Glutaminase 1 (GLS1) is a crucial enzyme that serves as the initial rate-limiting factor in glutaminolysis, a metabolic process that releases various factors that influence biological processes such as development, differentiation, and immune responses. Several studies have systematically investigated the crucial role of GLS1 in cancer. However, there is a lack of a comprehensive understanding of the relationship between GLS1 and inflammation. In this review, we present a detailed examination of GLS1, and discuss its structure, function, and role in inflammatory pathways. Here, we summarize the evidence supporting GLS1's involvement in several inflammatory diseases and explore the potential therapeutic applications of GLS1 inhibitors. We found that GLS1 plays a crucial regulatory role in inflammation by mediating glutaminolysis. Targeting GLS1, such as through the use of GLS1 inhibitors, can effectively alleviate inflammation induced by GLS1. Furthermore, we highlight the challenges and opportunities associated with investigating GLS1 function and developing targeted inhibitors, and propose practical solutions that offer valuable insights for the functional exploration and discovery of potential therapeutics aimed at treating inflammatory diseases.

Medicinal and edible plants (MEPs) have attracted increasing interest worldwide due to their natural origin, reliable efficacy, and minimal side effects in recent years. However, the complex and fluctuating levels of inherent chemical constituents and exogenous hazardous contaminants have triggered widespread concerns about their efficacy and safety. Developing analytical methods for both active components and exogenous contaminants concealed in these samples is central to the quality evaluation, in which sample preparation is crucial. This paper systematically reviewed the evolution of standard sample preparation methods, microextraction techniques based on novel solvents and nanomaterials, and innovative integrated techniques from 2019. Accordingly, their merits and weaknesses were discussed by showing fruitful applications in identifying and quantifying active components in these plants. Further, successful applications for analyzing exogenous contaminants were prominently showcased, highlighting the management of pesticides, heavy metals, mycotoxins, and polycyclic aromatic hydrocarbons (PAHs). Finally, forthcoming trends in sample preparation techniques were delineated to illuminate the development and implementation of more advanced sample preparation technologies.

Exosomes, small vesicles secreted by a wide range of cells, are found extensively in animals, plants, and microorganisms. Their excellent biocompatibility, efficient delivery capacity, and ease of membrane crossing have drawn significant interest as promising drug delivery carriers. Compared with their animal-derived counterparts, plant-derived exosomes (PDEs), in particular, stand out for their lower toxicity to human tissues, diverse sources, and enhanced targeted delivery capabilities. Advances in both in-depth research and technological development have enabled scholars to isolate exosomes successfully from various plants, exploring their potential in clinical therapies. However, the precise identification of PDEs and their drug delivery mechanisms remains an area of ongoing investigation. This review synthesizes the latest developments in the biogenesis, extraction, separation, and identification of PDEs, along with their engineering modifications and drug-loading strategies. We also delve into the therapeutic applications of exosomes and their future potential in drug delivery, aiming to elucidate the targeted delivery mechanisms of PDEs and pave new paths for clinical drug treatment.

The rising prevalence of multidrug-resistant pathogens poses a substantial threat to global healthcare systems, demanding urgent therapeutic interventions. Microorganisms exhibit diverse resistance mechanisms against various classes of antibiotics, highlighting the urgent need to discover novel antimicrobial agents for combating bacterial infections. Anti-virulence therapy has emerged as a promising therapeutic strategy that neutralizes pathogens by targeting their virulence determinants. The strategies for screening virulence arresting drugs (VADs) in bacteria represent a multifaceted approach that involves elucidating molecular pathogenesis mechanisms of bacterial pathogenicity, identifying evolutionarily conserved virulence factors across different pathogens, and employing integrated approaches combining in silico prediction with experimental validation. Recent technological advancements have established standardized protocols for effective identification and validation of anti-virulence compounds. This review systematically examines contemporary screening methodologies, primarily focusing on quorum-sensing disruption and biofilm suppression strategies, including in silico screening, activity-based screening with bioassays, in vitro and in vivo models. Additionally, we emphasize the imperative for standardized preclinical validation through physiologically relevant animal models, while proposing framework recommendations for developing next-generation VAD screening platforms. This synthesis not only outlines current best practices but also proposes innovative avenues for future antimicrobial discovery research.

Antibody drugs, such as monoclonal antibodies and antibody-drug conjugates, have shown significant potential in treating diseases due to their high specificity and affinity. In vivo analysis of antibody drugs with non-invasive and real-time techniques is of importance to understand dynamic behavior of drugs within living organisms, and help evaluate their pharmacokinetics and efficacies. This review summarizes the advances and in vivo analysis methods of antibody drugs, including the techniques of radiolabeled imaging, near-infrared fluorescence imaging and surface-enhanced Raman spectroscopy. The principles, applications, and challenges of each technique are discussed, which provides insights for the development of antibody drugs and in vivo analytical methods.

Proteins are indispensable to all biological systems and drive life processes through activities that are intricately linked to their three-dimensional (3D) structures. Traditional proteomics often provides static snapshots of protein expression, leaving unanswered questions about how proteins respond to stimuli and affect cellular functions. Limited proteolysis coupled with mass spectrometry (LiP-MS) has emerged as a powerful technique for exploring protein structure and function under near-natural conditions. Studies have revealed that LiP-MS is invaluable for structural and functional proteomics because it offers novel insights into protein dynamics. In this review, we summarise the current applications of LiP-MS in diverse areas such as the discovery and identification of drug targets, metabolite action mechanisms, proteome dynamics, protein interactions, and disease biomarkers. We also address the critical challenges in ongoing research and discuss their broader implications for advancing our understanding of protein biology and drug discovery. LiP-MS holds significant promise for accelerating biomarker and therapeutic target development as well as advancing molecular biology research in animals, plants, and microorganisms.

Rheumatoid arthritis (RA) is a systemic autoimmune condition that leads to chronic arthritis, disability, and reduced lifespan. Current therapies show limited effectiveness and often cause severe side effects, with up to 50% of patients discontinuing disease-modifying antirheumatic drugs (DMARDs) due to unsatisfactory outcomes. Natural bioactive compounds (NBCs), such as glycosides, alkaloids, terpenoids, flavonoids, polyphenols, and coumarins, have gained attention for their immunomodulatory and anti-inflammatory properties. However, challenges like poor solubility, high dosage requirements, short action duration, and low tissue specificity hinder their clinical use. Nanoparticle (NP)-based delivery systems, including lipid NPs (LNPs), polymer carriers, and inorganic nanocarriers, have been designed to address these challenges through passive, active, and stimuli-responsive strategies. NBC-loaded NPs target immune dysfunction, synovial hyperplasia, bone destruction, angiogenesis, inflammation, and oxidative stress (OS) in RA. This article highlights recent advancements in NBCs for RA treatment, nanoformulation design, and targeted mechanisms, while addressing challenges and future directions in this field. The integration of cutting-edge nanotechnology has demonstrated significant potential to overcome traditional barriers such as low bioavailability and off-target effects through intelligent NPs design. Future research should enhance artificial intelligence (AI)-driven modeling to predict drug-nanocarrier interactions, develop biomarker frameworks for precision nanomedicine, and optimize RA management.

Targeted drug delivery platforms are designed to enable spatiotemporal precision in transporting therapeutic agents to disease-specific sites, thereby optimizing therapeutic efficacy and mitigating off-target adverse effects. Despite their clinical promise, these platforms remain hindered by substantial translational barriers. Macrophages, with inherent biocompatibility and intrinsic tropism toward inflamed/diseased tissues, are critically involved in diverse pathological processes. Macrophage-based drug delivery systems (MDDSs) have emerged as promising platforms engineered via therapeutic cargo loading onto intact cells, cell-membrane coatings, extracellular vesicles (EVs), or hitchhiking mechanisms. This review delineates existing MDDS platforms, critically analyzing their respective merits and constraints. We further elucidate therapeutic mechanisms and clinical implementations of MDDSs for cancer, atherosclerosis (AS), and central nervous system (CNS) disorders, while establishing a systematic taxonomy of their biomedical applications. Specifically, we highlight the transformative potential of gene-editing technologies (exemplified by chimeric antigen receptor macrophage (CAR-M) therapy and antigen-independent strategies) in innovating next-generation MDDS architectures. We summarize state-of-the-art developments, persisting translational hurdles, and optimization roadmaps for MDDSs, providing a conceptual framework to guide their translational advancement.

Colorectal cancer (CRC) originates from biological events caused by gene mutations in normal intestinal epithelial cells (IECs). Sorting nexin 10 (SNX10) is a tumor suppressor in CRC that is involved in regulating chaperone-mediated autophagy (CMA) activity, which is implicated in the pathogenesis of CRC and glycolysis process. DEP domain containing 5 (DEPDC5) is a negative upstream regulator of mammalian target of rapamycin complex 1 (mTORC1). α-hederin has anti-CRC effects. We previously found that SNX10 knockdown in normal human IECs promoted glycolysis and decreased DEPDC5 expression, which was reversed by α-hederin. However, the specific mechanism has not yet been elucidated. Here, we aimed to investigate the specific regulatory mechanism of SNX10 on DEPDC5 expression, and the action of α-hederin on this process. We demonstrated that the degradation of DEPDC5 protein was accelerated after SNX10 knockdown, causing the activation of the mTORC1 pathway, which relied on CMA activation and lysosomal function enhancement. SNX10 interacted with DEPDC5 and recruited it to lysosomes for degradation, and the glycolysis level mediated by mTORC1 was elevated. Additionally, these phenotypes in shSNX10 IECs were compromised by SNX10 rescue. Moreover, α-hederin bound to the SNX10–DEPDC5 complex and impaired the interaction between SNX10 and DEPDC5, thereby inhibiting CMA-mediated DEPDC5 degradation, impairing the aberrant activation of mTORC1 signaling, and eventually reversing the elevation of glycolysis caused by SNX10 knockdown. Overall, we are the first to demonstrate that SNX10-mediated DEPDC5 degradation is a novel strategy for malignant transformation of normal human IECs, with α-hederin regulated during this process.

Barrier tissues such as the endothelium are critical in the regulation of mass transfer throughout the body. Trans-endothelium/epithelium electrical resistance (TEER) is an important bioelectrical measurement technique to monitor barrier integrity. Although available on the market, TEER sensors are usually expensive and bulky and do not allow customization around experimental setups like specific microfluidic settings. We recently reported a customizable TEER sensor built on Arduino. In this paper, we significantly advanced a new generation of TEER sensors characterized by 1) a large dynamic range of 242–11,880 Ω·cm² with high accuracy (>95%), which covers common needs for TEER studies, 2) a coupling three-dimensional (3D)-printed microfluidic system enabling modular cell integration and flow-based barrier studies, 3) customizable on-off cycles to significantly reduce cell exposure to the current, and 4) automated continuous measurements with customizable intervals. With this sensor system, we investigated how doxorubicin could impair the endothelium layer’s permeability, at a 1-min interval for 24 h. Endothelium toxicity is a new research direction under cardiotoxicity, with many aspects unknown. We found that a clinically relevant dosage did not change the endothelium integrity significantly until approximately 16 h of treatment, after that, the TEER started to drop (showing higher permeability), followed by a slight restoration of its barrier integrity. With an excess dosage (2.5 μM), the TEER started to drop significantly after 5 h and did not show recovery afterward, indicating endothelium toxicity. Overall, we report a new TEER sensor that can monitor continuous drug toxicity on barrier tissues. The customizable features make it translational for various other studies, such as personalized dosage determination on stem cell-derived tissue barriers, and transient barrier permeability variations under diseased conditions.

Vascular endothelial senescence is an important pathophysiological factor in the development and exacerbation of cardiovascular health problems linked to diabetes mellitus (DM). Accumulating evidence confirms that 1,8-cineole has multiple pharmacological properties, including anti-inflammatory, anti-microbial, and antioxidant activities. We investigated whether 1,8-cineole could ameliorate cardiovascular diseases and endothelial dysfunction, as the pharmacological properties and mechanism of diabetic vascular ageing remain unknown. Our results revealed notable senescence biomarkers in both in vivo and in vitro models. Treatment with 1,8-cineole alleviated lipid profiles and vascular senescence in mice with DM. Additionally, bioinformatics analysis suggested that peroxisome proliferator-activated receptor-γ (PPAR-γ) plays a crucial role in DM and ageing. We confirmed the binding capacity PPAR-γ with 1,8-cineole. Accordingly, experiments with the PPAR-γ agonist rosiglitazone, the PPAR-γ inhibitor GW9662, and PPAR-γ siRNA were performed to validate the pharmacological characteristics of 1,8-cineole. Finally, we clarified that 1,8-cineole can directly target PPAR-γ protein, as verified by cellular thermal shift assay, drug affinity responsive target stability, and surface plasmon resonance analyses. Taken together, these results provide the first evidence that 1,8-cineole ameliorates DM-induced vascular endothelial ageing via stabilising PPAR-γ protein by promoting deubiquitination at the Lys-466 site.

A supramolecular system of active pharmaceutical ingredients (APIs) can modify the physicochemical properties and enhance the synergistic efficacy of their components; however, the relevant underlying mechanisms in vivo remain unclear. This study employed a metabolomics-driven approach, combined with biological validation, to investigate the synergistic mechanisms of API-based supramolecular systems. Metabolic dysfunction exacerbates insulin resistance and obesity, contributing to hepatic steatosis and cardiac hypertrophy. A novel sodium-dependent glucose transporter 2 (SGLT-2)/peroxisome proliferator-activated receptor-γ (PPAR-γ) dual receptor (dapagliflozin-pioglitazone (DAP-PIO)) supramolecular system was selected as the model to explore the synergistic mechanism involved in the treatment of metabolic dysfunctions, diabetes and obesity. First, metabolomics analyses were performed to compare the effects of a simple physical mixture (PM) of DAP and PIO with the DAP-PIO supramolecular system after absorption into the bloodstream. The results demonstrated significant differences, with the supramolecular system activating the phosphatidylinositol 3-kinase (PI3K)/protein kinase B (AKT) and adenosine monophosphate-activated protein kinase (AMPK) signaling pathways. Ceramide (Cer), a key metabolite in sphingolipid metabolism, emerged as a critical mediator. Subsequently, the mechanisms underlying the DAP-PIO supramolecular system’s hypoglycemic effects and its ability to ameliorate hepatic steatosis and myocardial hypertrophy by reducing insulin resistance were evaluated and confirmed. These findings provide an innovative strategy for developing SGLT-2/PPAR-γ dual-receptor supramolecular systems to enhance the therapeutic outcomes for diabetes and obesity.

The Wnt/β-catenin signaling pathway is renowned for its contribution to the immunosuppressive microenvironment in non-small cell lung cancer (NSCLC). Consequently, inhibiting this pathway has emerged as a promising strategy to enhance immune activation and reinstate T cell responses in cancer treatment. In this study, we initially investigate the metabolic characteristics of Wnt-hyperactivated NSCLC using mass spectroscopic detection in a mouse in-situ model and unveil its significant feature of acid accumulation at tumor sites. Building upon this discovery, we design an acid-sensitive peptide-carnosic acid (CA) supramolecular droplet (Pep1@CA), which leverages the acidic microenvironment characteristic of NSCLC for controlled release. By doing so, we aim to enhance targeting efficiency while minimizing off-target effects. As anticipated, Pep1@CA demonstrates potent tumor-specific inhibition of the Wnt signaling pathway and effectively reactivates T cell immunity in Wnt-hyperactivated NSCLC. Importantly, comprehensive in vivo evaluations reveal significant antitumor efficacy alongside excellent biosafety profiles. Collectively, this study provides a therapeutic strategy with promising clinical translational potential for targeting the Wnt signaling pathway and offers theoretical support for its application in immunotherapy. This innovative approach underscores that targeting pathways beyond traditional immunotherapy can also activate tumor immunity, thereby expanding the potential of cancer immunotherapy.

We developed MaxQsaring, a novel universal framework integrating molecular descriptors, fingerprints, and deep-learning pretrained representations, to predict the properties of compounds. Applied to a case study of human ether-à-go-go-related gene (hERG) blockage prediction, MaxQsaring achieved state-of-the-art performance on two challenging external datasets through automatic optimal feature combinations, and successfully identified top 10 important interpretable features that could be used to model a high-accuracy decision tree. The models' predictions align well with empirical hERG optimization strategies, demonstrating their interpretability for practical utilities. Deep learning pre-trained representations have been demonstrated to exert a moderate influence on enhancing the performance of predictive models. Nevertheless, their impact on augmenting the generalizability of these models, particularly when applied to compounds possessing novel scaffolds, appears to be comparatively minimal. MaxQsaring excelled in the Therapeutics Data Commons (TDC) benchmarks, ranking first in 19 out of 22 tasks, showcasing its potential for universal accurate compound property prediction to facilitate a high success rate of early drug discovery, which is still a formidable challenge.

Melanoma, a common malignant skin tumor, faces challenges with multidrug resistance and high recurrence rates. Combining photodynamic therapy (PDT) and immunotherapy offers a promising personalized treatment approach. However, poor water solubility and significant side effects of photosensitizers and immune checkpoint inhibitors (ICIs) limit their application. Enhancing delivery efficiency while reducing adverse effects is crucial. Herein, we formulate BM@HSSC nanoparticles (NPs), which consist of a reduction-responsive hyaluronic acid (HA) backbone modified with photosensitizer chlorin e6 (Ce6) and loaded with the programmed cell death-ligand 1 (PD-L1) inhibitor BMS-1. This system synergistically integrates PDT, immunogenic cell death (ICD), and immunotherapy for melanoma treatment. BM@HSSC NPs target and accumulate at the tumor site via the CD44 receptor. The disulfide bonds (-S-S-) in the NPs react with high glutathione (GSH) concentrations in tumor cells, rapidly releasing Ce6 and BMS-1. Under 660 nm laser irradiation, BM@HSSC NPs generate cytotoxic reactive oxygen species (ROS), inducing cell apoptosis and triggering ICD via PDT damage-associated molecular patterns (DAMPs) and tumor-associated antigens (TAAs) released from ICD promote dendritic cell (DC) maturation, enhancing antigen presentation and activating cytotoxic T lymphocytes (CTLs). Meanwhile, BMS-1 blocks the programmed cell death-1 (PD-1)/PD-L1 pathway, countering the immunosuppressive tumor microenvironment (iTME) and inhibiting tumor cell immune escape. This strategy amplifies antitumor immune responses by enhancing immunogenicity and synergizing with ICIs, resulting in robust antitumor efficacy.

Evidences indicate that farnesoid X receptor (FXR) activation mitigates non-alcoholic fatty liver disease (NAFLD) by reducing endoplasmic reticulum (ER) stress. However, the mechanisms underlying FXR-ER stress interactions in combating NAFLD remain obscure. Moreover, few phytochemicals have been noted to improve NAFLD through this pathway. Here, we found that FXR activation directly induces the transcription of sarco/endoplasmic reticulum Ca²⁺ ATPase 2 (SERCA2), which acts as an ER stress repressor. This process leads to the dephosphorylation of the eukaryotic translation initiation factor 2 subunit α (eIF2α) within hepatocytes, consequently alleviating ER stress. Furthermore, through drug binding assays, luciferase reporter gene testing, gene expression analysis and biochemical evaluation, we identified the phytochemical atractylenolide II (AT-II) as a novel FXR agonist that effectively triggers SERCA2 activation. Our results showed AT-II effectively supresses accumulation of lipids and ER stress in palmitic acid-induced hepatocytes. In in vivo experiments, we demonstrated that AT-II attenuates fatty liver in diet- or chemical-induced NAFLD mouse models. Additionally, we showed that AT-II corrects diet-induced obesity, serum dyslipidemia, metabolic complications, and insulin resistance. Mechanistically, AT-II reduces ER stress, lipogenesis and inflammation and improves hepatic insulin signaling through stimulation of the hepatic FXR-SERCA2-eIF2α axis in mice. This conclusion was further reinforced by Serca2 knockdown both in vivo and in vitro, as well as FXR silencing in hepatocytes. Our findings provide new insights into the FXR-ER stress interplay in the control of NAFLD and suggest the potential of AT-II as an FXR agonist for the treatment of NAFLD through SERCA2 activation.

Various deep learning based methods have significantly impacted the realm of drug discovery. The development of deep learning methods for identifying novel structural types of active compounds has become an urgent challenge. In this paper, we introduce a self-supervised representation learning framework, i.e., Geometry-based Bidirectional Encoder Representations from Transformers (GEO-BERT). GEO-BERT considers the information of atoms and chemical bonds in chemical structures as the input, and integrates the positional information of the three-dimensional conformation of the molecule for training. Specifically, GEO-BERT enhances its ability to characterize molecular structures by introducing three different positional relationships: atom-atom, bond-bond, and atom-bond. By benchmarking study, GEO-BERT has demonstrated optimal performance on multiple benchmarks. We also performed prospective study to validate the GEO-BERT model, with screening for DYRK1A inhibitors as a case. Two potent and novel DYRK1A inhibitors (IC₅₀:<1 μM) were ultimately discovered. Taken together, we have developed an open-source GEO-BERT model for molecular property prediction (

) and proved its practical utility in early-stage drug discovery.

Reprogramming oncogenic signaling pathways to generate anti-tumor effects is a promising strategy for targeted cancer intervention, without significant off-target effects. Although reprogramming multi-oncoprotein interactions in a single signaling pathway axis has been shown to achieve sustained efficacy, there are several challenges that limit its clinical application. Herein, we transformed the mouse double minute 2 homolog (MDM2)-heat shock cognate protein 70 (HSC70) axis, a tumor-promoting pathway, into an activator of anti-tumor immunity using the Path-editor, an artificial selenoprotein. Once it enters the cell, Path-editor decomposes into PMI and PPI peptides: PMI inhibits MDM2-mediated p53 degradation and promotes HSC70 expression, while PPI binds to HSC70, enabling its ability to selectively degrade the programmed cell death ligand 1 (PD-L1). As a proof of concept, we tested its performance in microsatellite-stable (MSS) colorectal cancer, which typically displays limited responsiveness to immunotherapy. The results indicated that Path-editor effectively attenuated PD-L1 expression and reversed immune evasion in both CT26 allografts and humanized patient-derived tumor xenograft (PDX) models, thereby inhibiting tumor progression with high biosafety. Therefore, this paper introduces Path-editor as a paradigm for reprogramming oncogenic multi-protein pathways, utilizing selenium-assisted approach to achieve the rapid design of tumor-specific pathway editors. This strategy is expected to reverse immune escape in MSS colorectal cancer and treat difficult malignancies.

Diabetic foot ulcer (DFU) is an increasing global burden due to the rising prevalence of diabetes, and no specific pharmacological targets or satisfactory drugs are currently available for this devastating ailment. In this study, naringenin (NAR) was found to accelerate diabetic wound healing in diabetic C57BL/6J wild-type (WT) mice by reducing oxidative stress, as assessed through histological assay. NAR also alleviated the inhibition of proliferation, inflammation, cell senescence, and apoptosis in HaCaT cells induced by high glucose (HG). Mechanistically, the beneficial effects of NAR on wound healing are dependent on the E3 ubiquitin-protein ligase parkin (Parkin/PRKN/Prkn). NAR upregulated the expression level of Parkin and promoted its mitochondrial translocation, thereby activating Parkin-mediated mitophagy and maintaining mitochondrial quality control (MQC). Moreover, the wound healing-promoting effects of NAR were significantly diminished in Parkin knockdown HaCaT cells and Prkn knockout (Prkn^−/−) DFU mice. Inhibition of NAR binding to estrogen receptors (ERs) using tamoxifen (TAM) abolished the protective effects of NAR in HG-induced HaCaT cells. The luciferase reporter assay confirmed that NAR enhanced ERs binding to the estrogen response element (ERE), thereby upregulating Parkin transcription. Additionally, the cellular thermal shift assay (CETSA) revealed that NAR specifically bound to ERα. In conclusion, NAR promoted DFU wound healing by enhancing Parkin-mediated mitophagy via binding to ERα, highlighting its potential as a promising therapeutic candidate.

Therapeutic monoclonal antibodies (mAbs) have garnered significant attention for their efficacy in treating a variety of diseases. However, some candidate antibodies exhibit non-specific binding to off-target proteins or other biomolecules, leading to high polyreactivity, which can compromise therapeutic efficacy and cause other complications, thereby reducing the approval rate of antibody drug candidates. Therefore, predicting the polyreactivity risk of therapeutic mAbs at an early stage of development is crucial. In this study, we fine-tuned six pre-trained protein language models (PLMs) to predict the polyreactivity of antibody sequences. The most effective model, named PolyXpert, demonstrated a sensitivity (SN) of 90.10%, specificity (SP) of 90.08%, accuracy (ACC) of 90.10%, F1-score of 0.9301, Matthews correlation coefficient (MCC) of 0.7654, and an area under curve (AUC) of 0.9672 on the external independent test dataset. These results suggest its potential as a valuable in-silico tool for assessing antibody polyreactivity and for selecting superior therapeutic mAb candidates for clinical development. Furthermore, we demonstrated that fine-tuned language model classifiers exhibit enhanced prediction robustness compared with classifiers trained on pre-trained model embeddings. PolyXpert can be easily available at

.