Scaling covalent ligand discovery through dynamic combinatorial library-versus-proteome screening Authors: Yuchen Huang, Lexuan Hou, Ruiping He, Lin Zhu, Caifeng Xu, Shixiang Duan, Jing Hu, Xi Yang, Yu-Hsuan Tsai, Chu Wang, Chunmao He, Xiaoyu Li, Li Chen, Gang Li Published: 2026 ABSTRACT AbstractMass spectrometry-based chemical proteomics enables unbiased assessment of ligand potency and selectivity across the proteome. However, current approaches remain limited by the low throughput of single-compound screening and reliance on pre-synthesized libraries. Here we devise a mechanism-driven “library-versus-proteome” platform that couples dynamic combinatorial libraries with activity-based protein profiling, enabling real-time selection and optimization of ligands in complex biological systems. This approach increases screening throughput by 10- to 20-fold, streamlines library generation, and adopts a “screen first, synthesize later” paradigm. Applying this platform, we discover covalent inhibitors of serine hydrolases including PPME1, ABHD11 and PNPLA6, and reveal uncharacterized roles of PNPLA6 in lipid metabolism and cancer cell proliferation. We further extend the strategy to cysteine-targeting ligands by designing tailored warheads, enabling proteome-wide EC50 profiling of over 2600 ligandable cysteines and yielding inhibitors for NIT2, PRDX5, TXNDC17 and VCP. Focusing on VCP, we uncover a previously unrecognized signaling axis in which GPCR activity modulates activation of the ER stress-induced unfolded protein response. Using a gel-based “library-versus-proteome” assay, we screen over 800 analogues within two days and identify a more potent VCP ligand with nanomolar activity and in vivo antitumor efficacy. This work establishes library-versus-proteome screening as a scalable strategy for ligand discovery.Subject terms: Proteomics, Mass spectrometry, Combinatorial libraries Introduction Small-molecule probes are essential tools for elucidating protein function and serve as critical starting points for therapeutic development1. Currently, the majority of human proteins still lack selective chemical probes, and certain protein families such as transcription factors, RNA-binding proteins, and intrinsically disordered proteins remain challenging to target. This underscores the urgent need for new platform technologies to accelerate probe discovery2–4. Current strategies for chemical probe discovery can be broadly categorized as protein-centric or small molecule-centric methods. The protein-centric approach involves screening small molecule libraries against individual purified proteins using techniques like high-throughput screening5, DNA-encoded libraries6, or affinity selection-mass spectrometry7. On the other hand, the small molecule-centric method uses quantitative mass spectrometry (MS)-based chemical proteomics to identify proteins that interact with small molecules at the proteomic level and has been widely applied to covalent ligand discovery8,9. Covalent chemical probes offer unique advantages by combining molecular recognition with reactivity, enabling them to target shallow or cryptic binding sites and allowing discrimination between closely related nucleophilic residues through microenvironment-dependent reactivity10,11. Mass spectrometry-based chemical proteomics has transformed ligand discovery by enabling unbiased evaluation of ligand potency and selectivity across the proteome, making it a widely adopted method in both academic and industrial settings12,13. Among these platforms, activity-based protein profiling (ABPP) has been particularly effective for identifying selective covalent ligands in complex biological systems, including cell lysates, live cells, biofluids, and animal models14,15. A major advantage of ABPP for covalent ligand discovery is its target-agnostic nature, requiring no prior knowledge about target proteins or substrate assays, which enables ligand discovery even for poorly characterized proteins16,17. The mass spectrometry readout provides high resolution, broad dynamic range, and direct identification of labeled proteins or amino acid sites18. Despite these strengths, ABPP typically requires pre-synthesized compounds and one-at-a-time screening, which limits throughput and restricts exploration of chemical space. Addressing these limitations calls for scalable, proteome-wide screening strategies capable of evaluating compound libraries in a multiplexed fashion. “Library-versus-proteome” screening, where compound mixtures compete directly in the context of the proteome, offers a promising solution by enabling efficient identification of high-affinity ligand–protein pairs under physiologically relevant conditions. To enable the rapid construction of small molecule libraries rather than simply mixing multiple compounds, we turned to the dynamic combinatorial library (DCL) strategy. Since its introduction in the 1990s19, DCL has emerged as a powerful strategy for protein-targeted ligand discovery, with numerous successful examples of protein inhibitor identification20. DCL generates libraries through continuous interconversion of components via reversible chemical reactions such as imine formation, acylhydrazone exchange, and thiol-disulfide chemistry21–23. In this system, the presence of a protein template drives the formation of optimal ligands while less favorable compounds decompose, contributing to the recreation of the best binder. This adaptive process amplifies the most effective ligands, enhancing the efficiency of subsequent detection and deconvolution. However, due to limitations in detection platforms like LC-MS or NMR24–27, DCL has primarily been used for reversible inhibitor discovery against single protein targets, and expanding this approach to proteome-scale discovery remains a significant challenge in the field. In this work, we devise a mechanism-driven “library-versus-proteome” platform that synergistically couples the adaptive evolvability of DCL with the proteome-scale resolution of ABPP. This unified approach enables real-time selection and optimization of covalent ligands in native biological systems and offers several key advantages: (1) it increases the ligand discovery throughput by 10- to 20-fold compared to current MS-based screening methods while generating two-dimensional screening readouts for structure–activity relationship analysis; (2) it simplifies library construction by simply mixing initial building blocks rather than requiring lengthy organic synthesis; (3) it employs a “screen first, synthesize later” paradigm, focusing synthetic efforts only on active hits, eliminating the need for pre-synthesis of entire library; (4) it is compatible with native proteome samples, enabling proteome-wide screening under physiological conditions without requiring purified proteins. Using this approach, we discover covalent inhibitors of serine hydrolases, including PPME1, ABHD11, and PNPLA6, and reveal uncharacterized roles of PNPLA6 in lipid metabolism and cancer cell proliferation. We further extend the strategy to cysteine-targeting ligands by introducing tailored warheads, enabling proteome-wide EC50 profiling of over 2600 ligandable cysteine sites and identifying inhibitors for targets including NIT2, PRDX5, TXNDC17, and VCP. Focusing on VCP, we uncover a GPCR-modulated ER stress signaling axis involving the unfolded protein response (UPR) and rapidly identify a more potent ligand through iterative library screening. Together, these results establish library-versus-proteome screening as a scalable and efficient strategy for covalent ligand discovery in complex biological systems. Results Acylhydrazone-based DCL for serine hydrolase inhibitor discovery ABPP employs family-wide probes to target heightened nucleophilic active-site residues in enzyme families such as serine hydrolases28,29, kinases30–32, cysteine proteases33,34, metalloproteases35,36, glycosidases37, and tyrosine phosphatases38. For covalent inhibitor discovery, focused libraries with warheads specific to these enzyme families are competed against family-wide probes, where targets displaced by inhibitors are presumed to be the inhibitors’ intended targets29,39–41. We adapted this principle to develop a competitive DCL-ABPP platform, in which DCLs formed via acylhydrazone or imine chemistry inhibit members of an enzyme family. Enzymes occupied by DCL components are thereby protected from labeling by a family-wide probe. This probe, equipped with a clickable tag, enables subsequent enrichment and quantification via LC-MS/MS. Notably, LC-MS/MS resolution allows simultaneous assessment of binding interactions between all family members and all DCL components, enabling a highly efficient library-versus-proteome screening strategy for ligand discovery (Fig. 1A). We initially focused on the serine hydrolase family due to its involvement in diverse pathophysiological processes42,43. The family-wide probe fluorophosphonate (FP) exhibits high selectivity for serine hydrolases with minimal off-target reactivity41,44. Diphenylphosphonate (DPP) was selected as the warhead due to its moderate reactivity toward serine proteases, allowing tunability of potency and selectivity through recognition groups. DPP-peptide conjugates have previously enabled selective inhibition and activity imaging of serine proteases45,46. The DCL was constructed using aniline-catalyzed acylhydrazone chemistry, which is biocompatible, fully reversible, and rapidly equilibrates at pH 6.2; the products were stabilized by shifting to pH 8.047. We synthesized a DPP-containing acyl hydrazine warhead 1 (W1) and screened it against a commercial aldehyde library 1 with 103 aldehydes, divided into 11 groups with 7 shared aldehydes between groups 1 and 11 (Fig. 1B). Cheminformatics analysis revealed that 99% of the fragments had molecular weights <300 Da and complied with the “rule of three” (Fig. 1C and Supplementary Fig. 1), with most fragments containing <3 hydrogen-bond donors/acceptors, <3 rotatable bonds, and cLogP <3. Kinetic studies confirmed that the library reached equilibrium within 6 h (Supplementary Fig. 2A), with > 82% of acylhydrazone products detectable by LC-MS (Supplementary Fig. 3A). Competitive DCL-ABPP for serine hydrolase inhibitor discovery. A Schematic workflow of DCL-ABPP library-versus-proteome screening. Customized warheads react with fragments in a one-pot system to form dynamic libraries through reversible acylhydrazone or imine chemistry. The proteome shifts the equilibrium toward optimal ligands, which subsequently hinder their target proteins from labeling by family-wide probes. Probe labeling was quantified by LC-MS/MS after click chemistry conjugation. B Chemical structures of the diphenylphosphonate acyl hydrazine (warhead 1) and general formula of aldehyde library members (structures shown in Supplementary Information). C Cheminformatics analysis of the aldehyde fragment library 1 showing the distribution of molecular weight (MW), calculated LogP (cLogP), hydrogen bond donors (HBD), hydrogen bond acceptors (HBA), and rotatable bonds (RB). D, E Forward and reverse SILAC quantification of serine hydrolase inhibition by warhead alone (D) or warhead reacted with aldehyde fragments from group 2 (E). F Heatmap illustrating serine hydrolase inhibition across 11 groups. Inhibition folds were calculated as the average of forward and reverse SILAC results and normalized using the z-score formula within each group. Color scale indicates relative inhibition from weak (white) to strong (blue). G In-gel fluorescence analysis of PPME1 inhibition by group 8 library members via FP-rhodamine competition. PPME1-overexpressing HEK293T cell lysates were treated with in situ generated compounds. Coomassie Brilliant Blue (CBB) staining served as loading control. H Dose-response curves of PPME1 inhibition by compounds 18-(Z), 18-(E), 38 and 78. IC50 values were determined by nonlinear regression using Prism 9. N.D. = not determined. Data points are presented as mean ± SD (n = 3 biological replicates). I Chemical structures of 18-(Z), 18-(E), 38 and 78. Experiments were repeated at least twice with similar results. Source data are provided as a Source Data file. To quantify serine hydrolase inhibition, we treated SILAC-labeled HeLa cell lysates (light/heavy) with or without DCL, followed by FP competition and LC-MS/MS analysis. Forward (heavy: −DCL; light: +DCL) and reverse (heavy: +DCL; light: −DCL) experiments were performed to ensure data reliability (Supplementary Fig. 4). Serine hydrolases with SILAC ratios (H/L in forward and L/H in reverse) >1.5 were classified as inhibited. Screening of the 103 aldehydes, organized into 11 sub-libraries, revealed that the warhead alone (without aldehydes) did not significantly inhibit any serine hydrolases (Fig. 1D). However, in the presence of aldehydes, most sub-libraries in the DCL (except group 10) effectively blocked labeling of protein phosphatase methylesterase 1 (PPME1) by the FP probe (Fig. 1E, F and Supplementary Fig. 5A). Among these, group 8 showed optimal PPME1 inhibition and selectivity. Notably, although different aldehyde fragments modulated the degree of inhibition, the dominant target across most active sub-libraries remained PPME1. This result suggests that target preference is primarily dictated by the W1-based warhead scaffold, whereas the aldehyde fragments mainly tune inhibitory potency. To validate that the DCL is governed by thermodynamics, we pre-mixed each fragment in group 8 separately with the warhead and then combined the resulting products. The results were consistent with those from the direct mixing approach, with PPME1 being the only protein enriched in both conditions (Supplementary Fig. 5B). To identify the effective compounds, we generated individual acylhydrazone products in situ by reacting W1 with each aldehyde in group 8 and evaluated their activity. Using HEK293T cells transiently transfected with PPME1, we performed in-gel fluorescence assays with FP-rhodamine competition, which revealed substantial inhibition of PPME1 by compounds 18, 28, 38, 48, 78, and 98, while others showed partial effects (Fig. 1G). We synthesized representative hits (18, 38, 78) for further characterization. Compound 18 exhibited two distinct conformations due to intramolecular hydrogen bonding, which we identified as 18-(Z) and 18-(E) isomers through NOESY analysis (Supplementary Fig. 6). The two isomers were isolated and evaluated separately in competitive ABPP assays with FP-alkyne across a range of concentrations, and no detectable interconversion was observed under the assay conditions used. Dose-response assays demonstrated potent PPME1 inhibition with IC50 values of 7.0 μM (78), 13.8 μM (38), and 19.7 μM (18-(E)), while the 18-(Z) isomer showed no activity, highlighting the critical importance of stereochemistry (Fig. 1H, I and Supplementary Fig. 7A–D). Interestingly, NMR analysis revealed two sets of signals for these compounds at room temperature, which coalesced into single peaks at elevated temperatures, indicating the presence of conformational isomers. NOE studies confirmed that the trans-isomers of the acylhydrazone compounds were thermodynamically more stable (Supplementary Fig. 6). Intact protein mass spectrometry confirmed covalent modification of PPME1 by these compounds (Supplementary Fig. 8). To assess selectivity, we profiled compound 78 against the entire serine hydrolase family using quantitative LC-MS/MS. This analysis demonstrated remarkable selectivity for PPME1 over other family members (Supplementary Fig. 7E). To further determine whether our covalent DCL screening is driven by protein-templated selection rather than simple kinetic trapping, we monitored covalent adduct formation in a focused dynamic sub-library comprising aldehydes 18, 28, 38, 48, 58, and 68 using intact-protein ESI-MS. The results showed that, within this dynamic combinatorial library, PPME1 preferentially formed covalent adducts with compounds 18 and 38, whereas no signals corresponding to ineffective products were detected even when unreacted protein remained in the system. We further evaluated this point by varying the time of addition of aldehyde 18. When aldehyde 18 was added at 0, 0.5, 3, 5, or 6 h after the reaction had started, the corresponding PPME1-modified mass signal was still observed in a time-dependent manner (Supplementary Fig.9), indicating that the protein does not simply capture the earliest formed products, but instead continuously selects effective ligands from the evolving dynamic mixture. Besides, using PPME1 as a benchmark, we evaluated library formats ranging from 10 to 35 members. Data quality remained high in 20-member libraries (W1 × 20 fragments) but declined in 35-member formats (W1-W7 × 5 fragments) on the Q Exactive Plus mass spectrometer (Supplementary Fig. 10). These results suggest that our library-versus-proteome screening is well suited for up to 20-member libraries. Imine chemistry for ABHD11 and PNPLA6 inhibitor discovery The introduction of α-carbon substitutions to diphenylphosphonate (DPP) warheads has proven critical for developing selective serine hydrolase inhibitors, as exemplified by successful targeting of KLKs and neutrophil elastase46,48. Leveraging established synthetic methods, we incorporated α-substituents via α-amidoalkylation of triphenyl phosphite with benzyl carbamate and aldehyde derivatives, ensuring robust synthetic tractability. To expand our DCL approach, we explored imine chemistry, another widely used reversible reaction, by pairing amine-functionalized DPP warheads (W8 and W9) with a drug-like aldehyde library 2 (100 compounds). The imine library reached a stable composition within 2 h, a conservatively chosen equilibration period for the chemically diverse library under screening conditions (Supplementary Fig. 2B). After imine formation and equilibration, the resulting imine products were reduced with NaBH3CN prior to downstream screening and LC–MS analysis. LC–MS confirmed >80% product formation for W8 (Supplementary Fig. 3B), although the yield was slightly lower than that observed for acylhydrazone chemistry. Screening via SILAC-based competitive ABPP (10 aldehydes/group) revealed that W8 preferentially inhibited abhydrolase domain-containing protein 11 (ABHD11) across most groups (excluding groups 1, 2, 5, and 8) (Fig. 2A). Intriguingly, groups 3, 7, and 10 exhibited dual inhibition of ABHD11 and PPME1, while groups 4, 6, and 9 showed selective ABHD11 inhibition. To validate hits, we generated individual imine products in situ from group 9 aldehydes and W8. In-gel fluorescence analysis, facilitated by FP-rhodamine competition, using HEK293T cells overexpressing ABHD11 and PPME1 identified compounds 83 and 86 as potent ABHD11 inhibitors with minimal PPME1 activity (Supplementary Fig. 11A-B). Dose-response profiling demonstrated that 86 inhibited ABHD11 (IC50 = 1.9 µM) with eightfold selectivity over PPME1 (IC50 = 15.5 µM) (Fig. 2C and Supplementary Fig. 11C-D). Remarkably, 86 showed a 14-fold potency improvement over the parent warhead W8 (IC50 = 25.9 µM for ABHD11; no detectable PPME1 inhibition) (Fig. 2C and Supplementary Fig. 11E-F). Proteome-wide selectivity assessment by LC-MS/MS confirmed highly selective inhibition of ABHD11 over other serine hydrolases (Supplementary Fig. 11G), highlighting its potential as a selective chemical probe. 83 showed inhibitory activity against ABHD11 (IC50 = 15.1 μM) (Supplementary Fig. 12A–C). During synthesis, the imine intermediates (83-imine and 86-imine) displayed remarkable stability but exhibited reduced biological activity, likely attributable to their extended conjugated structures (Supplementary Fig. 11H–K and 12D–F). Discovery of ABHD11 and PNPLA6 inhibitors via imine chemistry. A Structures of the α-carbon substitutions of the DPP group (W8 and W9) used for serine hydrolase inhibitor screening, and their normalized heatmap of serine hydrolase inhibition across aldehyde groups. Inhibition folds (average of forward / reverse SILAC ratios) were normalized using the z-score formula within each group. The color scale indicates relative inhibition from weak (white) to strong (blue). W8: warhead 8; W9: warhead 9. B Chemical structure of compound 86. C Dose-response curves of 86 and W8 against ABHD11. IC50 values were determined by nonlinear regression using Prism 9. Data points show mean ± SD (n = 3 biological replicates). D Chemical structure of compound 70. E Dose-response curves of 70 and W9 against PNPLA6. Data points show mean ± SD (n = 3 biological replicates). F Cellular IC50 determination of 70 against PNPLA6 in HEK293T cells. Data points show mean ± SD (n = 3 biological replicates). G Enzymatic IC50 of 70 measured by phenyl valerate hydrolysis assay (colorimetric readout). Data points show mean ± SD (n = 3 biological replicates). H Independent validation of 70 inhibition using LPC (16:0) substrate in ELISA-based fatty acid release assay. Data points show mean ± SD (n = 3 biological replicates). I Untargeted lipidomics profiling of PNPLA6 substrates in lysates treated with or without 70. Heatmap shows normalized intensities of significantly altered lipids (n = 3; z-score normalized). J Hydrolysis kinetics of PNPLA6 toward lipid substrates: (Left) LPX substrates (LPC 16:0, LPS 18:0, LPE 16:0); (Right) LPE substrates (LPE 18:0, LPE 16:0, LPE 18:1). Bars show mean ± SD (n = 3 biological replicates). K Proliferation assays of HEK293T cells comparing wild-type (WT) and PNPLA6-overexpressing (PNPLA6-OE) lines following treatment with 70. Data normalized to untreated controls. Data points show mean ± SD (n = 3 biological replicates). Statistical significance was assessed by two-sided unpaired Student’s t-test. No multiple-comparisons adjustment was applied. L Metabolic pathway of LPC/LPE hydrolysis. Key enzymes are highlighted in blue; PNPLA6 is indicated in red. Experiments were repeated at least twice with similar results. Source data are provided as a Source Data file. SILAC-based screening of warhead W9 identified selective inhibition of patatin-like phospholipase domain-containing 6 (PNPLA6, also known as neuropathy target esterase, NTE) by compounds in groups 7 and 8 (Fig. 2A). Focusing on group 7 due to its superior potency and selectivity, we generated individual products by reacting W9 with each aldehyde in this group. FP-rhodamine competition assays revealed compound 70 as a potent PNPLA6 inhibitor (Supplementary Fig. 13A-B). The synthesized 70 exhibited dose-dependent inhibition of PNPLA6 in cellular proteomes with an IC50 of 786.5 nM, an 18-fold improvement over the parent warhead W9 (IC50 = 14.4 μM) (Fig. 2D, E and Supplementary Fig. 13C-D). LC-MS/MS profiling demonstrated high selectivity for PNPLA6 (Supplementary Fig. 13E). Importantly, 70 maintained potent cellular activity (IC50 = 1.1 μM) (Fig. 2F), indicating favorable membrane permeability and cellular stability. Validation using a PNPLA6-specific colorimetric assay (measuring phenyl valerate hydrolysis) confirmed its inhibition (IC50 = 1.7 μM) (Fig. 2G). As a previously unavailable PNPLA6 inhibitor, we pursued structure-activity relationship studies through medicinal chemistry optimization. Evaluation of 17 derivatives revealed that electron-withdrawing groups on the phenylsulfonyl moiety enhanced potency (Supplementary Fig. 14A-B). The most promising derivative, featuring a pyrimidine substitution, achieved 321.2 nM potency in lysates (Supplementary Fig. 14C-D), though cellular activity remained comparable to compound 70. PNPLA6 substrates profiling and functional characterization PNPLA6 is an integral membrane protein identified as the molecular target of neurotoxic organophosphates that cause axonal degeneration49. This 1327-amino acid enzyme plays essential roles in neural development and membrane lipid hydrolysis, featuring an N-terminal regulatory domain and a C-terminal catalytic domain containing the active serine residue (Ser 966). While its biological importance is well established, the complete repertoire of endogenous PNPLA6 substrates remains poorly characterized, with only 1-palmitoyl lysophosphatidylcholine (LPC 16:0) identified to date50. Using our PNPLA6 inhibitor 70 (IC50 = 832.1 nM as measured by LPC (16:0) hydrolysis assay) (Fig. 2H), we conducted comprehensive substrate profiling through untargeted lipidomics analysis of inhibitor-treated versus untreated cell lysates. This approach revealed PNPLA6’s broad specificity for LPC species with 14-24 carbon chains, while excluding longer 28-34 carbon lipids. Importantly, we identified several lysophosphatidylethanolamines (LPE) and lysophosphatidylinositols (LPI) species as potential PNPLA6 substrates (Fig. 2I). In contrast, phospholipids including phosphatidylcholine (PC), phosphatidylethanolamine (PE), and phosphatidylserine (PS) showed no detectable activity (Supplementary Fig. 15A). Subsequent kinetic analysis using purified PNPLA6 confirmed a clear substrate preference hierarchy of LPC > LPE > LPS, with distinct chain-length selectivity patterns also observed among the LPE substrates (Fig. 2J). These findings represent a significant expansion of PNPLA6’s substrate profile. Given its role in lysophospholipid hydrolysis, we next asked whether PNPLA6 affects cell growth and phospholipid homeostasis in cells. We observed that PNPLA6 overexpression enhanced cell proliferation in both HEK293T and MCF7 cells, while pharmacological inhibition with compound 70 reduced cell growth (Fig. 2K and Supplementary Fig. 15B). Pathway analysis suggested that compound 70 treatment affects phospholipid-related pathways, including PC and PE metabolism (Supplementary Fig. 15C). Based on known phospholipid metabolism, PC and PE can be hydrolyzed by phospholipase A (PLA) to generate LPC and LPE, which may be further processed to glycerophosphocholine (GPC) and glycerophosphoethanolamine (GPE) by PLA or PNPLA6 (Fig. 2L). Consistent with this biochemical framework, PNPLA6 overexpression in cells reduced levels of selected lysophospholipids, including LPC (16:0) and LPE (16:0), while increasing downstream metabolites (Supplementary Figs. 15D and 16). In addition, both PNPLA6 knockdown and compound 70 treatment shifted LPC and LPE levels in the opposite direction to PNPLA6 overexpression, consistent with inhibition of PNPLA6 activity by compound 70 (Supplementary Fig. 15E-F). Together, these results establish compound 70 as a functional chemical probe for PNPLA6 and identify additional lysophospholipid species as potential substrates of PNPLA6, further supporting its role in cellular lysophospholipid remodeling. Direct DCL-ABPP for cysteine covalent ligand discovery Recent advances in chemical proteomics have enabled the global profiling of nucleophilic residues using biocompatible chemistry. These methods extend beyond specific enzyme families to include reactive cysteines8,51,52, lysines53–55, tyrosines56,57, methionines58, histidines59, tryptophans60, arginine61, and acidic residues62,63. To adapt our DCL-ABPP strategy for such broad applications, we developed a direct screening approach that overcomes the limitations of competitive DCL-ABPP, which becomes impractical when analyzing hundreds of potential nucleophilic residues. Our solution integrates reactive warheads with alkyne-functionalized aldehyde fragments, facilitating both click chemistry-based enrichment and direct LC-MS/MS characterization of modified peptides (Fig. 3A). This approach enables simultaneous identification of modified residues and determination of the molecular mass of the bound ligands. Direct DCL-ABPP strategy for cysteine covalent ligand discovery. A Schematic of the direct DCL-ABPP strategy. Warheads and fragments containing aldehyde and alkyne groups undergo reversible reaction to form DCLs, which selectively bind nucleophilic residues. The alkyne handle enables click chemistry-based enrichment, followed by trypsin digestion, photocleavage, and LC-MS/MS analysis for site-specific and mass-resolved ligand identification. B In-gel fluorescence analysis of acrylohydrazone and chloroacetohydrazone products at varying concentrations of warheads (chloroacetohydrazide and acrylohydrazide) and aldehyde-alkyne fragments. Concentration-dependent labeling bands indicate cysteine-specific modification. Coomassie brilliant blue (CBB) staining confirms equal protein loading. (W10: warhead 10; W11: warhead 11; War.:Warhead; Frag.: Fragment). C Workflow for on-bead TMT labeling and LC-MS/MS quantification. Five concentrations of the DCL library (0.5 μM, 2 μM, 10 μM, 40 μM, and 200 μM), each in duplicate, were labeled using 10-plex TMT reagents following on-bead trypsin digestion. Photocleavage released TMT-labeled, DCL-modified peptides for LC-MS/MS analysis. D Heatmap showing EC50 values for 80 DCL compounds and their corresponding labeled cysteine sites. EC50 values were derived from dose-dependent TMT labeling intensities across five tested concentrations, and only cysteine sites with valid TMT quantification at ≥ 4 concentrations were included for fitting. E Representative EC50 curves for ligandable cysteine residues NIT2_C153, GAPDH_C152, PDIA3_C57, PRDX5_C100. Protein structures with highlighted (red) cysteine sites are shown alongside TMT ion intensity plots across five concentrations, demonstrating dose-dependent ligand engagement. Data points show mean ± SD (n = 2 biological replicates). F Validation of LC-MS/MS results by overexpression of wild-type and mutant enzymes, followed by click chemistry, biotin enrichment, and flag-tag detection. Specific labeling of wild-type proteins confirms selective binding of DCL molecules to cysteine residues. Experiments were repeated at least twice with similar results. Source data are provided as a Source Data file. For cysteine targeting, chloroacetamide and acrylamide are well-established8,64. To enable the direct DCL-ABPP workflow, we designed and synthesized chloroacetohydrazide (W10) and acrylohydrazide (W11) warheads and reacted them with aldehyde-alkyne fragments. We first examined whether the resulting products, chloroacetohydrazones and acrylohydrazones, could still react with cysteine and whether the resulting adducts would remain stable during LC-MS/MS analysis. Encouragingly, in-gel fluorescence analysis showed concentration-dependent labeling for both products, and open-search LC-MS/MS confirmed specific modification of cysteine residues (Fig. 3B and Supplementary Fig. 17). While direct aldehyde-mediated protein modification is a theoretical possibility, it was not supported by our open-search analyses. The DCL-derived conjugates were assigned primarily on the basis of intact precursor masses rather than diagnostic MS/MS fragmentation across the dynamic linkage, because partial cleavage of the hydrazone bond may occur under higher-energy collisional dissociation (HCD) conditions. In contrast, other active cysteine-targeting warheads 12-14 were also synthesized and tested but exhibited significantly weaker labeling signals (Supplementary Fig. 18). Previous studies assessing the activity of cysteine-targeting compounds typically relied on competitive assays, where test compounds competed with iodoacetamide (IAA) probes at a single concentration. Activity was inferred from competition ratios; however, this indirect approach cannot differentiate between fully competed ligands within a mixture of compounds and therefore requires additional deconvolution experiments to assign activity to individual members. To address these limitations and directly assess ligand activity while identifying binding sites, we developed an on-bead TMT-labeling workflow. In this workflow, after tryptic digestion, binding peptides immobilized on beads were labeled with TMT reagents. Five concentrations of the DCL library (0.5 μM, 2 μM, 10 μM, 40 μM, and 200 μM) were tested in duplicate using 10-plex TMT labeling to simultaneously read out labeled peptide abundances at different DCL concentrations. The labeled peptides were then photocleaved from the solid support and analyzed by LC-MS/MS (Fig. 3C). Using this workflow, we combined two warheads (W10, W11) with five aldehyde-alkyne fragments (from aldehyde library 5) per run. In just eight LC-MS/MS runs, we screened 80 compounds (W10, W11× 40 aldehyde-alkyne fragments) and obtained dose-response curves in duplicate. In total, we identified 2627 ligandable cysteine sites across 1474 proteins and calculated EC50 values (Fig. 3D and Supplementary Fig. 19). Ligands conjugated to chloroacetohydrazine and acrylohydrazine clustered closely, indicating high data quality. Several ligands, such as 5-ethynylnicotinaldehyde (G6-285, G6-299), and 1-(prop-2-yn-1-yl)−1H-imidazole-2-carbaldehyde (G7-288, G7-302), showed broad-spectrum labeling. EC50 values were derived from TMT ion intensities across the concentration range, as demonstrated for cysteine residues NIT2_C153, GAPDH_C152, PDIA3_C57, and PRDX5_C100 (Fig. 3E). To validate our LC-MS/MS findings, we overexpressed wild-type and mutant proteins, followed by click chemistry, enrichment, and flag-tag detection. This analysis confirmed specific labeling of the wild-type proteins (Fig. 3F). In summary, the direct DCL-ABPP strategy offers a robust, scalable approach for identifying cysteine-targeting covalent ligands. This method enables quantitative assessment of ligand activity and precise site-specific characterization with substantially improved throughput. Cysteine ligands enabled protein function studies To gain structural insights into the labeled cysteine residues, we performed a solvent-accessible surface area (SASA) analysis and found that over 50% of the labeled sites are buried within the proteins, as determined by their relative solvent accessibility (RSA) values (Supplementary Fig. 20A). For functional annotation, we aligned the identified cysteine residues with protein domains annotated in the PROSITE database. Several domain types were significantly enriched (Q < 0.01), including RNA-binding domains (eukaryotic RNA recognition motif and K homology domain), protein interaction and structural domains (tetratricopeptide repeat motif, calponin homology domain, LIM zinc-binding domain, and PCI domain), the actin-depolymerizing factor homology (ADF-H) domain, and enzymatic domains (thioredoxin family active site, ubiquitin-conjugating active site, and glutathione S-transferase N- and C-terminal domains) (Supplementary Fig. 20B). Interestingly, the majority (73%) of ligandable proteins identified in our dataset were not listed in DrugBank (Supplementary Fig. 20C). The DrugBank-listed proteins primarily included enzymes, cytoskeletal components, and transporter proteins. In contrast, the non-DrugBank proteins exhibited a broader functional distribution, with a notable proportion involved in metabolism, signal reception, cell adhesion, and unclassified categories. Analysis of secondary structural contexts using AlphaFold predictions revealed that ligandable cysteines were distributed across α-helices (37.6%), β-sheets (25.5%), and coil regions (36.9%), consistent with the overall cysteine distribution in the human proteome (Supplementary Fig. 20D). Furthermore, motif analysis showed no conserved local sequence features (Supplementary Fig. 20E), supporting the broad applicability of our method for global cysteine ligand discovery. Of the 1474 proteins identified, 239 were annotated with active sites in UniProt, and 235 of these had reliable structures in the AlphaFold Protein Structure Database65,66. Using these structures, we measured the distances between identified cysteines and annotated active-site residues. Our method directly labeled 45 active-site cysteines across 36 enzymes (distance = 0 Å) (Fig. 4A). Additionally, 10 and 47 labeled cysteines were located within 5 Å and 10 Å, respectively, of annotated active sites. Among these 36 direct labeled enzymes, 22% have known reversible inhibitors, while 42% have reported covalent inhibitors, highlighting the potential of our platform to identify covalent inhibitors for underexplored enzymes. Cysteine-targeting ligand discovery. A Heatmap showing labeled cysteines that overlap with annotated active-site cysteines (0 Å distance) in 36 enzymes. B–D Dose-dependent labeling of active-site cysteines by selected ligands in live cells, demonstrated for: B) G8-497 targeting NIT2, C) G4-397 targeting TXNDC17, and D) G8-324 targeting RPS6KA1. Specific labeling is observed in wild-type but not cysteine-mutant cells. E, F Enzymatic activity assays of PRDX5 (E) and TXNDC17 (F) after treatment with their corresponding ligands 2 h (G7-465: 20 μM; G4-397: 40 μM), confirming functional inhibition. Data points show mean ± SD (n = 3 biological replicates). Statistical significance was assessed by two-sided unpaired Student’s t-test. No multiple-comparisons adjustment was applied. G Fluorescence imaging of mitochondrial hydrogen peroxide using the 2 μM mitoPY1 probe after treatment with various concentrations of G7-465, in the presence of 100 μM hydrogen peroxide. Scale bar: 20 μm. Experiments were repeated at least twice with similar results. H Quantification of fluorescence intensity relative to DMSO-treated controls. Data points show mean ± SD (n = 8 cells). I EC50 determination for G7-465 based on mitoPY1 fluorescence intensity. Data points show mean ± SD (n = 3 biological replicates). J EC50 distribution of selected ligands across labeled cysteines. A threshold was applied to identify selective targets, defined as cysteine sites with EC50 values less than one-tenth of the average EC50 for all labeled cysteines by that ligand. All experiments were independently repeated at least twice with similar results. Source data are provided as a Source Data file. To demonstrate this potential, we synthesized a subset of DCL compounds and evaluated their potency and selectivity in live cells. Notably, compound G8-497, G4-397, and G8-324 selectively labeled active cysteines in NIT2, TXNDC17, and RPS6KA1, respectively, with strong labeling observed at concentrations as low as 0.1-0.5 μM (Fig. 4B–D). These results provide previously unavailable inhibitors for these enzymes. Enzymatic assays further confirmed that these ligands effectively inhibited PRDX5 and TXNDC17 activity (Fig. 4E, F). Given PRDX5’s role in decomposing mitochondrial hydrogen peroxide, we used the fluorescent probe mitoPY1 to monitor hydrogen peroxide levels as a functional readout. Treatment with G7-465 led to a dose-dependent increase in fluorescence, with an EC50 of approximately 23.4 μM, confirming selective functional perturbation of PRDX5 (Fig. 4G–I). To explore ligand binding beyond enzymatic proteins, we analyzed the EC50 distribution across all 80 compounds, applying a threshold where EC50 values were less than one-tenth of the average. This revealed several protein targets, including RPS6KA1_C432, VCP_C522, PPIA_C161 and ALDH3A1_C244 (Fig. 4D, J and Supplementary Fig. 21), suggesting opportunities for discovering ligands for previously unexplored targets. Lastly, we hypothesized that targeting surface-exposed cysteines at protein-protein interfaces could disrupt protein interactions. To test it, binary protein-protein complex structures were derived from a published database predicted by the FoldDock pipeline based on AlphaFold2 with pDockQ values greater than 0.23, as mentioned previously67. Interface cysteines were defined as those with ≥ 50% SASA reduction in binary complexes. Co-IP experiment successfully identified that the G7-302 disrupted the SKP1-FBOX44 interactions (Supplementary Fig. 22). Functional cysteine ligand for VCP regulated pathway study Cysteine-targeting ligands have been validated as modulators of proteostasis and autophagy68,69. To explore their potential as tool compounds for probing proteostasis, we synthesized ten ligands with high proteome selectivity and evaluated their cytotoxicity in HepG2 and MDA-MB-231 cells. One compound, G2-369, exhibited potent and selective cytotoxicity, with IC50 values of 1.5 μM in HepG2 and 5.8 μM in MDA-MB-231 cells (Fig. 5A and Supplementary Fig. 23A). In contrast, the IC50 values of G2-369 in corresponding noncancerous THLE-2 and MCF10A cell lines are 72.9 μM and 2.3 μM, respectively, indicating that G2-369 exhibits greater cytotoxic potency in liver cancer cells relative to MCF10A cells (Supplementary Fig. 23B-C). To elucidate the mechanism of action, we applied the isotopic tandem orthogonal proteolysis (isoTOP)-TMT-ABPP approach to identify cysteine targets of G2-369 in HepG2 cells. Cells were treated with varying concentrations of G2-369, followed by click chemistry, on-bead trypsin digestion, TMT labeling, photo-cleavage, and LC-MS/MS analysis. EC50 values were determined for the labeled cysteines (Fig. 5B). Among these, VCP (transitional endoplasmic reticulum ATPase, also known as TERA or p97) stood out due to its known role in protein quality control and degradation. To determine the cell death pathway induced by G2-369, we co-treated cells with various inhibitors of apoptosis, pyroptosis, necroptosis, autophagy, and paraptosis. Only cycloheximide (CHX), a known inhibitor of paraptosis70, rescued G2-369-induced cell death in HepG2 cells (Fig. 5C-D), indicating that G2-369 induces paraptosis. Biochemical assays confirmed that G2-369 inhibits purified VCP with an IC50 of 88.8 nM (Fig. 5E) and selectively labels wild-type VCP in cells but not a cysteine mutant (Fig. 5F). Compared to the known covalent VCP inhibitor NMS-859, G2-369 displayed a lower enzymatic IC50 (88.8 nM vs. 279.7 nM) and stronger cytotoxicity in HepG2 cells (IC50 = 38.2 μM for NMS-859), along with a more robust CHX-mediated rescue effect (Supplementary Fig. 23D–G). Functional cysteine ligand for VCP regulated signaling pathways study. A Cytotoxic profiling of selective cysteine-targeting ligands in HepG2 cells reveals potent activity of G2-369. Data points show mean ± SD (n = 3 biological replicates). B Identification of cysteine targets of G2-369 in live HepG2 cells using the isoTOP TMT-ABPP workflow. Cells were treated with a dose range of G2-369 (0-40 μM), followed by click chemistry, on-beads trypsin digestion, TMT labeling, and photo-cleavage to determine EC50 values for each labeled site. Cysteines with EC50 < 100 μM are shown. C Rescue effects of varying concentrations of cell death inhibitors (including the paraptosis inhibitor CHX and five other representative inhibitors) on HepG2 cells co-treated with 8 μM G2-369 for 24 h. Data points show mean ± SD (n = 3 biological replicates). D Chemical structure of ligand G2-369. E In vitro enzymatic assay showing that G2-369 inhibits purified VCP activity with an IC50 of 88.8 nM. Data points show mean ± SD (n = 3 biological replicates). F) Live-cell labeling of VCP_C522 (VCP-WT) by G2-369 in a dose-dependent manner; no labeling is observed in mutant cells lacking the reactive cysteine. G BONCAT-based fluorescent gel analysis to monitor nascent protein synthesis following 8 μM G2-369 or DMSO treatment for 4 h. Cells were pulsed with 4 mM L-AHA for varying durations (0–8 h) post-treatment. H Volcano plot of BONCAT-coupled TMT proteomics analysis showing the upregulation of ER stress response proteins (ATF4, DDIT3, HSPA5, and ATF3) as well as signaling modulators (GNG4 and MIDEAS) following VCP inhibition by G2-369. HepG2 cells were pre-treated with either DMSO or 8 μM G2-369 in methionine-free medium for 4 h, followed by incubation in medium containing 4 mM L-AHA for an additional 4 h to label newly synthesized proteins. Protein samples were then harvested and processed for TMT-based quantitative proteomic analysis. Statistical significance was assessed by two-sided unpaired Student’s t test. No multiple-comparisons adjustment was applied. I Western blot analysis showing that GNG4 knockdown attenuates ATF4 induction as well as ERK and PERK phosphorylation with or without treatment with 8 μM G2-369 for 8 h. J Time-course Western blot analysis demonstrating that GNG4 knockdown delays and reduces ATF4 upregulation and PERK phosphorylation upon treatment with 8 μM G2-369 in HepG2 cells. K Proposed mechanistic model illustrating how VCP inhibition by G2-369 induces paraptosis through dual activation of the PERK/ATF4 ER stress response and a parallel GNG4-mediated GPCR–ERK signaling pathway. This model was created with BioRender.com (Li, G., 2026; https://BioRender.com/mjxy8by). All experiments were repeated three times with similar results. Source data are provided as a Source Data file. VCP is a promising anticancer target, with inhibitors currently in Phase II clinical trials71. Its therapeutic relevance is supported by the finding that only cancer cells with an active mTORC-Akt axis can restore protein synthesis under proteotoxic stress, leading to paraptosis. To explore proteins re-synthesized upon VCP inhibition, we employed BONCAT (Bioorthogonal Noncanonical Amino Acid Tagging)72 combined with gel-based ABPP and proteomics (Fig. 5G). Gel analysis revealed reduced overall protein synthesis in G2-369-treated cells, with a few upregulated bands, suggesting selective re-translation of specific proteins. BONCAT-coupled TMT proteomics analysis identified significant upregulation of ATF4, DDIT3 (also called CHOP), ATF3, GNG4, HSPA5, and MIDEAS (Fig. 5H). Gene ontology (GO) analysis indicated strong enrichment of the ER stress-UPR pathway, prominently featuring HSPA5, ATF4, DDIT3, and ATF3 (Supplementary Fig. 23H). These findings point to the activation of the PERK axis, a well-characterized arm of the UPR that promotes transcriptional responses to ER stress. Interestingly, two less-characterized proteins, GNG4 and MIDEAS, were also significantly upregulated. GNG4, a G-protein γ subunit, is involved in GPCR signaling and regulation of the ERK pathway. We hypothesized that VCP inhibition may also engage a non-classical GPCR/ERK signaling axis to modulate ER stress-UPR. Consistent with this, GNG4 knockdown in HepG2 cells reduced ATF4 expression, ERK phosphorylation as well as PERK phosphorylation both basally and upon G2-369 treatment. These results suggest that GNG4 promotes PERK activation and mediates crosstalk between GPCR and ER stress signaling pathways (Fig. 5I). Notably, neither the cellular direct target-engagement profiling nor the desthiobiotin iodoacetamide (DBIA)-based competitive chemoproteomic analysis identified proteins involved in GPCR signaling as off-targets of the G2-369, supporting the interpretation that the observed effects arise from signaling crosstalk rather than direct engagement of the GPCR pathway (Supplementary Fig. 24A). Time-course analysis of protein expression showed that G2-369 treatment induced a gradual increase in ATF4 and PERK phosphorylation over time in control cells. In contrast, GNG4 knockdown delayed these responses, further confirming that GNG4 modulates both PERK pathway activation and ATF4 induction (Fig. 5J). These results suggest that VCP-induced paraptosis involves not only the ATF4/DDIT3 axes but also a GPCR/ERK signaling branch. Specifically, GNG4 upregulation activates ERK, which in turn enhances ATF4 expression, while MIDEAS expression is regulated downstream of ATF4, as confirmed by qPCR in ATF4 knockout cells (Supplementary Fig. 23I, J). Together, these interconnected signaling pathways coordinate the execution of VCP-dependent paraptosis (Fig. 5K). In summary, our findings establish G2-369 as a potent, cysteine-targeting tool compound for studying paraptosis and uncover previously unrecognized crosstalk between ER stress and GPCR signaling in cancer cell death. These insights may inform future therapeutic strategies exploiting proteostasis disruption in oncology. Competitive DCL-ABPP strategy for VCP inhibitor optimization Inspired by serine hydrolase inhibitor discovery strategies, we hypothesized that the discovered cysteine-targeting alkyne probes in Fig. 3 could function analogously to the fluorophosphonate probe, serving as competitive chemical reporters even if some of them lack optimal potency or selectivity. A follow-up gel-based competitive DCL-ABPP workflow would allow high-throughput screening of a large commercial aldehyde library to evolve these moderate inhibitors into more potent ligands rapidly for various protein targets. Given the clinical relevance of VCP in neurodegenerative diseases and cancers69, optimizing covalent VCP inhibitors holds significant therapeutic potential. To achieve this, we used G2-369 as a competition probe and combined DCL with gel-based ABPP to identify superior compounds with enhanced VCP-targeting capability. Specifically, we constructed a dynamic combinatorial library by pairing 403 commercially available aldehydes (from aldehyde library 1–4) with two warheads, chloroacetohydrazide (W10) and acrylohydrazide (W11), yielding 806 theoretical candidates (Fig. 6A). In the primary screen, 10 aldehydes were mixed with a single warhead and incubated with lysates from VCP-overexpressing cells, followed by equilibrium locking. Competitive binding was assessed by monitoring the reduction of VCP labeling by G2-369 via in-gel fluorescence, using its alkyne handle for click conjugation with rhodamine-azide. This streamlined workflow allowed all 806 library members to be screened within two days. Notably, competitive hits were identified from group 5 of aldehyde library 1 when paired with W11 (Fig. 6B). Competitive DCL-ABPP strategy for VCP inhibitor optimization. A Schematic workflow of competitive DCL-ABPP for optimization of the cysteine-targeting ligands when alkyne probes are available. B Gel-based ABPP screening of 806 acylhydrazone products derived from 403 aldehydes and two warheads W10 and W11. Each DCL mixture contained 10 aldehydes (1 μM each) and a single warhead (10 μM), incubated with VCP-overexpressing cell lysate for 6 h at room temperature. After pH adjustment to 8.0 to lock the equilibrium, the reaction continued for 2 h. G2-369 (10 μM) was added for competitive labeling of VCP for 1 h, followed by click chemistry with rhodamine azide. C Quantification of VCP inhibition based on gel fluorescence in Supplementary Fig. 25 A using ImageJ. Data points show mean ± SD (n = 3 biological replicates). D Chemical structure of compounds 65 and 95. E Enzymatic activity assays of compounds 65 and 95 using purified VCP protein. Data points show mean ± SD (n = 3 biological replicates). F IC50 values of compounds in lysates, quantified based on the gel fluorescence in Supplementary Fig. 25B using ImageJ. Data points show mean ± SD (n = 3 biological replicates). G EC50 of G2-369 in VCP-overexpressing HEK293T cells. Data points show mean ± SD (n = 3 biological replicates). H IC50 values of compounds in cells, quantified based on the gel fluorescence in Supplementary Fig. 25C using ImageJ. Data are presented as mean ± SD (n = 3 biological replicates). I, J Cell viability of AMO1 or HT29 cells following treatment with compound G2-369 or 95. Data points show mean ± SD (n = 3 biological replicates). K Inhibition of tumor growth in an HT29 colorectal cancer xenograft model by compound 95 in BALB/c nude mice. Tumor-bearing mice were treated with compound 95 by intraperitoneal injection (5 mg kg–1) three times per week during the treatment period. Data points show mean ± SD (n = 12 biological replicates). Statistical significance was assessed by two-way ANOVA. Source data are provided as a Source Data file. To identify the specific active compounds within this group, we generated the 10 corresponding products in situ by individually mixing each aldehyde from group 5 with W11 and evaluated their activity using gel-based ABPP in lysates (Supplementary Fig. 25A). Compounds 65 and 95 showed marked VCP inhibition (Fig. 6C, D). These compounds were selected for synthesis, substantially reducing synthetic burden compared to traditional screening approaches. Subsequent enzymatic assays using purified VCP protein confirmed nanomolar IC50 values for both compounds, with compound 95 achieving potency comparable to G2-369 (Fig. 6E). In lysate-based inhibition assays, compound 95 outperformed compound 65, achieving an IC50 of 129.7 nM in competition with 10 μM G2-369 (Fig. 6F and Supplementary Fig. 25B). In live cells overexpressing VCP, compound 95 retained strong and selective inhibitory activity, with a cellular IC50 of 359.3 nM under competition with 4 μM G2-369, outperforming compound 65 and G2-369 (Fig. 6H and Supplementary Fig. 25C). We further assessed the antiproliferative effects of G2-369 and compound 95 in additional cancer cell lines, including HT29 and AMO1. Compound 95 showed superior potency to G2-369 across these cell lines, with the strongest activity reaching 78.9 nM (Fig. 6I–J and Supplementary Fig. 25D). Consistent with its biochemical and cellular potency, compound 95 also suppressed tumor growth in an HT29 xenograft model in nude mice (Fig. 6K and Supplementary Fig. 25E–G), further supporting its potential as a lead compound for in vivo development. In conclusion, our work demonstrates that leveraging an alkyne-containing, cysteine-targeting probe, initially identified through direct DCL-ABPP, as a competitive reporter enables the rapid and efficient discovery of more potent covalent ligands. By integrating DCL with gel-based ABPP, this strategy dramatically accelerates the hit identification process, achieving a tenfold improvement in throughput over conventional ABPP workflows. Importantly, only validated hits, not the entire library, were synthesized, greatly reducing synthetic effort and enabling streamlined screening. This approach provides a powerful and broadly applicable platform for covalent ligand optimization, particularly when a suboptimal probe is already available. Discussion A global initiative, Target 2035, was launched in 2018 with the goal of developing a small-molecule or antibody probe for every protein in the human proteome73,74. Currently, chemical probes are available for only an estimated 4% of the human proteome, underscoring the urgent need for novel technologies to accelerate ligand discovery. ABPP coupled with LC-MS/MS has transformed ligand discovery by enabling unbiased, proteome-wide evaluation of ligand engagement in native biological contexts. Using mass spectrometry as the readout, ABPP can not only quantify the potency of a ligand but also provide crucial insights into selectivity profiles. However, its major limitation is low throughput, as it typically involves one-by-one compound screening and depends on pre-synthesized libraries, which are time- and resource-intensive. To overcome these bottlenecks, we developed a library-versus-proteome screening approach by integrating DCL with ABPP. Instead of evaluating individual ligands, this approach assembles compound libraries in situ by reacting electrophilic warheads with fragment libraries. This design markedly increases ligand discovery throughput by 10- to 20-fold compared to conventional ABPP workflows, while streamlining library construction by eliminating the need for compound pre-synthesis. By directing synthetic efforts only toward validated hits, the workflow becomes more resource-efficient. Furthermore, its compatibility with native proteomes enables proteome-wide profiling under physiologically relevant conditions, enhancing both biological relevance and translational potential. Notably, while our platform shares conceptual similarities with previous seminal work on “library-versus-library” competitive ABPP strategies41, it differs in both implementation and scope. The prior approach utilized pre-synthesized carbamate libraries screened against recombinantly expressed serine hydrolases, primarily via gel-based ABPP. In contrast, our method assembles dynamic libraries in situ through reversible covalent chemistry, allowing real-time ligand formation and selection directly in complex proteomes. Furthermore, our use of LC-MS/MS readouts enables quantitative, proteome-wide analysis with markedly improved throughput and resolution. These advances collectively establish our DCL-ABPP strategy as a scalable and generalizable platform for covalent ligand discovery. Using this approach, we first established a competitive platform for serine hydrolase inhibitor discovery, in which compound libraries competed against the family-wide probe FP-alkyne. This strategy is especially suitable for serine hydrolases since dozens of serine hydrolases were identified, yet only a few were inhibited, making follow-up validation feasible. As the only required building blocks were commercially available aldehydes, this platform is broadly accessible. Using acylhydrazone and reductive amination chemistry, we identified covalent inhibitors for PPME1 (IC50 = 7.0 μM), ABHD11 (IC50 = 1.9 μM), and PNPLA6 (IC50 = 786.5 nM). Notably, the PNPLA6 inhibitor is a previously unavailable small-molecule modulator of this enzyme, whose biological function is poorly understood. Lipidomics profiling of PNPLA6 inhibition revealed not only substrate identity but also lipid chain-length specificity and a role in regulating lipid metabolism to support cancer cell proliferation. Building on these results, we developed a direct DCL-ABPP platform for cysteine-targeting ligands. Competitive DCL-ABPP such as competition with IAA were impractical due to the difficulty in deconvoluting site-specific engagement among hundreds of labeled cysteines. Instead, we synthesized fragments bearing both aldehyde and alkyne functionalities, allowing mass-shift-based identification of covalent hits. We found that chloroacetohydrazone and acrylohydrazone warheads maintained cysteine reactivity and selectivity under physiological conditions, enabling robust DCL generation via acylhydrazone chemistry. Using TMT-based quantitative proteomics, we derived apparent EC50 values for about 2,622 cysteine sites and identified highly selective ligands for NIT2, PRDX5, TXNDC17, RPS6KA1, and VCP, most of which lack reported inhibitors. We anticipate that this cysteine-targeting dataset will serve as a valuable community resource, although only a small fraction of its potential has been explored. Given the rising interest in cysteine-reactive compounds for modulating proteostasis, phase separation, and targeted protein degradation, we applied this platform to explore cell death mechanisms. The compound G2-369, identified via direct DCL-ABPP, exhibited strong cytotoxicity and targeted VCP, a critical ATPase in protein quality control, leading to the accumulation of misfolded proteins. To investigate downstream effects, we used BONCAT proteomics to identify newly translated proteins in cancer cells following VCP inhibition. These data suggest that VCP inhibition can trigger crosstalk between the ER stress-UPR pathway and GPCR signaling in cancer cell death. Finally, leveraging G2-369 as a competitive probe in competitive DCL-ABPP gel-based platform, we rapidly screened 806 in situ-generated acylhydrazone products and identified compound 95 as a more potent VCP inhibitor. Looking forward, several improvements could further enhance the DCL-ABPP platform. Advanced mass spectrometers may alleviate current limitations in library size, boosting both throughput and spectral quality. Implementing data-independent acquisition (DIA) could also increase proteome coverage and depth. Importantly, this approach is broadly applicable to other enzyme families such as kinases, phosphatases, and deubiquitinases, and potentially to other nucleophilic amino acids such as lysine and tyrosine. However, certain limitations must be acknowledged. False negatives may arise in the selection process, leading to the potential omission of true inhibitors in serine hydrolase screening. An interesting unresolved question is whether aldehyde-templated competition between amines and hydrazides could reveal nucleophile preference under templating conditions. However, because the two systems require different catalytic and fixation conditions in the present workflow, a direct comparison is difficult to interpret. Future development of a unified dynamic covalent screening format may enable this question to be addressed more rigorously. Moreover, the current DCL-ABPP platform is limited to lysate-based screening and is not yet applicable to live-cell systems. In cysteine-targeting ligand discovery, only apparent EC50 values could be derived since reaction yields were not 100%, although the yield of aromatic aldehydes, which were exclusively present in the library, was reported to be high in aqueous systems47. In summary, we have established a high-throughput, cost-effective library-versus-proteome platform that significantly improves the efficiency of covalent ligand or inhibitor discovery. The serine- and cysteine-targeting maps generated here provide a foundation for future enzyme-targeted drug development, covalent ligand optimization, and functional study of proteostasis-related pathways. We envision that expanding this strategy to libraries containing thousands of aldehydes and dozens of warheads will broaden the ligandable proteome and accelerate probe and drug discovery efforts. Methods Cell culture HEK293T, HeLa, HepG2, MCF7, MDA-MB-231 and HT-29 cell lines were obtained from ATCC (CRL-3216, CRM-CCL-2, HB-8065, HTB-22, CRM-HTB-26, HTB-38). All cell lines tested negative for mycoplasma. HEK293T, HeLa, HepG2, MCF7, and MDA-MB-231 were cultured in DMEM supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin. HT-29 were cultured in McCOY’s 5 A supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin. THLE-2 and MCF10A cells were obtained from Procell (CL-0833, CL-0525) and maintained in the corresponding proprietary media (CM-0833, CM-H017). AMO1 cell was obtained from SSRCC, Shanghai (JY576) and was cultured in RPMI1640 with 10% fetal bovine serum and 1% penicillin/streptomycin. Animal studies All animal experiments were performed in accordance with protocols approved by the Animal Ethics Committee of Shenzhen Bay Laboratory under approval number AELG202401 and were conducted in accordance with relevant institutional and national guidelines and regulations. Mice were maintained under specific pathogen-free conditions with ad libitum access to food and water and housed under a 12-h light/12-h dark cycle. Animals were randomly assigned to experimental groups where applicable, and all efforts were made to minimize animal suffering. The study was not designed to evaluate sex-specific differences, and no sex-based comparison was performed. The maximal tumor size/burden permitted by the approved animal protocol was 1500 mm3, and the maximal tumor size/burden was not exceeded in any animal during the study. Antibody All antibody information is listed in a Supplementary Table 3. Chemical compounds The aldehyde ligand library (detailed structures provided in the Supporting Information) in this study comprised commercially available reagents unless otherwise noted. Each ligand was assigned a unique identifier following an ‘X-Y’ numbering system, where ‘1-1’ referred to ligand 1 in aldehyde library 1, ‘2-1’ represented ligand 1 in aldehyde library 2, and so on. Among these, aldehyde library 1 was used for acylhydrazone-based DCL screening, and aldehyde library 2 was employed in imine-based DCL screening. Aldehyde libraries 1–4 were utilized for the structural optimization screening of G2-369, and library 5, a dual alkyne-aldehyde functionalized set partially synthesized in-house, was applied for cysteine ligandability mapping. Detailed synthetic procedures for the chemical compounds that are not commercially available are provided in the Supporting Information. Cell viability assay Cell toxicity was assessed using the CellTiter-Glo® Luminescent Cell Viability Assay (Promega G7571). HEK293T, PNPLA6-stably expressing HEK293T and MCF7, HepG2, MDA-MB-231, MCF10A, and THLE-2 cells were seeded in 96-well plates. After 24 h, when the cells reached ~80% confluency, the medium was replaced with fresh medium containing test compounds, and cells were incubated for another 24 h. CellTiter-Glo reagent was then added at a 1:1 volume ratio, and plates were incubated at room temperature for 30 min. Luminescence was measured using an Agilent BioTek Synergy H1 plate reader following a 2-min shake. Plasmid construction Genetic constructs used in this study are listed in Supplementary Tables 1 and 2. Human PPME1 (NM_016147.3) and ABHD11 (NM_148912.4) DNA fragments were PCR-amplified from plasmids obtained via the Suzhou Research Institute plasmid-sharing platform and cloned into the pLenti-3×flag vector using Gibson assembly for overexpression in mammalian cells. DNA fragments for PNPLA6 (NM_001166114.2), NIT2 (NM_020202.5), PDIA3 (NM_005313.5), GAPDH (NM_001289746.2), PRDX5 (NM_012094.5), TXNDC17 (NM_032731.4), RPS6KA1 (NM_001006665.2), VCP (NM_148893.3), FBOX44 (NM_001014765.2) and SKP1 (NM_006930.3) were obtained from Horizon Discovery. Point mutations for NIT2 (C106S), GAPDH (C152S), PDIA3 (C57S), PRDX5 (C100S), TXNDC17 (C46S), RPS6KA1 (C432S), and VCP (C522S) were constructed via Gibson assembly and cloned into pLenti vectors (Supplementary Table 1). The target gene fragments of VCP and TXNDC17 were cloned into the pET-21a vector through homologous recombination by attaching homologous arms, successfully constructing prokaryotic expression plasmids. For knockdown experiments, shRNAs targeting GNG4 and ATF4 were cloned into the pLKO.1 vector (Supplementary Tables 1 and 2). Stable cell line generation For generating stable overexpression cell lines, HEK293T cells were plated at 2.0 × 105 cells per well in 6-well plates. After 24 h, cells at approximately 80% confluency were transfected with 2.4 μg of recombinant lentiviral plasmid (pLenti or pLX-304 backbone) and packaging plasmids (3 μg of psPAX2 and 1 μg of pMD2.G) using Lipo8000. The medium was replaced with fresh medium 24 h post-transfection. Viral particles were harvested 48 h later, and the viral supernatant was filtered through a 0.45 μm filter. The filtered supernatant was used to transduce HEK293T cells, which were then selected with 5 μg/mL blasticidin for at least 7 days. Overexpression was confirmed by Western blot. For generating stable knockdown cell lines, HEK293T cells were plated at 2.0 × 10⁵ cells per well in 6-well plates. After 24 h, cells at approximately 80% confluency were transfected with 2.4 μg of recombinant lentiviral plasmid (pLKO backbone) and packaging plasmids (3 μg of psPAX2 and 1 μg of pMD2.G) using Lipo8000. The medium was replaced with fresh medium 24 h post-transfection. Viral particles were harvested 48 h later, and the viral supernatant was filtered through a 0.45 μm filter. The filtered supernatant was used to transduce HepG2 cells, which were then selected with 2 μg/mL puromycin for at least 7 days. Knockdown efficiency was validated by Western blot or qPCR. DCL-ABPP for serine hydrolases inhibitor discovery via reversible acylhydrazone reaction and LC-MS/MS To construct the DCL, aldehyde ligands were prepared as 40 mM stock solutions. Ten aldehydes were mixed with a serine hydrolase-targeting warhead. To initiate the reversible acylhydrazone exchange, 40 mM ammonium acetate buffer (pH 6.2) was added, and the final concentration of each component was adjusted to 1 mM and the warhead to 10 mM. To prepare the proteome sample, HeLa cells were cultured in heavy and light SILAC medium75. When reached ~95% confluency, cells were washed twice with PBS, scraped, and resuspended in 50 mM ammonium acetate (pH 6.2). Lysis was performed via sonication (1 s on, 2 s off, 35% amp) for 1 min followed by centrifugation. The protein concentration of the resulting lysate was quantified to 2 mg/mL using a BCA assay. To minimize potential false positives arising from mass spectrometry competition effects, particularly during the detection of low-abundance proteins, we introduced the following adjustments to the isotope-labeled lysates prior to screening: 20% of the light-labeled lysate was spiked into the heavy-labeled sample, and conversely, 20% of the heavy-labeled lysate was added to the light-labeled sample. A total of 2 mg of protein (heavy or light) was used. For the forward labeling experiment, light proteome was incubated with a 100-fold diluted DCL library and 10 mM aniline, while the heavy proteome was mixed with DMSO and 10 mM aniline. For the reverse labeling experiment, the labels were swapped. Samples were incubated at room temperature for 6 h on a rotating apparatus. The reaction equilibrium was then locked by raising the pH above 8 using 5 μL 1 M NaOH. Heavy and Light samples were combined in a 1:1 ratio, then incubated with 2 μL 100 μM FP-Alkyne (final concentration) for 30 min at room temperature. Click chemistry was performed by adding 24 μL 100 μM (final concentration) Biotin-PEG-Azide, 24 μL 1 mM (final concentration) freshly prepared TCEP, 70 μL100 μM (final concentration) TBTA, and 24 μL1 mM (final concentration) CuSO4. The reaction was shaken at room temperature for 1 h. Samples were precipitated using a methanol: chloroform: water mixture (4:1:3 mL), followed by centrifugation at 4500 × g for 10 min at room temperature. The protein pellet was washed twice with 1 mL methanol and resuspended in 1 mL of 8 M urea in 25 mM ammonium bicarbonate (ABC) buffer. Proteins were reduced with 10 μL 10 mM (final concentration) DTT at 55 °C for 40 min, alkylated with 50 μL 15 mM (final concentration) iodoacetamide in the dark at room temperature for 30 min, and quenched with an additional 5 μL 5 mM (final concentration) DTT. Samples were diluted with 5 mL PBS and incubated with 50 µL pre-washed NeutrAvidin agarose resin (Thermo Fisher Scientific, 29204) for 4 h at room temperature. The beads were sequentially washed: three times with 5 mL PBS containing 0.2% SDS, three times with 5 mL PBS containing 1 M urea, and three times with 5 mL PBS alone. Beads were then resuspended in 300 µL of 25 mM ABC buffer containing 1 M urea, 1 mM CaCl2, and 20 ng/µL trypsin, and digested overnight at 37 °C with shaking. After centrifugation at 6010 × g for 5 min, the supernatant was collected, and the pH was adjusted to 2-3 using 20% formic acid. Peptides were desalted using a SOLAµTM SPE plate (Thermo Fisher Scientific, 60309-001), then dried using a vacuum concentrator at 37 °C and redissolved in 10 µL of 0.1% formic acid in water for quantification via NanoDrop. For LC-MS/MS analysis, 1 µg peptides were analyzed using a Q Exactive Plus mass spectrometer with nano-ESI and Xcalibur 4.3 software. Chromatographic separation was carried out on a 150 µm × 20 cm capillary column packed with 1.9 µm C18 resin using an EASY-nLC 1200 UHPLC system. Peptides were eluted using a 95-min gradient from 8% to 50% solvent B (A: 0.1% FA in water; B: 0.1% FA in 80% acetonitrile), followed by a 6-min ramp to 98% B at 300 nL/min. Data were acquired in data-dependent acquisition (DDA) mode with the following parameters: MS1 scan range 350-2000 m/z, resolution 120,000, AGC target 4 × 105, and maximum injection time 150 ms. The spray voltage was set at 2.0 kV, and the temperature of the ion transfer capillary was 320 °C. The top 10 most abundant precursor ions were selected for fragmentation using HCD at 30% normalized collision energy, with a quadrupole isolation window of 1.6 m/z. MS2 scans were acquired at 30,000 resolution, AGC target 5 × 104, and 150 ms maximum injection time. Dynamic exclusion was set to 30 s, and ions with a charge of 1+ and > 7+ were rejected for MS/MS. The FragPipe platform was utilized to analyze the raw data using a SILAC workflow. Due to the experimental design, in which 20% light-labeled sample was spiked into the heavy-labeled sample and vice versa prior to SILAC-based screening, the raw protein ratio (A) obtained from database searches was adjusted using the formula: (4 A-1) / (4-A), to yield the theoretically corrected fold-change. DCL-ABPP for serine hydrolases inhibitor discovery via reversible imine reaction and LC-MS/MS To construct the DCL, aldehyde ligands (ten per group) were prepared as 40 mM stock solutions. An imine-reactive warhead was then added, and the final concentrations were adjusted to 1 mM for each aldehyde and 10 mM for the warhead by dilution with 4× PBS (pH 6.2). HeLa cell lysates were prepared at 2 mg/mL in 4× PBS (pH 6.2). For the forward labeling experiment, 2 mg of light-labeled lysate was incubated with 100-fold diluted DCL, while 2 mg of heavy-labeled lysate was incubated with an equivalent volume of DMSO. For the reverse labeling experiment, the labeling conditions were swapped. Samples were incubated at room temperature with gentle rotation for 2 h. To stabilize the reversible imine equilibrium, 1 μL 1 mM (final concentration) sodium cyanoborohydride (NaBH3CN) was added, and the reaction was allowed to proceed for an additional one hour at room temperature. All subsequent steps, including FP-alkyne probe labeling, click chemistry, enrichment, digestion, and LC-MS/MS analysis, were carried out as described above. Competitive gel-based profiling of candidate ligands via in situ generation of subgroup members Reversible reactions were performed individually between the warhead and each of 10 ligands from the effective subgroup. For imine formation, 1 μL of warhead solution (40 mM) was mixed with 1 μL of individual ligand (40 mM) and incubated at room temperature for 2 h. To stabilize the imine products, 1 μL of sodium cyanoborohydride (NaBH₃CN, 400 mM) was added, followed by an additional one hour incubation. The resulting reaction mixture was diluted with 4× PBS (pH 6.2) to a final concentration of 500 μM and then added to a lysate containing the overexpressed target protein, yielding a final compound concentration of 10 μM. The mixture was incubated for 2 h at room temperature. For acylhydrazone reaction, the ligand and warhead were incubated under similar conditions as described above, except that the mixtures were incubated for 6 h at room temperature in the presence of 10 mM aniline as a catalyst, followed by pH adjustment to 8.0 to lock the equilibrium. All samples were subsequently treated with 5 μL 2 μM (final concentration) FP-Rho probe and incubated in the dark with gentle shaking for 1 h. Samples were then heat-denatured at 95 °C and analyzed by SDS-PAGE. DCL-ABPP for cysteine ligand discovery via reversible acylhydrazone chemistry and TMT-based LC-MS/MS analysis HeLa cell lysate was prepared at 4 mg/mL in ammonium acetate buffer (pH 6.2) supplemented with a protease inhibitor cocktail. For DCL construction, 25 μL of each acylhydrazide-based warheads (W10 and W11, each at 40 mM) and 10 μL of each five alkynyl-functionalized aldehydes (40 mM each) were used to construct the DCL. The mixture was added to the lysate, and the final concentrations of each library component were adjusted to 200 μM, 40 μM, 10 μM, 2 μM, and 0.5 μM in the presence of 10 mM aniline. Two biological replicates were prepared for each condition. Samples were incubated at room temperature with rotation for 6 h. The pH was then adjusted to 8.0 using 1 M NaOH, and samples were further incubated for 2 h to lock the equilibrium. Click chemistry was performed by conjugating Biotin-PC-azide to the alkyne-tagged proteins for 1 h at room temperature. Protein precipitation was carried out using a mixture of methanol, chloroform, and water (4 : 1 : 3 v / v / v), followed by centrifugation at 4500 × g for 10 min at room temperature. The resulting protein pellets were washed twice with 1 mL methanol and resuspended in 1 mL of 8 M urea in 25 mM HEPES buffer (pH 8.0). Reduction was performed with 10 μL 10 mM (final concentration) DTT in 25 mM ABC buffer at 55 °C for 40 min, followed by alkylation with 50 μL 15 mM (final concentration) freshly prepared iodoacetamide in the dark at room temperature for 30 min. The reaction was quenched by adding 5 μL 5 mM (final concentration) DTT. Samples were diluted with 5 mL PBS and incubated with 50 μL of pre-washed NeutrAvidin agarose beads for 4 h at room temperature with gentle mixing. Beads were washed sequentially with 5 mL PBS containing 0.2% SDS (×3), 5 mL PBS containing 1 M urea (×3), and 5 mL PBS (×3). For on-bead digestion, beads were resuspended in 300 μL of 25 mM HEPES buffer (pH8.0) containing 1 M urea and 1 mM CaCl2 and supplemented with 4 μg of trypsin. Digestion was performed overnight at 37 °C. Following digestion, beads were washed four times with 1 mL deionized water. For TMT labeling, beads were resuspended in 100 μL of 100 mM TEAB buffer (pH 8.5), and 240 μg TMT 10-plex reagent (Thermo Fisher Scientific, 90110), dissolved in 30 μL ultra-dry acetonitrile, was added76. Labeling was carried out for 2 h at room temperature with shaking. The reaction was quenched by adding 5 μL ABC to a final concentration of 5 mM and shaked for 15 min. The beads were subsequently washed four times with 1 mL deionized water. All labeled samples were pooled and the beads were resuspended in 70% methanol in water, followed by UV irradiation (365 nm) for one hour using a crosslinker. The supernatant was collected and dried in a vacuum concentrator at 37 °C. For LC-MS/MS analysis, dried peptides were reconstituted in 10 μL 0.1% formic acid and 1 μg sample was injected into a Thermo Scientific Ultimate 3000 UHPLC system coupled to an Orbitrap Exploris™ 480 mass spectrometer equipped with a Nanospray Flex ion source. Peptides were first loaded onto a precolumn (75 μm × 2 cm; Thermo Fisher Scientific, 164535) and then separated using a 25 cm analytical column (75 μm ID, Thermo Fisher Scientific, 164941) packed with 2μm C18 resin. Chromatographic separation was performed using a linear gradient from 10 to 28% buffer B over 88 min, followed by 30-50% B over 17 min, and then ramped to 99% B over 3 min at a flow rate of 300nL/min (Buffer A: 0.1% FA in water; Buffer B: 0.1% FA in 80% acetonitrile). Mass spectra were acquired in a data-dependent acquisition (DDA) mode alternating between full-scan MS and MS2 scans. The spray voltage was set at 2.1 kV and the temperature of ion transfer capillary was 320 °C. The full MS1 scans (m/z 350-1500) were recorded at a resolution of 120,000 with an AGC target of 3 × 106 and a maximum injection time of 100 ms. Selected ions were sequentially fragmented in a 2 s cycle by HCD (35% normalized collision energy) with a quadrupole isolation window of 0.7 m/z, and MS2 spectra were collected at a resolution of 30,000, using an AGC target of 2 × 105 and a 120 ms maximum injection time. Dynamic exclusion was set to 60 s, and ions with charge states from 2+ to 7+ were selected for MS/MS. Raw data were analyzed using the FragPipe platform following a standard TMT-based quantitative proteomics workflow. TMT proteomics for identification of cysteines labeled by covalent ligand The covalent ligand G2-369 was dissolved in DMEM and incubated with cells at final concentrations of 0.1, 0.5, 2, 10, and 40 μM (0.1% DMSO) for 2 h at 37 °C. Following treatment, cells were washed twice with PBS, harvested, and lysed. Protein concentration was normalized to 2 mg/mL. A total of 2 mg protein from each sample was subjected to a copper-catalyzed azide-alkyne cycloaddition (CuAAC) reaction with 24 μL 100 μM (final concentration) biotin-PC-azide for one hour at room temperature. Proteins were then precipitated using the methanol-chloroform method to remove excess reagents. Subsequent sample processing including reduction, alkylation, enrichment using NeutrAvidin resin, tryptic digestion, TMT labeling, and cleanup was performed according to the standard TMT sample preparation workflow described above. Labeled peptides were analyzed on a Thermo Scientific Orbitrap ExplorisTM 480 mass spectrometer. Time of flight mass spectrometry for labeled protein analysis A 10 μL aliquot of purified PPME1 protein (500 ng in 50 mM ammonium acetate buffer pH = 6.2) was incubated with 1.1 μL 100 μM (final concentration) test compounds (18-(E), 38, and 78) or DMSO control on ice for one hour, followed by 2 h shaking at room temperature prior to mass spectrometry analysis. Mass spectrometric analysis of labeled proteins was performed using a Waters SYNAPT XS Ion Mobility Quadrupole Time-of-Flight Mass Spectrometer (IM-QTOF MS) equipped with an ESI source. Chromatographic separation was carried out on a Thermo MAbPac TMRP column (4 μm, 3.0 × 100 mm) maintained at 40 °C. A linear gradient from 20% to 80% acetonitrile (ACN) was applied over 5 min at a flow rate of 0.2 mL/min. Data were acquired in positive ion mode with a mass range of 400 – 4000 Da. The acquisition was conducted in centroid mode with a scan time of 0.2 s per spectrum. IC50 determination of candidate molecules by gel-based ABPP Overexpression lysate of the target protein was prepared at a concentration of 1 mg/mL. Candidate molecules were subjected to a serial threefold dilution series across ten concentrations from final 100 μM to 5 nM. For each concentration, 49 μL of lysate was mixed with 1 μL of the respective compound solution. Samples were incubated at 25 °C with shaking at 1050 rpm for 6 h. Following incubation, the 5 μL FP-Rho probe was added to a final concentration of 2 μM, and samples were gently shaken in the dark for 1 h. After labeling, 4× SDS loading buffer was added, and the samples were heat-denatured at 95 °C for 10 min. Proteins were resolved by SDS-PAGE. Fluorescent gel images were acquired, and band intensities were quantified using ImageJ software. IC50 values were determined by nonlinear regression analysis of the fluorescence intensity versus compound concentration. Uncropped and unprocessed scans of the gel results are provided in the Source Data file. Validation of ligand-dependent target labeling in live cells by gel-based analysis HEK293T cells were transiently transfected with flag-tagged wild-type (WT) or mutant (Mut) constructs of the target protein. After 24 h, cells were harvested and evenly seeded into 6-well plates. About 24 h later, cells were treated with the corresponding ligand at final concentrations of 0, 0.1, 0.5, 2, 10, and 40 μM. Ligand incubation was carried out at 37 °C for 4 h. Cells were then washed twice with PBS and collected by centrifugation. After protein concentration determination via BCA assay, lysates were subjected to a copper-catalyzed azide-alkyne cycloaddition (CuAAC) reaction using 1.2 μL Rhodamine-azide (final concentration: 100 μM) for one hour at room temperature. SDS loading buffer was added, and samples were denatured at 95 °C for 10 min. Proteins were separated by 12.5% SDS-PAGE and visualized by in-gel fluorescence scanning. In parallel, input lysates were analyzed by Western blot using anti-flag antibody to assess expression levels of the WT and mutant target proteins. Purification of recombinant VCP or TXNDC17 from E. coli The coding sequences of VCP and TXNDC17 were cloned into the pET-21a vector using homologous recombination, generating prokaryotic expression constructs encoding C-terminal 6×His-tagged proteins. The plasmids were then transformed into Escherichia coli BL21 (DE3) competent cells. Transformed cells were cultured in LB medium supplemented with 50 μg/mL ampicillin at 37 °C until the optical density at 600 nm (OD600) reached ~ 0.5. For protein expression, the cultures were cooled to 16°C for VCP (or 20 °C for TXNDC17), induced with 1 mM isopropyl β-D-1-thiogalactopyranoside (IPTG), and incubated with shaking for 16 h. Cells were harvested by centrifugation at 4000 × g for 10 min at room temperature (RT) and lysed in Buffer A (PBS, 10 mM imidazole, 1 mM DTT, pH 7.5) supplemented with protease inhibitor cocktail, using a high-pressure homogenizer. The lysate was clarified by centrifugation, and the supernatant was applied to nickel-nitrilotriacetic acid (Ni-NTA) agarose resin for affinity purification. The column was washed with 50 mL of wash buffer (PBS with 10 mM imidazole), and bound proteins were eluted with stepwise imidazole concentrations (50 mM, 100 mM, 150 mM, and 250 mM) in wash buffer. Protein-containing fractions were pooled, buffer-exchanged into storage buffer (20 mM HEPES, pH7.4, 250 mM KCl, 1 mM MgCl2, 5% glycerol, 1 mM DTT), aliquoted, flash-frozen in liquid nitrogen, and stored at −80°C. Purification of target protein from HEK293T cells HEK293T cells were transiently transfected with plasmids encoding the target protein and cultured for 48 h. Cells were washed twice with PBS and lysed by scraping in 1 mL of immunoprecipitation (IP) buffer. For PPME1, the lysate was rotated at 4 °C for 30 min and then centrifuged at 13,523 × g for 10 min to remove cell debris. Fifty microliters of anti-DYKDDDDK (flag) magnetic beads (Thermo Fisher Scientific, A36798) were washed three times with 1 mL IP buffer (Thermo Fisher Scientific, 87787). The clarified lysate was incubated with the beads at room temperature with rotation for 40 min. Following incubation, beads were washed three times with 1 mL PBS. The bound protein was eluted by adding 50 μL of 3×flag peptide (0.3 mg/mL dissolved in 50 mM ammonium acetate buffer, pH 6.2) and incubating with gentle shaking in the dark for 30 min. The supernatant was collected using a magnetic rack. For PNPLA6, the lysate was rotated at 4 °C for 30 min, followed by sonication (1 s on, 2 s off, 20% amp) for 40 s, then centrifuged at 13,523 × g for 10 min. The supernatant was collected, and 50 μL of anti-DYKDDDDK magnetic beads were washed three times with 1 mL IP buffer. The lysate was incubated with beads at room temperature under rotation for 40 min. Beads were washed three times with 1 mL pre-cooled PBS. Elution was performed by incubation with 3×flag peptide (1.5 mg/mL dissolved in 4 × PBS, pH 6.2) at 4 °C overnight. The temperature was then gradually increased with shaking as follows: 10 °C for 5 min, 15 °C for 5 min, 20 °C for 5 min, and 25 °C for 20 min. The eluate was collected using a magnetic rack. PNPLA6 activity assay For the chemical substrate (pentyl ester) assay, 10 μL PNPLA6 protein (50 ng) was incubated with 1.1 μL various concentrations of inhibitor 70 for 2 h with shaking. After incubation, an equal volume of pentyl ester substrate (stock concentration 5.3 mM) was added. The reaction mixture was incubated at 37 °C for 20 min. To terminate the enzymatic reaction, an equal volume of 1.23 mM 4-aminoantipyrine was added, followed by the addition of an equal volume of potassium ferricyanide (4 mg/mL) as a colorimetric reagent. The solution was mixed thoroughly, and absorbance was measured at 486 nm using a microplate reader. For Biological substrate (LPC (16:0)) assay, 5 μL purified PNPLA6 protein (1 μg) was incubated with 5 μL lysophosphatidylcholine (LPC (16:0)) at a final concentration of 20 μM in the presence of 40 μL 3.2 mM CHAPS and 1 μL varying concentrations of compound 70. The reaction was shaken at 37 °C for 4 h. The release of free fatty acids, indicative of PNPLA6 activity, was quantified using a free fatty acid ELISA kit following the manufacturer’s instructions. VCP enzyme activity assay The 200 μL recombinant VCP protein (800 nM) was pre-incubated with 4 μL varying concentrations of inhibitor (0–50 μM) in reaction buffer containing 40 mM Tris (pH 7.5), 20 mM MgCl2, and 0.1 mg/mL BSA. The pre-incubation was carried out at 25 °C for 2 h. Subsequently, 100 μL mixture was taken and diluted with an equal volume of buffer (100 μL) and then incubated with 1 μL 100 μM ATP in the same reaction buffer for 30 min at 37 °C to initiate ATPase activity. After the enzymatic reaction, the 50 μL mixture was brought to a final volume of 200 μL with deionized water. Then, 30 μL of phosphate assay reagent (Abcam, ab65622) was added to each reaction well and mixed thoroughly. The mixture was incubated in the dark for 30 min, and the release of inorganic phosphate was measured by absorbance at 650 nm using a spectrophotometer. All reagents added (except reaction buffers) are reported as final concentrations. PRDX5 enzyme activity assay The enzymatic activity of PRDX5 was assessed by monitoring NADPH consumption in a reaction mixture containing 90 μL 50 mM HEPES-NaOH (pH 7.0), 20 μL 250 μM NADPH, 10 μL 50 nM thioredoxin reductase (TrxR), 20 μL 3 μM thioredoxin (Trx), and 25 μL 2 μM PRDX5. All components were prepared in 50 mM HEPES buffer (pH 7.0). The reaction mixture was pre-equilibrated at 30 °C, and the reaction was initiated by the addition of 10 μL 0.5 mM H2O2. The decrease in absorbance at 340 nm, corresponding to NADPH oxidation, was continuously monitored using a spectrophotometer. A standard curve of NADPH was constructed using serial dilutions (250 μM, 50 μM, 10 μM, 2 μM, and 0.5 μM) to correlate absorbance values with NADPH concentration, allowing quantification of consumption rates and calculation of enzymatic kinetic parameters. All reagents added (except reaction buffers) are reported as final concentrations. TXNDC17 enzyme activity assay 49 μL purified 5 μM TXNDC17 protein was pre-incubated with either 1 μL 40 μM compound G4-397 or an equivalent volume of DMSO for 2 h at room temperature with continuous shaking. Then 5 μL pre-treated protein was then added to reaction mixtures containing 126 μL TE buffer (50 mM Tris-HCl, 2 mM EDTA (pH 7.5)), 2.25 μL 150 μM NADPH, 1.875 μL 150 nM thioredoxin reductase (TrxR) and 15 μL 160 μM insulin. Parallel reactions without TXNDC17 were included as background controls. Parallel reactions without TXNDC17 were included as background controls. Reactions were incubated at 37 °C for 4 h, and NADPH consumption was monitored by measuring absorbance at 340 nm. A standard curve was generated using NADPH serial dilutions (250, 50, 10, 2, and 0.5 μM) to convert absorbance values into concentration units. All data were background-corrected, and TXNDC17 activity was quantified by fitting the kinetic data to the standard curve. All reagents added (except reaction buffers) are reported as final concentrations. Pull-down and Western blot validation To validate covalent ligand binding to cysteine residues identified by proteomics, wild-type and mutant forms of representative proteins (PRDX5, NIT2, GAPDH, and PDIA3) were overexpressed in HEK293T cells. 98 μL cell lysates (400 μg) were collected and incubated with 2 μL purified covalent ligands at varying concentrations for 2 h with shaking. The pH was adjusted to 8.0 using 1 M NaOH, followed by click chemistry conjugation with 2.4 μL 100 μM (final concentration) biotin-PC-azide for 1 h at room temperature. Proteins were precipitated using 500 μL pre-cooled acetone, and the pellet was resuspended in 0.2 mL of 8 M urea in 25 mM ABC buffer, then diluted with 4 mL PBS. The solution was incubated with 25 μL of pre-washed NeutrAvidin agarose resin beads for 4 h, followed by washing as previously described. After a final wash, 20 µL 4× SDS-PAGE loading buffer was added to the beads, which were then denatured at 95 °C for 30 min. Samples were separated by SDS-PAGE and transferred to PVDF membranes for Western blot analysis. Membranes were blocked with 5% non-fat milk in TBST (TBS + 0.1% Tween-20), incubated with primary antibodies (1:1000 dilution) at 4 °C overnight, and then incubated with HRP-conjugated secondary antibodies (1:5000 dilution) for one hour at room temperature. Protein bands were visualized using chemiluminescence on a ChemiDoc MP Touch imaging system. Uncropped and unprocessed scans of the immunoblots are provided in the Source Data file. Co-immunoprecipitation and Western blot analysis HEK293T cells were transiently transfected with flag-tagged SKP1 or V5-tagged FBXO44 expression plasmids. After 48 h, cells were washed with PBS and lysed in the presence of protease/phosphatase inhibitors. For the binding assay, 100 μL flag-tagged lysates (300 μg) were pre-incubated with 1 μL test ligands (0, 2, 10, 40 μM final concentration) with constant shaking at room temperature for 2 h. 100 μL V5-tagged lysates (300 μg) were then added and incubated for an additional hour at room temperature to allow complex formation. Protein complexes were immunoprecipitated by incubating with pre-washed 5 μL anti-flag magnetic beads (Thermo Fisher Scientific, A36797) overnight at 4 °C with rotation. Beads were subsequently washed three times with PBST 10 min at room temperature. Bound proteins were eluted in 1× SDS-PAGE loading buffer by heating at 95 °C for 10 min. For Western blot analysis, samples were resolved by SDS-PAGE and transferred to PVDF membranes. Membranes were probed with anti-flag (1:1000) and anti-V5 (1:1000) primary antibodies overnight at 4 °C, followed by appropriate HRP-conjugated secondary antibodies (1:5000). Protein bands were visualized using enhanced chemiluminescence. Identification of unknown lipid substrates of PNPLA6 For the experimental group, 2 μL 7.7 μM (final concentration) compound 70 was added in HeLa cell lysate, while the control group received an equivalent volume of DMSO. LPC (16:0) was then added to a final concentration of 20 μM, along with 1 μL 3.2 mM (final concentration) CHAPS. The reaction mixtures were gently shaken at 37 °C for 4 h. Following incubation, 600 μL of a methanol:isopropanol (1:1, v/v) solution was added to each sample and mixed thoroughly. Samples were centrifuged at 21,130 × g for 10 min at 4 °C. The resulting supernatants were collected and transferred to LC-MS vials for subsequent lipidomic analysis. Metabolomics and lipidomics analysis of cells WT, stable PNPLA6-overexpressing, and PNPLA6-knockdown HEK293T cells, as well as WT HEK293T cells treated with compound 70 (10 μM), were cultured in 6-cm dishes for 3 days, with the culture medium refreshed every 24 h. The culture medium was then removed, and the cells were washed with PBS. A total of 405 μL of methanol:water (5 : 4, v/v) was added, and the cells were scraped into 1.5 mL Eppendorf tubes using a cell scraper. Subsequently, 750 μL of methyl tert-butyl ether (MTBE) was added, and the mixture was vortexed for 1 min to ensure complete mixing. The samples were centrifuged at 4 °C and 15,000 rpm for 2 min to separate the phases. 500 μL of the upper organic phase was collected, dried under nitrogen gas, and reconstituted in 100 μL of methanol:isopropanol (1 : 1, v / v) for lipidomics LC-MS analysis. 100 μL of the lower aqueous phase was mixed with 300 μL of acetonitrile:methanol (1 : 1, v/v), vortexed for 10 s, and incubated at –80 °C overnight for protein precipitation. After incubation, the samples were vortexed again, centrifuged at 4 °C and 15,000 rpm for 10 min, and the supernatants were collected for metabolomics LC-MS analysis. Metabolomics profiling was performed using an AB SCIEX TripleTOF 6600+ mass spectrometer coupled with an iHILIC-(P) Classic column (5 μm, 150 × 2.1 mm, 200 Å). Mobile phase A consisted of 20 mM ammonium acetate with ammonium hydroxide and 2.5 μM medronic acid, and mobile phase B was 100% acetonitrile. The flow rate was 0.2 mL/min. The gradient elution was programmed as follows: 85% B from 0 to 2 min, decreased to 60% B from 2 to 7 min, reduced to 35% B from 7 to 12 min, and washed at 20% B for 4 min. Mass spectrometry analysis was conducted in both positive and negative ion modes. The gas settings (GS1 and GS2) were set at 60, the curtain gas (CUR) at 35, and the ion source temperature (TEM) at 500 °C for positive mode and 550 °C for negative mode. The ion spray voltage (ISVF) was set to +5000 V for positive mode and −4500 V for negative mode. For the TOF MS scan, the mass range was set to 60-1200 m/z in positive mode and 70–1200 m/z in negative mode, with an accumulation time of 250 ms. The declustering potential (DP) was ±60 V, and the collision energy (CE) was ±10 V. For MS/MS scans, the mass range was 25–1200 m/z with an accumulation time of 50 ms, CE was set to ±30 V, and the collision energy spread (CES) was 15. The resolution for Q1 was set to UNIT. The charge state was set to 1, and ions with intensities above 100 counts per second (cps) were selected. Ions were excluded for 4 s after two occurrences, and a maximum of six candidate ions were monitored per cycle. The mass tolerance was set to 20 ppm, and isotopes within 4 Da were excluded. The injection system was configured with a sampling speed of 5 μL/s, a cooler temperature of 6 °C, and a needle stroke of 52 mm. Rinsing was performed before and after aspiration using 500 μL of solvent with a dip time of 5 s and a rinse speed of 35 μL/s. The column oven temperature was maintained at 30 °C with a maximum temperature limit of 45 °C. Metabolite annotation was achieved using the NetID algorithm77. Selected metabolite annotations were further supported by matching experimental MS/MS spectra against reference spectra from HMDB78, MassBank of North America (MoNA; RRID:SCR_015536), and the NIST Tandem Mass Spectral Library79. Annotated metabolites were mapped to Small Molecule Pathway Database (SMPDB) pathways80 for pathway-level interpretation. Lipidomics analysis was performed using a Vanquish F liquid chromatography system coupled to an Orbitrap ExplorisTM 480 mass spectrometer (Thermo Fisher Scientific). Chromatographic separation was achieved on an Acclaim C30 column (3 μm, 2.1 × 150 mm) maintained at 45 °C. The aqueous mobile phase consisted of 10 mM ammonium formate and 0.1% formic acid in water:acetonitrile (40 : 60, v / v), while the organic phase consisted of 10 mM ammonium formate and 0.1% formic acid in isopropanol:acetonitrile (90 : 10, v / v). The flow rate was set at 0.26 mL/min with a 10 μL injection volume. The gradient program was as follows: 32% B at 0 min, held at 32% B until 1.5 min, increased to 45% B at 4 min, 52% B at 5 min, 58% B at 8 min, 66% B at 11 min, 70% B at 14 min, 75% B at 18 min, and 97% B at 21 min, held until 25 min, followed by re-equilibration to 32% B from 25.01 to 33 min. Mass spectrometry data were acquired in Full MS with data-dependent MS/MS (ddMS2) mode in both positive and negative ion modes. The ion source parameters were set as follows: spray voltage +3200 V (positive mode) and −3000 V (negative mode); sheath gas 40 arb, auxiliary gas 10 arb, and sweep gas 1 arb. The ion transfer tube was maintained at 300 °C, and the vaporizer temperature was set to 350 °C. Lipid species were annotated using LipidSearch 4.0 software81. For lipid class-level analysis, the abundance of each lipid class, such as PC and PE, was calculated by summing the peak intensities of all annotated lipid species assigned to that class in the LipidSearch result table. Immunofluorescence imaging HeLa cells were seeded into 35 mm confocal dishes at a density of approximately 2.5 × 105 cells per well. After 12 h, the culture medium was removed, and the cells were washed twice with PBS. The compound G7-465 was dissolved in DMEM (0.5% DMSO) at final concentrations of 0, 0.1, 0.5, 2, 10, and 40 μM and applied to the cells, followed by incubation at 37 °C for 2 h. The cells were then washed twice with PBS, and mitoPY1 was added at a final concentration of 2 μM in PBS. Following a 30-min incubation at 37 °C, the cells were again washed twice with PBS, and Hoechst was added at a final concentration of 10 ng/mL in PBS. After 15 min of incubation at 37 °C, the cells were washed twice with PBS. Hydrogen peroxide was then added at a final concentration of 100 μM in PBS, and the cells were incubated at 37 °C for 10 min. After two final PBS washes, the samples were preserved in HBSS. Immunofluorescence images were acquired using a ZEISS LSM 980 confocal microscope equipped with ZEN 3.5 software. Non-canonical amino acid metabolic labeling detected by gel-based ABPP HepG2 cells were cultured in methionine-depleted medium and treated with 8 μM G2-369 or DMSO as a vehicle control for 4 h. After two washes with PBS, cells were pulse-labeled with 4 mM L-azidohomoalanine (L-AHA) for various time points (0, 0.5, 1, 2, 4, or 8 h). Following labeling, cells were collected and subjected to copper-free click chemistry using a final concentration of 100 μM Alkyne-Rhodamine (alkyne-Rho) for one hour. The samples were subsequently denatured and analyzed by SDS-PAGE. All reagents added (except reaction buffers) are reported as final concentrations. Non-canonical amino acid metabolic labeling detected by MS-based ABPP Metabolic labeling was performed under the same conditions as previously described. After click chemistry conjugation, samples were prepared for proteomic analysis using the established TMT-based workflow described above. Control (DMSO-treated) and experimental (G2-369-treated) samples were labeled with distinct channels of TMT-10-plex reagents, with three biological replicates per condition. Mass spectrometric analysis was conducted on an Orbitrap ExplorisTM 480 mass spectrometer operating in data-dependent acquisition (DDA) mode. qPCR analysis HepG2 cells and ATF4-knockdown cells (ATF4-sh2) were treated with either DMSO (vehicle control) or 8 μM (final concentration) G2-369 for 8 h. Total RNA was extracted using RNAiso Plus (TaKaRa, 9109), followed by genomic DNA removal and reverse transcription into cDNA using the FastKing cDNA First-Strand Synthesis Kit (TIANGEN, KR116-01). Quantitative PCR was conducted using SYBR Green Pro Taq HS Premix (ACCURATE BIOLOGY, AG11718) on a real-time PCR system, with fluorescence detection performed at 60 °C. Relative MIDEAS expression levels were normalized to β-actin as the internal control, and expression in the DMSO-treated HepG2 group was used as the reference (fold change = 1) for comparative analysis. In vivo evaluation of compound 95 in an HT-29 xenograft model Six-week-old female BALB/cAnSlac-nu mice (stock no. SM-014; Shanghai Model Organisms Center, Inc., Shanghai, China) were used for the HT-29 xenograft study. A total of 24 mice were used in this experiment, with n = 12 mice per group. Female mice were used for the HT-29 xenograft study to maintain consistency with the established xenograft protocol. Therefore, sex was not considered as an experimental variable in the study design or statistical analysis. All mice were housed in a Specific Pathogen-Free (SPF) grade animal facility, maintained at 22 ± 2 °C with 45–60% relative humidity and a 12 h light/dark cycle (lights on from 7:00 a.m. to 7:00 p.m.). HT-29 cells (3 × 106 cells per mouse) were subcutaneously inoculated into the right dorsal flank of nude mice. Once the tumors reached approximately 100 mm3, the mice were randomly divided into the blank control group and the experimental treatment group. Mice in the experimental group received intraperitoneal injection of compound 95 at a dose of 5 mg/kg (formulated in 5% DMSO and 95% corn oil) three times a week, on Monday, Wednesday and Friday. Tumor length and width were measured with a caliper, and tumor volume was calculated as width2× length × 0.5. After 21 days of treatment, tumor growth monitoring was terminated. Tumor tissues were dissected and recorded, and the mice were sacrificed humanely at the end of the experiment. SASA and secondary structure location analysis Structural models utilized for the SASA analysis were obtained from the AlphaFold Protein Structure Database66. To ensure reliability, only models with an average per-residue predicted Local Distance Difference Test (pLDDT) score above 70 and a total sequence length under 2600 amino acids were selected for secondary structure mapping66. Relative solvent accessibility (RSA) values and secondary structure classifications of the labeled cysteine residues were determined using the DSSP software (Version 4.0 + )82,83. Domain enrichment analysis Labeled cysteine residues were mapped to functional domains using the Prosite domain definitions as annotated in the UniProt database. A labeled cysteine was considered a domain “hit” if it overlapped with any annotated domain region. Consistent with the criteria established by Hsu et al.56, multiple labeled sites within the same domain annotation were counted as a single hit, while sites overlapping with multiple domain annotations contributed a hit to each corresponding domain. In this analysis, the UniProt reference database contains a total of N domains. Within our dataset, K of these domains were matched. For a given domain, n represents its total frequency in UniProt, and k denotes the number of times it appeared in our labeled dataset. The statistical significance of enrichment for each domain was assessed using a binomial test, commonly employed in gene ontology (GO) enrichment analysis84: Resulting P values were corrected for multiple comparisons using the Benjamini-Hochberg procedure to control the false discovery rate at 1%. Domains with adjusted Q-values below 0.01 were considered significantly enriched and visualized in a bubble plot generated using R. This enrichment analysis was performed separately for proteins with and without DrugBank annotations that contained labeled cysteines. Data analysis and processing Raw data acquired in DDA mode were searched against the corresponding SwissProt-reviewed protein databases, supplemented with common contaminants, using Fragpipe (v18.0). Peptide identification required full tryptic specificity with up to two missed cleavage sites permitted. Carbamidomethylation of cysteine was set as a fixed modification, while methionine oxidation was included as a variable modification. Mass tolerances were set to 10 ppm for precursor ions and 0.02 Da for fragment ions (MS2 Orbitrap). Contaminant proteins were removed, and protein identifications were filtered to maintain a false discovery rate (FDR) below 1%. Subcellular localization was assessed via Gene Ontology (GO) analysis using DAVID Bioinformatics Resources. For cysteine site-level analysis in the closed search, peptides containing modified cysteines were collapsed into a single representative peptide. Peptides were also consolidated regardless of methionine oxidation status. Miscleaved peptides at lysine or arginine residues were merged into single identifications, and intensity values were averaged accordingly. Peptides with zero intensity across all photosensitizer conditions were excluded. For peptides containing multiple cysteine sites, the signal was deconvoluted and assigned to individual cysteine residues as appropriate. Statistics and reproducibility Statistical analyses were performed using GraphPad Prism 9, and data processing was conducted in Python (version 3.10.11). Comparisons between two groups were evaluated using two-sided Student’s t tests. Error bars represent the mean ± standard deviation (SD). No statistical method was used to predetermine sample size. Experiments were not randomized, and investigators were not blinded to group allocation or outcome assessment. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Reporting Summary Transparent Peer Review file Source data Source Data Acknowledgements We thank the mass spectrometry, imaging, sequencing, and NMR core facility in Shenzhen Bay Laboratory for their assistance in running samples and collecting data. Author contributions G.L. conceived of the study and supervised research. Y.H., L.H., C.H., C.W., X.L. and G.L. designed and analyzed biological experiments. R.H. and L.C. designed and analyzed the lipidomics and metabolomics experiments. L.Z., C.X., J.H., S.D., Y.T. designed and synthesized the chemical compounds. X.Y. wrote the Python and R code for data processing, bioinformatics and chemical informatics analysis. Y.H. and G.L. wrote the manuscript with input from all authors. Peer review Peer review information Nature Communications thanks the anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available. Funding This work was supported by the National Natural Science Foundation of China [22477082 to G.L., 32422048 to L.C., and 22277029 to C.H.], the Youth Science Fund Program (Category B) of the National Natural Science Foundation of China [22522703 to G.L.], the Guangdong Special Support Plan for Outstanding Young Talents [2023TQ07A238 to G.L.], the Research Grants Council of Hong Kong SAR, China [17301125, N_HKU702/23, and T12-705-24-R to X.L.], and the Laboratory for Synthetic Chemistry and Chemical Biology under the Health@InnoHK Program [to X.L.]. Data availability The mass spectrometry data generated in this study have been deposited to the ProteomeXchange Consortium via the iProX85 partner repository with the dataset identifier PXD064043 [https://www.iprox.cn//page/SCV017.html?query=PXD064043]. The metabolomics and lipidomics data generated in this study have been deposited in MassIVE under the dataset identifier MSV000102047 [https://massive.ucsd.edu/ProteoSAFe/dataset.jsp?task=d7e4f822603b4077b8e4308d37b929d1]. Source data are provided with this paper. Code availability The Python code and the corresponding dataset have been deposited to Zenodo with 10.5281/zenodo.15448066 [https://zenodo.org/records/17667171]. Alternatively, it is available from the G.L. upon request. Competing interests G.L., Y.H., L.H. and C.H. are named inventors on patent applications related to the DCL-ABPP strategy described in this study, including US patent application no. US 19/253,687 and Chinese patent application no. 202410077968.8. G.L., Y.H., L.H. and Z.L. are named inventors on Chinese patent application no. 202510918817.5, which is related to protein inhibitors and their uses, including compounds described in this study. The remaining authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. These authors contributed equally: Yuchen Huang, Lexuan Hou, Ruiping He, Lin Zhu. Contributor Information Chunmao He, Email: hecm@scut.edu.cn. Xiaoyu Li, Email: xiaoyuli@hku.hk. Li Chen, Email: chenli_imib@fudan.edu.cn. Gang Li, Email: ligang@szbl.ac.cn. Supplementary information The online version contains supplementary material available at 10.1038/s41467-026-74672-y. References Associated Data Supplementary Materials Supplementary Information Reporting Summary Transparent Peer Review file Source Data Data Availability Statement The mass spectrometry data generated in this study have been deposited to the ProteomeXchange Consortium via the iProX85 partner repository with the dataset identifier PXD064043 [https://www.iprox.cn//page/SCV017.html?query=PXD064043]. The metabolomics and lipidomics data generated in this study have been deposited in MassIVE under the dataset identifier MSV000102047 [https://massive.ucsd.edu/ProteoSAFe/dataset.jsp?task=d7e4f822603b4077b8e4308d37b929d1]. Source data are provided with this paper. The Python code and the corresponding dataset have been deposited to Zenodo with 10.5281/zenodo.15448066 [https://zenodo.org/records/17667171]. Alternatively, it is available from the G.L. upon request.