from 2.3 to 0.13 μM). Crystal structure complex revealed VVD-133214 occupies a hydrophobic pocket at the RecA1-RecA2 domain interface via covalent binding to Cys727 and stabilizes a closed conformation through a hydrogen bond network.

ABPP continues to drive the evolution of covalent drugs toward heightened selectivity, allosteric intervention, and responsiveness to the live-cell microenvironment. Looking ahead, the integration of technologies such as in situ target engagement quantification, stereochemical optimization guided by chemoproteomics, and artificial intelligence-driven prediction of allosteric pockets holds significant promise. This convergence is anticipated to further unlock traditionally intractable targetsincluding transcription factors and phase-separation proteinssolidifying ABPP’s role as an essential underlying engine for the development of precision covalent therapeutics.

3.3. Covalent DNA-Encoded Chemical Library

The systematic long-term development of covalent inhibitors is constrained by off-target risks. DNA-encoded library (DEL) technology significantly accelerates screening efficiency. First proposed by Brenner and Lerner in 1992, DEL uses DNA barcodes to tag small molecules, enabling the screening of trillion-compound libraries (Figure ). However, traditional DEL relies on reversible noncovalent interactions (like hydrogen bonding or hydrophobic effects). Covalent inhibitors, in contrast, function by electrophilic warheads attacking nucleophilic residues on target proteins. This poses a challenge in the discovery of covalent ligands via DEL technology: The covalent warheads may react preferentially with abundant nucleophilic sites (oxygen atoms in the phosphate backbone, amino groups in bases) in the DNA tags, leading to library degradation and screening failure. Winssinger et al. addressed the challenge of warhead reactivity conflicting with DEL chemistry by developing a strategy using a preinstalled warhead library, strict washing discrimination, and postscreening off-DNA validation. , This approach enabled the use of DELs for screening covalent inhibitors. Neri et al. utilized DNA self-assembling chemical libraries (DSCLs) to identify bivalent ligands, which contain one pharmacophore binds the target protein and linked warhead near the binding site, enabling efficient and selective covalent modification of the target protein even under reducing environments, solving the problem of warhead selectivity in complex biological settings. These two key advancesrespectively solving the core issues of chemical compatibility and screening methodologydrove the development of DEL screening for covalent inhibitors. They also spurred subsequent rapid progress in warhead diversification and targeting strategies.

5.

(A) and (B) Schematic of reversible and irreversible DNA-encoded library selection approaches. Reproduced from ref Copyright 2023 American Chemical Society. (C) Evolution of covalent DNA-encoded libraries: (i) Warhead Diversification (ii) Diverse Library Construction (iii) ABPP-Guided Design.

Recent years have witnessed rapid advances in DEL technology for covalent drug discovery, driven primarily by innovations in three key areas: warhead chemistry and library design, screening strategy enhancement, and data/AI-driven optimization. The evolution of covalent DEL (CoDEL) technology exemplifies a significant transition from conceptual validation to sophisticated multiresidue targeting capabilities. The field’s foundation was laid in 2017 by Liu’s IDUP (Interaction Determination Using Unpurified Proteins) strategy, which employed self-pairing screening between DNA-encoded protein and compound libraries to identify a covalent inhibitor targeting MAP2K6 kinase. This pioneering work first demonstrated DEL’s feasibility for covalent compounds screening, though early efforts were constrained by limited warhead diversity (dominated by acrylamides) and narrow residue targeting (primarily cysteine). To overcome these limitations, Lu’s team pioneered multidimensional expansion of covalent warhead capabilities. Their triazine-scaffolded library innovatively unified structural and warhead functions, utilizing chlorine atoms on the triazine ring as electrophilic warheads modifiable by nucleophilic residues like cysteine and lysine. Efficient screening against BTK, JAK3, and Pin1 demonstrated how this strategy broadened the scope of warhead chemistry. In 2023, Lu and Zhou’s work achieved targeting of nonclassical residues through sulfonyl fluoride warheads engineered for specific tyrosine engagement via phenolic hydroxyl reactivity, confirming DEL’s capacity for difficult-to-modify residues. A milestone was reached in Lu’s work, where systematic evaluation of 17 warheads against 9 residues in FGFR2 enabled construction of a 24.8-million-compound multiwarhead library. This yielded covalent inhibitors targeting cysteine (validating classical sites), lysine (revealing novel reactive positions), the first arginine-targeted inhibitor (overcoming V564F resistance with FGFR2/3 selectivity), and glutamate-targeting leads (confirming this residue’s druggability). Collectively establishing the multiwarhead CoDEL platform’s versatility, this work significantly expands targetable residue and protein space to address clinical resistance challenges, ultimately transforming the technology from a proof-of-concept tool into a generalized platform for residue-selective covalent drug discovery.

The breakthrough evolution of CoDEL technology lies in the systematic advancement of screening strategies, driven by deep integration with ABPP and proteomics. Lu and Zhou’s team pioneered the ABPP-CoDEL strategy in 2023, innovatively coupling target discovery with ligand screening: First, sulfonyl fluoride probes (F1/F2) functionally scan the entire proteome to rationally identify tyrosine-targetable proteins (e.g., PGAM1, GSTP1) based on reactivity and abundance data. Next, a cognate warhead strategy was employed to design a covalent DEL library (67 million compounds) for immobilized enrichment screening against preselected targets. This approach reduces false positives through ABPP-guided target prioritization and ensures binding specificity via the use of a conserved warhead chemotype, solving the challenge of discovering noncysteine covalent inhibitors.

Lu’s team further advanced the technology into a “proteomics-guided CoDEL”: By integrating public ABPP data sets (covering chemotype- and warhead-based lysine reactivity landscapes), it systematically identifies 447 high-potential targets (e.g., PGAM1, UBE2N). Leveraging lysine microenvironment features (pK _a perturbation, spatial accessibility), it precisely matches 8 warhead types to build a 10.7-million-member library, covering full reaction mechanismsphotoactivatable (o-nitrobenzyl alcohol), reversible (aromatic aldehyde), and irreversible (squaramate/sulfonyl fluoride). Its innovation enables “single-target multi-warhead” parallel screening, breaking traditional warhead limitations and revealing novel mechanisms like allosteric inhibition. Together, these works mark CoDEL screening’s entry into a “proteomics-driven rational design” era: ABPP provides proteome-wide reactivity mapping, proteomics data guide warhead matching, and DEL enables high-throughput functional screening. This “target-warhead-mechanism” trinity strategy delivers a standardized solution for historically undruggable targets.

CoDEL technology has evolved into a powerful platform targeting diverse protein residues (such as Cys, Lys, Arg, Glu) by overcoming core challenges in chemical compatibility and screening methods. Looking ahead, CoDEL technology will continue to enhance its platform capabilities, exploring a wider range of targets, warhead mechanisms, and screening strategies. In the future, the deep integration of AI with DEL technology, including CoDEL, will significantly advance the analysis and prediction capabilities for massive screening data sets. This opens new avenues for rapidly discovering highly selective covalent lead compounds against novel and challenging targets, particularly historically “undruggable” ones.

3.4. Phage and mRNA Display Approaches

In 1985, Smith pioneered phage display technology by inserting a foreign gene into gene III of filamentous phage f1, enabling the fusion expression of exogenous peptides with the phage coat protein. The core principle links the phenotype (peptide/protein displayed on the phage surface) directly to its genotype (the encoding DNA sequence within the phage). Specific binding ligands are enriched through multiple rounds of “adsorption-washing-elution-amplification” against a target (e.g., protein, receptor), enabling high-throughput screening in vitro. However, traditional phage display generates only linear peptides composed of natural amino acids, limiting screening to noncovalent binding ligands.

Recent technological advances have extended phage display to covalent ligand screening. Bogyo’s team pioneered this approach by constructing rigid cyclic peptide libraries modified with residue-specific covalent warheads: vinyl sulfone (VS) for cysteine targeting and diphenyl phosphonate (DPP) for serine targeting. Following recombinant protein expression and iterative cycles of binding, stringent washing, elution, and amplification, they identified high-affinity covalent cyclic peptides. Validation through target binding assays, comparison with analogous inhibitors, and proteomic analysis confirmed the successful discovery of nanomolar-potent, highly specific covalent peptide inhibitors (e.g., against TEV protease and FphF). This work demonstrated the compatibility of genetically encoded covalent warheads with phage display, establishing a foundation for applications against complex targets. Building on this, the same team reported an extension of the technology in 2025, detailing a novel phage display-driven strategy for discovering covalent macrocyclic inhibitors. They employed sulfonyl fluoride (SF) electrophilic warheads, utilizing sulfur­(VI) fluoride exchange (SuFEx) chemistry to target Tyr, Lys, and His residues enriched at PPI interfaces. A variable peptide sequence was inserted between two engineered Cys on the phage pIII protein. Chemical cross-linking cyclized the peptide, fixing its conformation and significantly enhancing binding stability and protease resistance. The strategy innovatively combined binary screening (direct incubation with the target) and ternary screening (preincubation of target with a blocker before adding the library). After five rounds of progressive panning, ligands targeting functional sites were precisely identified using a “positive-to-counter selection enrichment ratio”. This methodology identified the covalent macrocyclic compound CP_SW3A, which efficiently labels the SARS-CoV-2 spike protein and blocks spike-ACE2 interactions. This achievement underscores the broad applicability of phage display in covalent inhibitor screeningnot only for developing inhibitors against traditional targets like kinases but also providing an innovative paradigm for designing covalent macromolecules targeting PPI interfaces.

Furthermore, researchers have recently developed a novel screening platform based on phage display and the reversible covalent chemistry of 2-acetylphenylboronic acid (APBA). APBA can reversibly bind the ε-amino group of lysine residues under physiological conditions to form an iminoboronate conjugate. The researchers synthesized APBA conjugates (APBA-1 and APBA-3) by coupling APBA to dichloroacetone (DCA) via an oxime linkage. They utilized DCA as a bis-cysteine cross-linker to construct cyclic peptide libraries, thereby enhancing peptide rigidity. Validation confirmed that APBA-1 enables efficient phage-peptide cross-linking. Applying this technology, the researchers constructed a CX₆C library (where X represents any amino acid) on M13 phage based on the size of the sortase A (SrtA) active site pocket. They recombinantly expressed biotinylated SrtA (using an AviTag for cotranslational biotinylation), immobilized it on streptavidin-coated magnetic beads, and performed phage panning, which successfully identified peptide ligands that effectively inhibit SrtA activity. When extending this approach to screen against the receptor-binding domain (RBD) of the SARS-CoV-2 spike protein, high-affinity RBD-binding peptides were discovered, although no inhibitors blocking the RBD–hACE2 interaction were identified. These RBD-binding peptides functioned as useful probes for coronavirus detection. In summary, phage display has demonstrated its utility in the discovery of inhibitors employing both irreversible covalent warheads (as exemplified by the work of the Bogyo group) and reversible covalent strategies (such as the APBA platform described here).

The success of phage display demonstrates the significant value of genotype-phenotype linkage for efficient molecular screening (as previously discussed). However, its reliance on living cells for protein expression and display limits library complexity (typically < 10¹¹) and hinders the screening of molecules toxic to host cells or structurally unstable. , Furthermore, the noncovalent physical linkage poses a potential dissociation risk under stringent screening conditions, such as strong denaturants or rigorous washing steps. To address these limitations and explore broader molecular diversity, mRNA display technology was developed. This approach eliminates cellular systems, performing translation entirely in vitro and permanently linking the mRNA to its encoded product via a stable covalent bond. , This fundamentally different paradigm enables the construction of ultralarge libraries (up to 10¹⁴) and the screening of molecule types challenging for traditional methods (Figure ).

6.

(A) General workflow for discovering covalent inhibitors via phage display. A covalent cyclic peptide library was constructed on phage by modifying cysteine-containing peptides with diverse warhead linkers. The library is iteratively screened: target-bound phage is purified, washed, and eluted by proteolysis. The eluted phage is then amplified for subsequent rounds or sequenced to identify covalent binders. (B) Advantages of mRNA display. mRNA display combines the use of very large libraries of protein variants; the ability to readily incorporate unnatural amino acids; and to perform selections under a wide range of conditions. Reproduced from ref Copyright 2020 American Chemical Society. (C) Chemical structures of FphB-OX-14 and FphB-OX-5 and their oxadiazolone warhead.

The core mechanism utilizes puromycina molecule mimicking the structure of a tRNA terminusattached to the 3′ end of the mRNA. During translation, puromycin is incorporated into the ribosomal A-site. Upon translation completion, it forms a stable peptide bond with the C-terminus of the newly synthesized protein, thereby creating a forced covalent linkage between the genotype (mRNA) and phenotype (protein). This enforced coupling ensures that during subsequent functional screening (e.g., affinity-based enrichment), protein variants with desired properties can be efficiently identified and enriched through their linked mRNA. , Consequently, mRNA display serves as a robust in vitro screening and directed evolution platform. It allows the selection of functionally optimized molecules from libraries containing trillions of protein variants in a single experiment, opening new avenues for fields like covalent ligand discovery.

The technological evolution of mRNA display covalent inhibitor platforms mainly involves upgrades in warhead design and cyclization strategies. Bogyo’s team developed an integrated platform combining genetic code reprogramming with mRNA display. They designed an oxadiazolone warhead (JJ-OX-009), converted it to the keto-bearing unnatural amino acid OxF-CME, and site-specifically incorporated it into peptides using engineered aminoacyl-tRNA synthetase/tRNA pairs. Libraries of 6–10 amino acid macrocycles were constructed, where each peptide underwent N-terminal chloroacetylation to enable cyclization via reaction with a downstream cysteine, while embedding the OxF-CME warhead. Through five rounds of mRNA display selection against the bacterial phosphatase FphB, high-affinity covalent cyclic peptides (e.g., FphB-OX-14 and FphB-OX-5) were enriched. Subsequent mass spectrometry and Cys53 mutagenesis confirmed covalent engagement of the warhead at the enzyme’s active site, achieving irreversible inhibition. This pioneering work established the first mRNA display platform for discovering irreversible covalent macrocyclic inhibitors, enabling targeting of challenging proteins.

After that, the Bower’s team developed a method for discovering covalent cyclic peptide inhibitors using mRNA display technology, employing dehydroalanine (Dha) as a naturally occurring electrophile. This approach involves incorporating the Dha precursor, phenylselenocysteine (PhSec), into the nascent peptide chain during translation via an orthogonal aminoacyl-tRNA synthetase (ORS). Subsequently, library cyclization is achieved using Flexizyme-mediated thioether bond formation. A post-translational oxidative elimination step then reveals the reactive electrophile, Dha. In contrast to phage display methodologies, this strategy allows precise positional control and overrepresentation of the Dha electrophile through strategic design of the mRNA library. The authors established a comprehensive workflow encompassing library construction and selection, which includes the use of guanidine hydrochloride denaturation to enrich for covalent binders. This method was successfully applied to develop potent covalent cyclic peptide inhibitors for two model targets: Calcium- and Integrin-Binding Protein 1 (CIB1) and Melanoma-Associated Antigen 4 (MAGE-A4). The study establishes Dha as a versatile electrophile for covalent inhibitor discovery and demonstrates the synergistic integration of multiple library diversification strategiesORS, post-translational chemistry, and Flexizymesthereby expanding the application of mRNA display to covalent inhibitor discovery. Building upon the foundational work of the Bogyo team on phage display for covalent inhibitors, this research deepens the exploration of electrophilic warheads, enabling more flexible control over covalent targeting.

Based on prior advancements in warhead design and cyclization techniques, the Dickinson group developed an mRNA display platform using TEV protease as a model system. This platform employs the chemical probe DBA-VS (1,3-dibromoacetone-vinyl sulfone) to simultaneously cyclize peptide libraries and install covalent warheads. Compared to the traditional DCA-VS probe (a bifunctional probe used by the Bogyo group in phage display for screening covalent cyclic peptides), the upgraded DBA-VS exhibits higher reactivity, enabling efficient dual functionalization. The researchers utilized an NNK-encoded tricysteine library (Cys­(NNK)­nCys, n = 9, 12, 15) and incorporated a post-translational purification step targeting a C-terminal HA tag to ensure sequence integrity. During the selection process, the inactive mutant TEV-C151A was introduced as a competitor, effectively enriching inhibitors targeting the active site cysteine C151. Through screening combined with AlphaFold3 structural prediction, the most potent covalent TEV inhibitor to date, cTEV6–2 (IC₅₀ = 81.7 nM), was identified. This platform enhances the efficiency of chemical modifications in mRNA display via a chemistry-driven approach, providing a powerful method for developing highly selective covalent macrocyclic peptide inhibitors.

Phage display and mRNA display technologies, leveraging genotype-phenotype coupling mechanisms, have driven innovation in covalent inhibitor development. Phage display integrates both reversible and irreversible covalent chemistry, alongside innovations in peptide cyclization and screening strategies. This significantly enhances the targeting capabilities of macrocyclic peptides, offering new solutions for challenging targets like PPIs. mRNA display, benefiting from its cell-free expression system and covalent genotype-phenotype linkage, overcomes library size limitations. Through the synergy of diverse chemical warheads and cyclization techniques, it enables efficient inhibition of high-value targets such as kinases. Looking ahead, advances in artificial intelligence will facilitate the intelligent integration of these technologies. Tools like AlphaFold will aid in optimizing warhead design and analyzing macrocyclic peptide conformations, accelerating rational inhibitor design. Furthermore, progress in reversible and controllable covalent targeting of specific amino acid residues will broaden the scope of applicable warheads. This includes warheads targeting noncanonical amino acids like methionine and tryptophan. Ultimately, the deep integration of these display technologies, combined with the convergence of chemical, computational, and biological tools, will propel covalent inhibitor development toward higher selectivity and more intelligent design strategies.

3.5. Sulfur­(VI) Fluoride Exchange Approaches

Sulfonyl fluorides have served as molecular tools for covalent inhibitors since the mid-20th century. Early examples include protease protectants in cell lysates, like PMSF and AEBSF. These work by irreversibly modifying the serine residue in the active site of hydrolases, becoming crucial tools in cell biology. Similarly, aryl sulfonyl fluoride covalent inhibitors developed by Baker’s team in the 1960s against targets like dihydrofolate reductase and trypsin demonstrated the unique value of these molecules in achieving potent enzyme inhibition through irreversible active site modification. ,

However, traditional sulfonyl-fluoride-based inhibitors have long been hampered by intricate synthetic routes, a limited capacity for selectivity modulation, and suboptimal biocompatibility, constraints that have collectively hindered their progression into viable therapeutic agents. The development of covalent drugs urgently required a general technology enabling efficient and controllable sulfur-fluorine bond exchange to meet the demands for modular construction, precise targeting, and in vivo stability. In 2014, Sharpless’ team introduced SuFEx click chemistry as a solution with prefunctionalized sulfur fluoride warheads (e.g., Ar-OSO₂F, RN = S­(O)­F₂). The S­(VI)-F bond efficiently exchanges with nucleophiles (e.g., amines, phenols) to form stable covalent bonds (e.g., sulfamides, fluorosulfates). Grimster’s subsequent study quantified SuFEx warhead (e.g., ArOSO₂F) reactivity, demonstrating a preferential and controllable reaction with Lys and Tyr at physiologically suitable rates, which minimizes off-target toxicity by avoiding nonspecific binding. The establishment of SuFEx technology and its controllable reactivity under physiological conditions overcame the historical limitations of traditional sulfonyl fluorides. It enabled rapid, highly selective biomolecular conjugation under physiological conditions, which laid a solid chemical foundation for the rational design of a new generation of covalent inhibitors featuring modular construction, precise targeting, and excellent in vivo stability.

The core strategy for screening covalent inhibitors based on SuFEx technology lies in utilizing the unique properties of sulfonyl fluoride functional groups, such as high chemical stability and controllable reactivity selectivity. Wu and Sharpless pioneered the concept of a “SuFExable warhead library”, which involves late-stage functionalization of drug molecules with sulfuryl fluoride (SO₂F₂) or thionyl tetrafluoride (SOF₄) via highly selective modification of phenolic hydroxyl or amino groups. Liquid-phase in situ SuFEx allows direct synthesis of an aryl fluorosulfate library (39 compounds) in 96-well plates, while bioactivity screening can be performed without purification. When phenolic drugs modified by SO₂F₂ bind to target proteins, the sulfonyl fluoride group forms a covalent bond via S⁶⁺-F exchange. Crucially, this does not alter the parent molecule’s target specificity. This modular construction method based on the SuFEx reaction system provides a new strategy for rapidly building covalent compound libraries.

Lead compound optimization is lengthy and challenging, especially for covalent drugs. Leveraging the biocompatibility of SuFEx click chemistry, Wolan and Sharpless et al. developed an integrated “synthesis-screening” high-throughput platform. Starting from a SpeB (bacterial cysteine protease) inhibitor 1 (K _i = 8 μmol/L), they introduced an iminosulfur oxydifluoride (RN = S­(O)­F₂) hub. Within 24 h, a one-pot reaction with 230 amines in microplates generated 460 covalent compounds containing sulfamide/sulfimidoyl fluoride bonds. Crude products were directly used for enzyme inhibition assays, which leads to identified superior SpeB inhibitor 5 with 480-fold enhancement in enzyme inhibition (K _i = 18 nM) (Figure ). This work utilized a click chemistry strategy to achieve high-throughput modular screening, and this platform increased library capacity by 10-fold and reduced single-reaction scale to the picomole level (1536-well plate, 400 pmol per well) compared to the previous work.

7.

(A) Sulfonyl fluoride (SO₂F) chemistry and sulfur-fluoride exchange (SuFEx) click chemistry. (B) Schematic representation of the mechanism associated with an SF-containing ligand targeting a nucleophilic amino acid residue. (C) Advanced SuFExable warhead library progressed through two iterations to achieve high-throughput screening. Reproduced from ref Copyright 2018 American Chemical Society and ref Copyright 2020 American Chemical Society.

The inherent modularity of SuFEx chemistry was also validated for the rapid discovery of bioactive molecules with therapeutic potential. Using phenols or polyphenol analogs as precursors, Zhang et al. synthesized a series of aryloxysulfonyl fluoride derivatives with sulfuryl fluoride via SuFEx chemistry. Screening via broth microdilution assays identified several aryloxysulfonyl fluorides that showed significantly reduced minimum inhibitory concentrations (MICs) against MRSA compared to their parent compounds. SAR analysis confirmed the -OSO₂F group is essential for the antibacterial activity of these compounds. Broader spectrum testing revealed that several derivatives exhibit potent antibacterial activity against Gram-positive bacteria, including VRSA, VRE, MRSE, and Bacillus subtilis.

Optimization of covalent lead compounds entails navigating dual challenges: demonstrating efficacy while ensuring specificity. SuFEx click chemistry, with its modular construction, bioorthogonal reactivity, and stable bond formation, offers a unique chemical toolbox when deeply integrated with core drug discovery platform technologies to overcome these hurdles. To further fine-tune the reactivity and amino acid selectivity of the warhead, research has evolved from classic SuFEx to their derived sulfonyltriazoles (SuTEx). In 2025, the Hsu team employed a covalent inhibitor strategy based on SuTEx, combined with chemical proteomics to assess cellular activity and selectivity. They designed a series of SuTEx probes and identified several featuring cyclopropyl triazole groups (e.g., RJG-2043) that exhibited enhanced tyrosine selectivity and potent targeting of AKR1C3 in phenotypic screens using Colo205 cells. The tailored probe RJG-2121 successfully enabled specific detection of endogenous AKR1C3 protein modification at tyrosine residue Y24 in cells (verified by gel fluorescence and mass spectrometry). Finally, compound RJG-2051, featuring a cyclohexyl triazole group, stood out with optimal AKR1C3 inhibitory activity and selectivity after detailed SAR study. The entire development process integrated SuTEx technology with ABPP, enabling efficient discovery of the covalent lead compound while effectively ensuring its bioactivity and target selectivity.

Based on Xiang group’s breakthrough study, SuFEx-compatible in vitro selection platform integrates SuFEx chemistry with a trillion-scale oligonucleotide library. By leveraging the reversible phosphorothioate substitution reaction, SuFEx covalent warheads are dynamically installed or removed during screening cycles to prevent polymerase deactivation and enable library regeneration. This approach facilitates multiround selection across trillion-scale (10¹²) SuFEx-functionalized oligonucleotide libraries. SuFEx in vitro selection employs an alternating strategy of positive and negative selection cycles to ensure optimal balancing of the dual performance metricshigh target specificity and high reactivityfor the discovered covalent inhibitors. This approach was successfully applied to screen covalent inhibitors targeting the SARS-CoV-2 spike protein–ACE2 PPI and the human complement C5 (hC5) protein, a therapeutic target for diseases like geographic atrophy. This work not only demonstrates the compatibility and reactivity of SuFEx functional groups within the in vitro selection context, but more importantly, pioneers a novel pathway for discovering covalent inhibitors targeting traditionally “undruggable” PPIs using SuFEx click chemistry.

In summary, these cutting-edge studies signify that SuFEx click chemistry has entered a mature stage of application in covalent drug discovery. This technology has successfully expanded the scope of targets to include highly challenging PPIs. It has also identified lead or candidate compounds with significant activity in crucial therapeutic areas like antibacterial and anticancer research. The core strengths of SuFExmodular synthesis, high-throughput screening, and controllable bioconjugationare being deeply integrated with other key drug discovery platforms. This integration is significantly improving the throughput and selectivity of covalent inhibitor screening. Advances in efficient and stereoselective sulfur­(VI) fluoride exchange methodologies have been reported. , Future convergence of chemical biology technologies with organic synthesis methods will further enhance and refine the application of SuFEx technology in covalent inhibitor screening. This progress paves the way for developing candidate drugs with improved selectivity and metabolic stability. This progress paves the way for developing candidate drugs with improved selectivity and metabolic stability. The inherent stability of SuFEx-derived linkages (e.g., sulfamides, sulfonate esters) against hydrolysis directly enhances metabolic stability by resisting enzymatic degradation. Furthermore, the modularity of SuFEx chemistry enables the rational structural optimization of lead compounds, allowing medicinal chemists to systematically eliminate metabolic soft spots and fine-tune drug-like properties. At last, we provide a personal overview on some potential advantages and limitations of the aforementioned five technologies, and the comparison of the advantages and limitations is summarized in Table .

1. Summary and Comparison of Major Platforms in Covalent Inhibitor Discovery.

technology	features and advantages	limitations	rep. refs.
Covalent Tethering	Utilizes reversible covalent bonds to capture and optimize weak-binding fragments, enabling the targeting of transient pockets.	-Requires a nucleophilic residue at the target site.	,,,,
		-Requiring further structural optimization to achieve desired function.
ABPP	Employs activity-based probes to directly quantify functional enzyme activity in native systems, enabling global assessment of inhibitor selectivity and off-target effects.	-Requires careful matching of warhead to enzyme class.	,,,
		-Coverage can be incomplete for low-abundance or insoluble proteins.
CoDEL	Seamlessly integrates diverse covalent warheads with DNA-encoded libraries for ultrahigh-throughput screening of vast libraries, enabling the targeting of multiple residue types.	-Warheads must be stable with DNA tags.	,−
		-Requires specialized library construction and screening protocols.
Phage/mRNA Display	Evolves covalent macrocyclic peptides from vast genetic libraries to target protein–protein interfaces (PPIs) via flexible warhead incorporation.	-Peptide therapeutics may have poor pharmacokinetic properties.	,,−
		-Chemical modification steps can be complex and inefficient.
SuFEx	Leverages S(VI)-F warheads for selective covalent exchange under physiological conditions, offering excellent biocompatibility and modular synthesis to target noncysteine residues	-Dependent on specific local residue environments.	,,,
		-Synthetic complexity can limit rapid candidate generation.

4. Conclusion and Outlook

Over the past few decades, the evolution of covalent drugs from “serendipitous discovery” to “rational design” signifies the entry of targeted therapy into an era of precision covalent therapeutics. This perspective systematically summarizes five key technology platformsABPP, covalent tethering, CoDEL, phage/mRNA display, and SuFExwhich collectively establish a comprehensive technological roadmap for covalent drug discovery spanning “target identification → molecular design → safety optimization” through complementary innovations.

As highlighted in the previous sections, every therapeutic technique poses its own challenges and opportunities for covalent drug discovery. Covalent tethering captures transient pockets (e.g., the S-IIP of KRAS^G12C) via dynamic covalent chemistry (disulfide/imine bonds), providing structural foundations for challenging targets. This tethering technology has been successfully used and validated in the discovery of KRAS^G12C inhibitors. ABPP leverages functional proteomic mapping (e.g., discovery of the WRN hel-icase allosteric pocket) while SuFEx enables tyrosine-selective targeting. These platforms synergistically tackle traditionally undruggable targets such as transcription factorand PPI interfaces. In addition, CoDEL achieves multiresidue targeting (Cys/Lys/Arg/Glu) through trillion-compound libraries, advancing screening from random exploration to data-driven rational design when integrated with ABPP-guided proteomics. Display technologies direct the evolution of covalent macrocyclic peptides via genotype-phenotype coupling (phage/mRNA) and customized warheads (e.g., Dha, OSF), enabling precise engagement of complex targets including PPIs. Furthermore, multiple technological approaches address off-target toxicity concerns in covalent inhibitors. ABPP globally evaluates off-target effects and SuFEx ensures controlled reactivity under physiological conditions (preferentially binding Tyr/Lys), systematically reducing covalent drug toxicity risks. Moreover, reversible covalent strategies integrated with tethering and display platforms provide new pathways for therapeutic window optimization. The discovery of covalent drugs remains largely confined to targeting cysteine, lysine, and tyrosine. To overcome this limitation, emerging electrophilic warheads are demonstrating considerable potential. Histidine, with its highly nucleophilic imidazole group, represents a promising covalent target. In addition to aryl fluorosulfates, the α-cyano keto warhead has successfully achieved covalent modification of His315 located in the NADPH-binding pocket of wild-type isocitrate dehydrogenase (IDH1). Glutamate and aspartate are also prevalent in many undruggable targets. However, their weak nucleophilicity under physiological conditions has historically hindered the development of warheads capable of targeting carboxyl groups, particularly those that form covalent bonds in a cellular context. Recent advances have introduced carbodiimide warheads as novel reactive groups that selectively target aspartate, and ynamide electrophile, which serve as efficient coupling reagents capable of forming a key α-acyloxyenamide intermediate with carboxylic acids, thus representing a promising electrophilic warhead for selectively modifying carboxyl residues in proteins. Moving forward, the integration of these residue-diverse novel warheads into established screening platforms will be essential for achieving comprehensive proteome coverage and ultimately overcoming “undruggable” targets.

The groundbreaking advances in covalent drug development stem from the deep integration of multiple technologies. In target discovery, the “ABPP-Tethering-CoDEL” integrated platform achieves an efficient closed loop: ABPP globally identifies functional residues (e.g., Cys727 in the WRN allosteric site), covalent tethering subsequently screens weakly binding fragments (e.g., KRAS^G12C lead compound 6H05), and CoDEL ultimately optimizes high-selectivity molecules through trillion-compound library iterations. The modular “Display-SuFEx” strategy enables mRNA display to generate covalent macrocyclic peptide leads, followed by SuFEx click chemistry for late-stage functionalization, rapidly constructing drug-like molecules.

The deep integration of AI and covalent technology has emerged as a frontier in drug discovery. High-accuracy protein structure prediction tools such as AlphaFold2/3 provide new avenues for identifying targetable residues (e.g., Cys, Lys, Tyr) that may become exposed in dynamic conformations on the protein surface, significantly enhancing the precision of covalent tethering and activity-based probe development. In warhead optimization, machine learning-based reactivity prediction models (e.g.,FP-Stack and GNN) enable accurate assessment of the reactivity between various warheads and cysteine within specific local microenvironments, thereby advancing the selectivity optimization of covalent inhibitors. Furthermore, this capability can be extended to the rational design of CoDEL and SuFEx-based compound libraries. Through the application of trained AI models for virtual warhead assignment prior to library construction, optimal warhead types with superior reactivity and selectivity can be identified based on features such as the electrostatic environment and solvent accessibility of the target residue. This approach is expected to guide rational design and significantly enhance the efficiency. It should be pointed out that AI has not only empowered the technology for discovering covalent hit compounds but has also played a significant role in covalent drug design. During the critical optimization stage from hit compounds to clinical drug candidates in covalent drug development, AI technology serves as a core driving force that systematically addresses the unique challenges of covalent inhibitor optimization. As demonstrated in the CDK12/13 inhibitor project, guided by the “PandaOmics” AI platform and quantum chemical calculations, chemists balanced the covalent warhead’s reactivity, successfully increasing the lead compound’s oral bioavailability nearly 10-fold while significantly reducing its toxicity. In response to the challenge of coronavirus variants, the research team employed their proprietary AI platform “Chemistry42” for de novo molecular design. This computational approach guided the development of the broad-spectrum covalent inhibitor ISM3312, enabling its core scaffold to adapt to the key protein microenvironments of different mutant strains. ISM3312 has entered the clinical stage, demonstrating the great potential of AI-generated models in covalent drug discovery. On the other hand, natural products containing potential reactive electrophilic groups can covalently react with proteins to modulate their biological action. However, target identification and lead optimization of covalent natural products remain challenging tasks. Recently, researchers utilize the TransGenGRU generative model to identify E1, a covalent NLRP3 inhibitor that targets Cys280, from structurally complex guaianolide sesquiterpenes. This success establishes a novel path for developing covalent inhibitors from natural products. AI is systematically empowering key aspects of covalent drug discoveryfrom warhead reactivity optimization and broad-spectrum activity design to the mining of novel natural product scaffoldscollectively driving the field toward a more intelligent and predictive new paradigm, moving beyond traditional trial-and-error approaches.

On the other hand, PROTAC technology, long constrained by limited E3 ligase ligands and high-affinity binding requirements for target proteins (POI), is revolutionized through covalent PROTAC. This approach stabilizes the POI-PROTAC-E3 ternary complex via covalent bonding, effectively extending target residence time. The integrated strategy extends PROTAC applications to traditionally “undegradable” targets including transcription factors and nonenzymatic scaffold proteins, demonstrating covalent technology’s potential to reshape therapeutic paradigms. While several covalent PROTACs have been developed, − their irreversible binding characteristics may negate the catalytic nature of the PROTAC activity. We believe that reversible covalent PROTACs hold promise for selective degradation of challenging targets, since a proof of concept study about BTK reversible covalent PROTAC has been demonstrated by London and co-workers. Although the new technology platforms accelerate the covalent ligand discovery, target-based or phenotype-based screening of cysteine-reactive covalent ligands library (consisting of acrylamides, maleimide and chloroacetamides, etc.) is still a highly efficient method. As prime examples, Nomura and our group independently reported the discovery of small fragment covalent ligands against challenging targets such as oncogenic transcription factor or m⁶A demethylase through target-based or phenotype-based screening, and the target validation has been performed by chemical biology approach (e.g., ABPP). More importantly, subsequent medicinal chemistry efforts are warranted to enhance the potency and druggability of the compounds presented in these studies.

It is worth mentioning that a small number of drugs react with enzyme cofactors via covalent bonding, which is contrary to the conventional nucleophile–electrophile dynamic observed in covalent drugs. , For instance, phenylcyclopropylamine­(PCPA)-based LSD1 inhibitors that covalently bind to cofactor FAD through a single electron transfer mechanism within the LSD1 catalytic pocket have been well studied and already entered into late-stage clinical trials. , Despite their great potential in covalent drug discovery, ABPP probes targeting these kinds of electrophilic enzyme cofactors are still in the early stages of development, because most ABPP probes utilizing electrophilic groups that tag nucleophilic amino acid residues. Recently, Matthews and co-workers developed a versatile class of probes containing an electron-rich hydrazine motif, which enables covalent targeting important classes of enzyme cofactors, paving a new way for chemical proteomics technologies utilizing nucleophilic covalent probes.

In summary, the evolution of covalent inhibitors has driven continuous technological advancements to address challenges in target discovery and safety profiles. This perspective aims to provide researchers with a technological framework for covalent drug development. By clarifying the core functionalities and synergistic logic across platforms, it facilitates efficient drug design and risk management, ultimately advancing targeted therapy toward greater precision and enhanced safety.

Glossary

Abbreviations

ABPP

activity-based protein profiling

AEBSF

4- (2-aminoethyl)­benzenesulfonyl fluoride hydrochloride

AMD

age-related macular degeneration

APBA

2-acetylphenylboronic acid

BME

β-mercaptoethanol

BTK

Bruton’s tyrosine kinase

CIB1

calcium- and Integrin-Binding Protein 1

CoDEL

covalent DEL

DCA

dichloroacetone

DEL

DNA-encoded library

DPP

diphenyl phosphonate

DSCLs

DNA self-assembling chemical libraries

EGFR

Epidermal Growth Factor Receptor

HTS

high-throughput screening

IA-DTB

iodoacetamide-desthiobiotin

IAA

iodoacetamide

IDUP

Interaction Determination Using Unpurified Proteins

isoTOP-ABPP

Isotopic Tandem Orthogonal Proteolysis–Activity-Based Protein Profiling

KRAS

Kirsten ratsarcoma viral oncogene homologue

LSD1

lysine-specific demethylase 1

MAGE-A4

Melanoma-Associated Antigen 4

MICs

minimum inhibitory concentrations

ORS

orthogonal aminoacyl-tRNA synthetase

PMSF

phenylmethanesulfonyl fluoride

PPI

protein–protein Interaction

PROTAC

protein-targeting chimera

RBD

receptor-binding domain

PhSec

phenylselenocysteine

SuFEx

sulfur­(VI) fluoride exchange

SuTEx

sulfonyltriazole

TCIs

targeted covalent inhibitors

TS

thymidylate synthase

VS

vinyl sulfone

The authors declare no competing financial interest.