Neisseria gonorrhoeae is a common Gram-negative pathogen with increasing resistance to all recommended antibiotics. There is a critical need to improve the efficiency of the antibiotic hit discovery process to replenish the drug development pipeline. Here, we show that deep learning models can augment high-throughput screens to identify readily available molecules with narrow-spectrum activity against difficult-to-treat strains of N. gonorrhoeae. We phenotypically tested 38,650 small molecules for N. gonorrhoeae growth inhibition to train a predictive graph neural network (GNN) model. We benchmarked the model's performance against other architectures, including a large language model, and found that GNNs more accurately identify active, drug-like molecules that are structurally distinct from the training set and known antibiotics. Using the model to virtually screen ~6 million compounds, we identified 213 compounds for experimental validation and found that 83 (39%) inhibited N. gonorrhoeae growth. Two of these compounds were structurally dissimilar to existing antibiotics, maintained potency against multidrug-resistant N. gonorrhoeae strains in vitro, exhibited promising selectivity indices, and were rapidly bactericidal with low frequencies of resistance. Proteomic studies revealed their distinct mechanisms of action, with one compound targeting alanine racemase, an enzyme involved in the essential process of peptidoglycan synthesis. Furthermore, the compounds showed early promise in reducing N. gonorrhoeae titers in a human vagina-on-a-chip infection model and a mouse vaginal infection model. Our work establishes the deep learning-enabled discovery of selective antibacterial compounds against N. gonorrhoeae as a much-needed hit discovery tool to address the growing crisis of antimicrobial resistance for this pathogen.

INTRODUCTION: Alzheimer's disease (AD) progresses over decades, yet plasma biomarkers that resolve disease stage rather than simply detect disease remain scarce. This distinction is clinically consequential because effective AD intervention depends on identifying patients before disease biology has progressed beyond a therapeutically responsive stage. METHODS: We used small-molecule-modulated protein corona proteomics to profile plasma from 90 individuals in the Australian Imaging, Biomarker and Lifestyle cohort, stratified by Centiloid (CL) Aβ-amyloid burden (30 amyloid-negative, CL < 15; 30 moderate-to-high, CL 26 to 100; 30 very high, CL > 100). We quantified 3,176 proteins and applied differential abundance and actual causality analyses to identify stage-specific and candidate causal proteins. RESULTS: Differential protein abundance was exclusively captured during the moderate-to-high AD transition, revealing a discrete proteomic "switch." The switch was marked by accumulation of the autophagy receptor CALCOCO1, together with coordinated depletion of the S100A8/S100A9 calprotectin complex and core erythroid-cytoskeletal network structural markers (e.g., SPTA1, SPTB, ANK1). Adhesion G protein-coupled receptor G6 (ADGRG6) showed a significant moderate positive monotonic association with absolute CL burden, providing a proportional molecular anchor for cumulative disease burden. Actual causality analysis identified COL6A2, FOXRED2, P3H1, PRR4, and GOLGA5 as candidate upstream drivers linking matrix remodeling, Golgi trafficking, and collagen processing to AD progression. DISCUSSION: These findings suggest a candidate blood-accessible framework for staging AD by active disease biology, which, if replicated in independent cohorts, may have implications for therapeutic selection and mechanism-guided clinical trials.

The gut bacterial microbiota is increasingly recognized as a key modulator of colorectal cancer (CRC) initiation, progression and response to therapy. However, the mechanisms by which bacteria influence the response to anticancer drugs remain poorly understood. Here, we investigate the effects of microbiota-driven signaling on the tumor suppressor p53 and its impact on chemotherapy. We uncover a mechanism by which lipopolysaccharide (LPS) from Klebsiella pneumoniae and other Enterobacteria impairs p53 activity and promotes chemoresistance via paracrine signaling from the tumor microenvironment. While direct exposure to LPS did not alter the drug response of CRC cells, conditioned media from LPS-stimulated macrophages or fibroblasts suppressed p53 accumulation and attenuated the response to chemotherapeutic agents. Deep quantitative proteomics further revealed selective inhibition of a subset of p53 targets by inflammation. This same subset negatively correlated with inflammatory signature and immune infiltration in patients and was associated with improved survival following chemotherapy. Mechanistically, our data suggest that macrophage-derived extracellular vesicles contribute to p53 degradation in cancer cells. Overall, our findings reveal a microbiota-driven mechanism of p53 suppression via the microenvironment that contributes to chemoresistance, highlighting the impact of bacteria on tumor cell fate and therapeutic efficacy in CRC.

Identifying how drugs interact with proteins is fundamental to understanding their therapeutic effects and side effects. While numerous chemical proteomics methods exist for determining protein targets of drugs, each exhibits “blind spots,” necessitating complementary approaches. We introduce Above-Filter Digestion Proteomics (AFDIP), which monitors trypsin digestion rates that decrease at ligand-binding sites, while potentially increasing elsewhere. Molecular dynamics simulations showed that these changes relate to backbone flexibility. Using AFDIP, we identified targets of various drugs and metabolites, allowing two-dimensional analysis with the drug concentration as the second dimension. The method identifies binding sites within ≤10 Å of crystallography-determined locations with improved resolution (≤5 Å) for larger proteins. Compared with existing proteolysis approaches, AFDIP offers simpler sample preparation, deeper proteome analysis, and broader sequence coverage. AFDIP addresses the blind spots of current techniques and provides structural insights, enhancing the chemical proteomics toolkit.

The human plasma proteome contains extensive diagnostic information but remains difficult to interrogate because protein concentrations span more than ten orders of magnitude, with highly abundant proteins such as albumin masking low-abundance biomarkers. Nanoparticle (NP) enrichment strategies partially address this limitation, but their analytical depth is still constrained the incorporation of albumin and other highly abundant proteins within the protein corona. Here we show that the NP protein corona can be rationally reprogrammed by pre-incubating plasma with a chemically diverse panel of ligands designed to bind albumin, attenuate its interaction with NPs, and promote enrichment of lower-abundance proteins. Molecular docking and structural analyses demonstrate that these ligands induce allosteric conformational changes and surface charge redistribution on albumin, thereby reducing its association with the NP surface. Importantly, this strategy is highly tunable, as individual ligands shift protein corona composition toward different broad families of proteins, revealing overall enrichment trends rather than strict selectivity for any single class. Using this approach, we quantified over 6,600 proteins from a single human plasma sample across a curated panel of conditions, substantially exceeding the depth achieved with untreated NP enrichment. These findings establish a versatile chem-bio strategy for ligand-directed protein corona engineering, enabling high-depth plasma proteome profiling and expanding access to low-abundance proteins and candidate biomarkers across human diseases.

Voltage dependent anion channels (VDACs 1, 2 and 3) in the outer mitochondrial membrane control the flux of anions and oxidizable substrates that sustain mitochondrial metabolism. Nicotinamide adenine dinucleotide (NADH) closes VDAC by binding to a pocket, conserved in all isoforms, located in the inner wall of the channel. Previously, we identified the small molecule SC18 that targets the NADH-binding pocket of VDAC1 employing computational analysis. Here, we explored the interaction between SC18 and VDAC1 using high-resolution nuclear magnetic resonance spectroscopy and molecular dynamics simulations. Atomically resolved data precisely confirmed the computational results, showing that SC18 binds to a site on VDAC1 that partially overlaps with the NADH binding pocket. SC18, in the presence of NADH blocked the conductance of VDAC1 reconstituted in lipid bilayers. To determine the metabolic effect of SC18, we combined readouts of mitochondrial metabolism and glycolysis with functional metabolomics and proteomics. Short-term treatment with SC18 inhibited mitochondrial metabolism and adenosine triphosphate production. Treatment over 24 h and 48 h further reduced mitochondrial uptake of pyruvate and glutamine, utilization of tricarboxylic acid cycle intermediates, as well as lipid, DNA and amino acid synthesis. Concomitant with the inhibition of mitochondrial metabolism, cellular uptake of glucose and glutamine increased in parallel with augmented lactate release. These results indicate that compensatory enhanced glycolysis sustains adenosine triphosphate production after impaired mitochondrial function induced by SC18 blockage of VDAC1. Our work sets a mechanistic foundation for VDAC1 inhibition as a novel strategy to target and reprogram cancer metabolism through modulation of the biosynthetic ability of mitochondria.

The nanoparticle (NP) protein corona is considered the biological identity that determines NP fate, safety, targeting, and therapeutic effectiveness in biofluids. Nonetheless, standard corona isolation workflows assume the recovered protein signature originates primarily from plasma proteins adsorbed directly onto the NP surface, while largely overlooking co-isolation of endogenous nanoscale biological structures such as extracellular vesicles (EVs). This oversight can distort the apparent “biological identity” of the NP. Here, we show that EVs are a major hidden contributor to the perceived protein corona composition in human plasma. Using highly monodispersed polystyrene NPs (50-1000 nm) and superparamagnetic beads, we compared corona formation in standard human plasma and plasma depleted of an EV- enriched sedimentable fraction by ultracentrifugation at 100,000 × g for 2 h, with the recovered vesicles subsequently characterized by MACSPlex immunoaffinity analysis. Mass spectrometry revealed that EV depletion reduced the number of proteins identified on polystyrene NPs by 60-75% and on magnetic beads by 45-50%, demonstrating a substantial fraction of the conventionally assigned corona proteome arises from EV- associated carryover. EV depletion also restructured the apparent abundance hierarchy, increasing the relative prominence of soluble plasma proteins such as albumin and shifting dominant signals away from intracellular cytoskeletal component proteins that are characteristic of EV carryover towards genuine soluble plasma proteins and complement factors. These results highlight that standard corona workflows can inadvertently co-isolate a vast array of EV-associated material and thereby yield inaccurate assignments of protein origin. Distinguishing proteins adsorbed from the soluble phase from EV-surface and intravesicular material is essential for accurate interpretation of NP-biofluid interactions, biomarker discovery, and therapeutic targeting because molecular compartment determines both analytical meaning and drug accessibility. Significance Statement The nanoparticle (NP) “protein corona” defines how engineered nanomaterials interact with living systems, influencing therapeutic safety, efficacy, and diagnostic utility. Conventionally, corona isolation workflows assume that recovered proteins were adsorbed directly from the fluid phase. This study reveals a major, previously overlooked source of analytical distortion: standard separation techniques routinely co-isolate endogenous extracellular vesicles (EVs), drastically distorting the perceived biological identity of NPs. Depletion of an EV-enriched sedimentable fraction by ultracentrifugation reduced identified corona proteins by up to 75% and restructures the apparent proteomic hierarchy. Distinguishing true soluble adsorbates from vesicular carryover is essential for accurately predicting NP behavior in vivo and prevents false positives in nano- diagnostics, establishing a critical new standard for high-fidelity biomarker discovery and nanomedicine.

Advancements in high-throughput techniques such as Thermal Proteome Profiling and the high-throughput Proteome Integral Solubility Alteration assay have revolutionized our understanding of drug-protein interactions. Despite these innovations, the absence of an integrative platform for cross-study analysis of stability and solubility alteration data represents a significant bottleneck. To address this gap, we introduce Drug-target interactOmics Resource based on Stability/Solubility Alteration Assay (DORSSAA), an interactive and expandable web-based platform for the systematic analysis and visualization of proteome stability and solubility alteration assay datasets. Currently, DORSSAA features 1,135,985 records spanning 38 cell lines and organisms, 135 compounds, and 40,742 protein targets. Through its user-friendly interface, the resource supports comparative drug-protein interaction analysis and facilitates the discovery of actionable therapeutic targets. Through two case studies, methotrexate target profiling in A549 cells and combinatorial-therapy drug-target interactions in leukemia cell lines, we demonstrate DORSSAA's utility for identifying protein-drug interactions across diverse experimental contexts. This resource empowers researchers to accelerate drug discovery and enhance our understanding of protein behavior. Compared with data repositories and interaction databases, DORSSAA provides direct protein-level evidence of mechanisms of action with strict statistical control for each study. This enables more reliable identification of drug targets, off-target effects, and potential drug combinations.

Precise characterization of proteoforms within the protein corona is essential for developing safer and more effective nanomedicines for diagnostic and therapeutic applications. Here, we advance the characterization of proteoforms within the protein corona by integrating mass spectrometry (MS)-based top-down proteomics (TDP) and bottom-up proteomics (BUP). TDP of the protein corona formed on polystyrene nanoparticles (PSNPs) identifies over 5000 proteoforms of 400 genes from breast cancer plasma samples, creating one of the largest human plasma TDP datasets and the most comprehensive proteoform database of the protein corona. Combining TDP and BUP improves the characterization quality for about 35% of identified proteoforms containing mass shifts, producing a more precise proteoform landscape of protein corona. It also enables the discovery and precise characterization of potential proteoform biomarkers of breast cancer. The approach will eventually enhance our understanding of the protein corona, offer valuable insights into nanoparticle-biosystem interactions, and advance proteoform biomarker discovery.

Ginseng is widely praised for its benefits on cancer patients, often attributed to its metabolite compound K (CK). Here, we synthesized a derivative (CKD-4) that, compared with CK, exhibited enhanced cellular uptake, threefold greater cytotoxicity, and improved pharmacokinetics. CKD-4 induced significant growth inhibition on lung cancer patient-derived organoids, and on cell line-derived xenografts with minimal systemic toxicity. CKD-4 also suppressed orthotopic lung tumor growth in immunocompetent mice with enhanced antitumor immune infiltration. Using proteome integral solubility alteration and ProTargetMiner analyses, the mitochondrial phospholipid transfer protein PRELID3B was unbiasedly identified as a shared anticancer target of CK and CKD-4. PRELID3B is a potential pancancer therapeutic target and prognostic biomarker supported by cancer genetics and transcriptomics evidence. Both CK and CKD-4 stabilize PRELID3B in cellular thermal shift assay and bind PRELID3B with Kd of 23 µM and 5 µM, respectively, measured by biolayer interferometry. Multiomics analyses revealed that CK and CKD-4 share similar anticancer mechanisms, involving mitochondrial phospholipid depletion, integrated stress response activation, and immunomodulatory pathways induction associated with PRELID3B inhibition. This study provides the basis for the immunomodulatory and anticancer effects of ginseng metabolites through targeting PRELID3B, and illustrates the application of orthogonal proteomics in target identification of natural compounds.

The blood plasma proteome is a rich reservoir of potential biomarkers, yet its analysis is hindered by a protein concentration dynamic range exceeding ten orders of magnitude. Traditional fractionation and enrichment methods often present limitations in detecting rare proteins masked by the presence of abundant proteins. A promising frontier technology has emerged to overcome such limitations: leveraging nanoparticle protein coronas to enrich low-abundance plasma proteins and unveil previously inaccessible layers of the proteome. This Perspective highlights the capabilities of the protein corona technology in enhancing proteome coverage, addresses its current limitations and outlines future directions in proteoform analysis and causal inference.

Abstract The vaginal microbiome is a critical determinant of women’s health. We investigated the genetic basis of common vaginal microbiome species and their biofilm formation. Genomic analysis of Gardnerella vaginalis ( Gv ) revealed a fundamental phylogenetic split correlating with high- versus low-biofilm phenotypes, driven by clade-specific genomic islands and allelic variants. In a dual-species coculture model of five key vaginal bacteria, Gv achieved numerical dominance, triggering extensive, asymmetric proteomic reprogramming in partner species while showing limited shifts itself. Proteins from biofilm-associated modules showed functional divergence, supported by AI-predicted structural variations in a type II secretion/Tad pilus system, which is first discovered from Gv strains. Integrated metabolomics identified a methyl- β -carboline compound that is elevated in cocultures containing Prevotella bivia ( Pb ). This compound acts as a potent and selective inhibitor of Gv and Pb biofilms, sparing Lactobacillus crispatus . This work establishes a direct genomic basis for Gv virulence and demonstrates how interspecies interactions govern community dynamics and antimicrobial metabolite production. Highlights: Comprehensive genomic resource comparing with high-quality long-read whole genomes and reference Gardnerella vaginalis and Lactobacillus iners strains. Integrated multi-omics and functional analysis on the most common vaginal microbiome species using 16S rRNA gene sequencing, proteomics, metabolomics, and in vitro assays. Key phenotypes quantified, including biofilm formation and polymicrobial interactions. Conserved Type II Secretion/Tad Pilus System identified across all Gardnerella vaginalis strains, with AI-predicted structural modeling. Evaluation of growth inhibition using metabolites against a panel of relevant microbes, including vaginal microbes and opportunistic pathogens.

Understanding how cellular proteins interact with their environment, including endogenous and exogenous molecules, is critical for elucidating mechanisms of cellular regulation and drug action. Partial proteolysis-based techniques offer peptide-level resolution of ligand-induced conformational changes but are limited by modest proteome coverage and depth, as well as sensitivity to the experimental conditions. To overcome these limitations, we developed high-ratio partial proteolysis with carrier proteome (HOLSER), an efficient workflow that features an extended digestion time for reduced peptide yield variability as well as tandem mass tag multiplexing that includes full digests for enhanced proteome depth and sequence coverage, as well as higher precision of peptide abundance measurements. We demonstrate HOLSER capabilities of probing structural changes on the scale of specific binding sites for kinase target mapping, individual protein domains for structural mapping of the FKBP-mTOR complex in response to rapamycin, as well as global proteome structure profiling.

The protein corona influences the in vivo biodistribution of ionizable lipid nanoparticles (LNPs) in nucleic acid delivery, yet their structural architecture remains poorly defined. Using cryo-transmission electron microscopy, we visualized LNP-protein interactions in their native state. We show that, unlike the discrete "fuzzy" shells observed on hard nanoparticles, LNPs displayed no peripheral protein shell. Instead, controlled incubation and competitive "dual-particle" assays, supported by molecular dynamics simulations, indicate that LNP membranes undergo localized thickening and electron-dense remodeling consistent with lipoprotein integration rather than surface adsorption. Similar features were observed in extracellular vesicles, suggesting that this behavior is shared among lipid-based carriers, and proteomic analysis identified apolipoproteins as the dominant associated proteins. Together, these findings support a model in which the biological identity of LNPs arises through membrane remodeling rather than shell-like adsorption and provide a framework for the rational design of targeted nanomedicines.

Since 2018, ionizable lipid nanoparticles (LNPs) have revolutionized nucleic acid therapeutics. However, achieving potent extrahepatic delivery remains a formidable challenge, primarily due to rapid hepatic uptake driven by apolipoprotein adsorption. While analyzing the LNP protein corona is essential for engineering organ-specific tropism, these soft materials present unique analytical hurdles. Co-isolation of blood-borne contaminants, such as extracellular vesicles and lipoproteins, often masks the true corona composition. This perspective examines the critical need for refined proteomic strategies to distinguish genuine corona proteins from impurities. We propose tailored investigative approaches, suggesting the LNP protein corona significantly differs from the rigid shells observed on inorganic nanoparticles.

The antimicrobial resistance crisis necessitates structurally distinct antibiotics. While deep learning approaches can identify antibacterial compounds from existing libraries, structural novelty remains limited. Here, we developed a generative artificial intelligence framework for designing de novo antibiotics through two approaches: a fragment-based method to comprehensively screen >107 chemical fragments in silico against Neisseria gonorrhoeae or Staphylococcus aureus, subsequently expanding promising fragments, and an unconstrained de novo compound generation, each using genetic algorithms and variational autoencoders. Of 24 synthesized compounds, seven demonstrated selective antibacterial activity. Two lead compounds exhibited bactericidal efficacy against multidrug-resistant isolates with distinct mechanisms of action and reduced bacterial burden in vivo in mouse models of N. gonorrhoeae vaginal infection and methicillin-resistant S. aureus skin infection. We further validated structural analogs for both compound classes as antibacterial. Our approach enables the generative deep-learning-guided design of de novo antibiotics, providing a platform for mapping uncharted regions of chemical space.

The protein corona is a layer of biomolecules—primarily proteins—that adsorbs to nanoparticle (NP) surfaces in biological fluids. If the purpose of the NP is therapeutic, this can have a profound effect on its biological activity and function in vivo. Protein corona formation can also be exploited for diagnostic purposes and to differentially enrich proteins for biomarker discovery. For all of these applications, it is useful to determine which proteins, and which specific proteoforms, bind to different types of NP. The traditional mass spectrometry (MS)-based bottom-up proteomics does not accurately identify specific proteoforms within the protein corona. This limitation impedes the nanomedicine field’s ability to precisely predict the biological fate and pharmacokinetics of nanomedicines and their effectiveness in early-stage biomarker discovery and disease detection because many different proteoforms of the same gene could exist in the corona, and they have divergent biological functions. Here, we describe how to use capillary zone electrophoresis (CZE)–MS-based top-down proteomics to characterize the proteoform landscape of the protein corona. Our procedures detail the recovery of intact proteoforms from NP surfaces by using detergent-assisted proteoform elution and the measurement of these proteoforms by using CZE–tandem MS (MS/MS) and CZE–high-field asymmetric waveform ion mobility spectrometry (FAIMS)–MS/MS. The entire workflow is completed within 3–4 d. Using this protocol, hundreds of proteoforms from the protein corona of polystyrene NPs can be identified. Distinct protein corona proteoform profiles were observed from NPs with different physicochemical properties. The addition of FAIMS is beneficial for more in-depth proteoform characterization. It is important to understand how nanoparticles interact with biological systems and behave in vivo. This protocol describes top-down proteomics for protein corona profiling using advanced capillary zone electrophoresis–tandem mass spectrometry.

In contemporary studies on the role of the protein corona in specific biological applications, identifying correlation is widely used to draw conclusions from observations and statistical methods, yet it merely identifies associations without establishing a direct influence between variables. This over reliance on observation can lead to spurious connections where co-occurrence does not imply causation. In contrast, a causality-focused approach asserts the direct impact of one variable on another, offering a more robust framework for inference and the drawing of scientific conclusions. This approach allows researchers to better predict how changes in a nanoparticle’s physicochemical properties or biological conditions will affect protein corona composition and decoration, in turn affecting their safety and therapeutic/diagnostic efficacies. As a proof of concept, we explore the concept of “actual causality” (introduced by Halpern and Pearl) to mathematically prove how spiking small molecules, including metabolites, lipids, vitamins, and nutrients, into plasma can induce diverse protein corona patterns on identical nanoparticles. This approach significantly enhances the depth of plasma proteome profiling. Our findings reveal that among the various spiked small molecules, phosphatidylcholine was the actual cause of the observed increase in the proteomic depth of the plasma sample. By considering the concept of causality in the field of protein coronas, the nanomedicine community can substantially improve the ability to design safer and more efficient nanoparticles for both diagnostic and therapeutic purposes.

Posttranslational modifications (PTMs) are of significant interest in molecular biomedicine due to their crucial role in signal transduction across various cellular and organismal processes. Characterizing PTMs, distinguishing between functional and inert modifications, quantifying their occupancies, and understanding PTM crosstalk are challenging tasks in any biosystem. Studying each PTM often requires a specific, labor-intensive experimental design. Here, we present a PTM-centric proteome informatic pipeline for predicting relevant PTMs in mass spectrometry-based proteomics data without prior information. Once predicted, these in silico identified PTMs can be incorporated into a refined database search and compared to measured data. As a practical application, we demonstrate how this pipeline can be used to study glycoproteomics in oral squamous cell carcinoma based on the proteome profile of primary tumors. Subsequently, we experimentally identified cellular proteins that are differentially expressed in cells treated with multikinase inhibitors dasatinib and staurosporine using mass spectrometry-based proteomics. Computational enrichment analysis was then employed to determine the potential PTMs of differentially expressed proteins induced by both drugs. Finally, we conducted an additional round of database search with the predicted PTMs. Our pipeline successfully analyzed the enriched PTMs, and detected proteins not identified in the initial search. Our findings support the effectiveness of PTM-centric searching of MS data in proteomics based on computational enrichment analysis, and we propose integrating this approach into future proteomics search engines.