A Five-Gene Signature Identified by Integrating Single-Cell Analysis and Machine Learning Benchmarking Enables Rheumatoid Arthritis Diagnosis and Resveratrol Target Discovery

A Five-Gene Signature Identified by Integrating Single-Cell Analysis and Machine Learning Benchmarking Enables Rheumatoid Arthritis Diagnosis and Resveratrol Target Discovery | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article A Five-Gene Signature Identified by Integrating Single-Cell Analysis and Machine Learning Benchmarking Enables Rheumatoid Arthritis Diagnosis and Resveratrol Target Discovery Yan Zhou¹, Jia Guo², Xiaoqing Bao¹, Nan Yu¹ This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9098862/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background Rheumatoid arthritis (RA) is a systemic autoimmune disease featured by persistent synovial inflammation and multi-organ injury, with T lymphocytes as the core driver of its pathological progression. However, the molecular mechanisms of abnormal T cell functional regulation and the exact biological targets of resveratrol (a natural immunomodulatory polyphenol) in RA synovial microenvironment remain unclear. This study aimed to construct a T cell-focused RA diagnostic classifier and identify resveratrol-modulated molecular targets via multi-omics integration. Methods We integrated scRNA-seq and bulk transcriptome data of RA synovial tissues to characterize T cell-specific gene expression, intersected resveratrol targets from pharmacological databases with RA-related differentially expressed genes, and performed systematic benchmarking of 10 machine learning algorithms to screen stable diagnostic gene signatures (validated in external cohorts). CIBERSORT quantified synovial immune cell infiltration, and molecular docking evaluated resveratrol-target protein binding interactions. Results Distinct T cell subsets with differential resveratrol target gene expression were identified in RA synovium. A 5-gene signature ( PTTG1 , NUSAP1 , PDE4D , ITGA4 , BIRC3 ) was screened, showing excellent diagnostic performance (training cohort AUC = 0.983; external cohorts GSE55457 AUC = 0.908, GSE77298 AUC = 0.902). Its expression was positively correlated with activated T cell and pro-inflammatory macrophage infiltration. Molecular docking revealed stable binding of resveratrol to the five target proteins (binding free energy: -5.5 to -8.0 kcal/mol). Conclusion We constructed a robust T cell-centered 5-gene signature for RA diagnosis, and these five genes are potential molecular targets of resveratrol in RA immunotherapy. This study provides a novel molecular basis and research direction for developing synovitis-targeted therapeutic strategies for RA. Rheumatoid arthritis Single-cell RNA sequencing T cells Machine learning benchmarking Resveratrol Molecular docking Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Highlight Box Key findings Five immune-related genes ( BIRC3 , ITGA4 , NUSAP1 , PTTG1 , PDE4D ) were identified as a robust diagnostic signature for rheumatoid arthritis (RA), and a multi-algorithm machine learning model demonstrated high and stable diagnostic performance across training and external validation cohorts. What is known and what is new? T cells and inflammatory signaling are central to the pathogenesis of rheumatoid arthritis. This study integrates single-cell analysis, machine learning benchmarks, and computational pharmacology to identify a five-gene signature for RA diagnosis and uncover potential resveratrol targets. What is the implication, and what should change now? The five-gene signature could serve as a reliable tool for early and accurate diagnosis of rheumatoid arthritis. The identified signature and resveratrol targets offer new insights for mechanistic research and personalized treatment of RA. Introduction Rheumatoid arthritis has a global prevalence of approximately 0.5% to 1% in the adult population, and its typical pathological feature is persistent synovial inflammation that can induce cartilage erosion and irreversible articular damage if not effectively intervened in the early stage [ 1 , 2 ]. The pathogenesis of RA is a complex process jointly mediated by genetic susceptibility, environmental inducing factors and immune system disorder, with all these pathological factors converging and exerting effects in the synovial tissue microenvironment [ 3 , 4 ]. Although conventional disease-modifying antirheumatic drugs (DMARDs) and biological agents have greatly enriched the clinical therapeutic strategies for RA, a considerable proportion of patients still cannot obtain satisfactory disease control effects in clinical practice [ 5 , 6 ]. T lymphocytes act as the core regulatory component of the inflammatory network in RA, exerting their pathological effects through antigen recognition, pro-inflammatory cytokine secretion, and functional crosstalk with other immune cells in the synovial microenvironment [ 7 – 9 ]. The Th1 and Th17 subsets of CD4⁺ helper T cells promote the development and persistence of synovial inflammation by secreting pro-inflammatory cytokines such as IFN-γ, IL–17 and TNF-α [ 10 , 11 ]. In contrast, regulatory T cells (Tregs) in RA patients exhibit impaired immune suppressive function, which fails to effectively inhibit excessive inflammatory responses and leads to persistent synovial inflammation [ 12 , 13 ]. Although our grasp of RA pathogenesis has deepened considerably, molecular biomarkers suitable for routine diagnosis remain elusive [ 14 , 15 ]. Clinicians currently diagnose RA using ACR/EULAR criteria that blend symptoms, imaging, and serology—chiefly rheumatoid factor (RF) and anti-citrullinated protein antibodies (ACPA) [ 16 ]. Neither marker is ideal: RF specificity hovers around 60–80% because titers rise in other autoimmune states and even in healthy older adults; ACPA, although more specific, is undetectable in ~ 30% of patients (seronegative RA) [ 17 , 18 ]. Worse still, serological titers seldom mirror real-time disease activity or forecast therapeutic outcomes [ 19 , 20 ]. High-throughput transcriptomics now permits discovery of gene-expression signatures that may aid diagnosis and prognosis [ 21 , 22 ]. Several research teams have proposed interferon response modules and inflammation-related pathway gene panels as potential diagnostic biomarkers for RA [ 23 , 24 ]. However, these gene signatures have obvious limitations in clinical application: they are mostly derived from peripheral blood samples rather than the pathogenic site of RA (synovial tissue), their detection results cannot be stably replicated in different patient cohorts, and their intrinsic biological correlation with the pathophysiological process of RA remains unclear [ 25 , 26 ]. A robust, synovium-derived signature with clear biological underpinnings is therefore still needed. The stilbene resveratrol (trans-3,5,4′-trihydroxystilbene), found in grapes, berries, and peanuts, has attracted interest owing to reported anti-inflammatory and immunomodulatory effects [ 27 , 28 ]. In preclinical RA models, resveratrol has been shown to alleviate synovial inflammation through multiple mechanisms, including inhibiting the activation of NF-κB signaling pathway, reducing the secretion of pro-inflammatory cytokines, and promoting the differentiation and functional enhancement of Tregs [ 29 , 30 ]. Nevertheless, the specific molecular targets of resveratrol that mediate these anti-inflammatory and immunomodulatory effects in the RA synovial microenvironment have not been clearly identified [ 31 , 32 ]. Network pharmacology wedded to transcriptomics provides a rational route to map resveratrol targets onto RA-dysregulated genes [ 33 ]. Machine learning now underpins many efforts to distill high-dimensional omics data into clinically useful classifiers [ 34 , 35 ]. Early studies on RA gene biomarkers mostly constructed diagnostic models based on a single machine learning algorithm, which easily leads to overfitting of the discovery cohort data and thus the model cannot be effectively validated in external independent cohorts [ 36 , 37 ]. In contrast, the simultaneous benchmarking of multiple machine learning algorithm combinations can effectively avoid overfitting and construct a diagnostic model with better generalization ability in unknown patient cohorts [ 38 , 39 ]. To address the above-mentioned deficiencies in current RA biomarker research and resveratrol target exploration, we designed an integrated multi-omics analysis pipeline for this study. Our specific research objectives are as follows: (i) to analyze the heterogeneity of T cell subsets in RA synovial tissue by scRNA-seq technology; (ii) to perform intersection analysis of T cell-specific marker genes, RA-related dysregulated genes and known resveratrol molecular targets; (iii) to screen a stable RA diagnostic gene signature by benchmarking multiple machine learning algorithms; (iv) to explore the clinical relevance of this gene signature through immune cell infiltration profiling and molecular docking simulation. Materials and Methods Data Acquisition and Single-Cell RNA Sequencing Analysis The scRNA-seq dataset (GSE296117) derived from RA synovial biopsy samples was retrieved from the Gene Expression Omnibus (GEO) database, and we implemented all data preprocessing procedures with the Seurat R software package. We applied strict quality control to exclude low-quality cells, with the filtering criteria defined as 200 < n Feature RNA < 4000 and mitochondrial gene expression proportion < 10%. Following data normalization, unsupervised clustering was performed to distinguish distinct cell populations, and each population was annotated based on canonical marker genes compiled in the Cell Marker2 database. Differential Expression Analysis and Drug Target Prediction We designated the bulk transcriptomic dataset GSE89408 as the training cohort for differential expression analysis, and the limma software package was utilized to compare gene expression differences between RA patients and healthy controls, with the threshold for differentially expressed genes set as adjusted P 1. In parallel, we compiled potential molecular targets of resveratrol from the Comparative Toxicogenomics Database (CTD) and Gene Cards, and leveraged the chemical structure of resveratrol to predict supplementary targets via the Super-PRED and Swiss Target Prediction online platforms. Identification of Key Candidate Genes The Venn Diagram software package was adopted to compute the intersection of three gene sets, namely (1) T cell-specific marker genes identified from scRNA-seq data; (2) RA-related differentially expressed genes (DEGs) obtained from bulk transcriptomic analysis; and (3) predicted resveratrol target genes. The overlapping genes were designated as key therapeutic targets and subjected to Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis using cluster Profiler . Machine Learning Benchmarking and Model Construction For the purpose of constructing a robust and reliable RA diagnostic gene signature, we adopted an integrated machine learning analytical framework with the GSE89408 dataset as the training cohort. A total of 19 machine learning algorithms covering regularization-based, tree-based, linear, non-linear and ensemble categories were combined into 120 distinct classifier models with optimized hyperparameter combinations using the caret and glmnet packages, including lasso, Stepglm, glm Boost, Enet, rf, svm Radial, svm Linear, kknn, gbm, pls Rglm, lda, knn, nnet, naive_bayes, pls, step QDA, glm Step AIC, step LDA, and Logit Boost. To minimize the interference of random sampling bias on model performance evaluation, each constructed classifier model was subjected to strict 10-fold cross-validation with three repeated validation cycles. The model with the highest average value of the area under the receiver operating characteristic (ROC) curve in all cross-validation experiments was defined as the optimal diagnostic model for subsequent research. We conducted further validation on the diagnostic efficacy and stability of the identified optimal model using two independent external datasets, GSE55457 and GSE77298. Immune Infiltration and Single-Gene Gene Set Enrichment Analysis (GSEA) We quantified the infiltration levels of 22 immune cell subsets in the GSE89408 dataset via the CIBERSORT algorithm, with parameters set as permutation number = 1000, confidence interval = 95%, and deconvolution results with P < 0.05 deemed statistically significant and reliable for subsequent analyses. Pearson coefficients linked z-scored signature expression to cell-type fractions (two-tailed P < 0.05). Single-gene gene set enrichment analysis (GSEA) was performed with the cluster Profiler package (KEGG database v7.5.1; 1000 permutations) to explore the biological pathways associated with each gene in the diagnostic signature. Molecular Docking Simulation The three-dimensional structures of target proteins were obtained from the RCSB PDB database, while the molecular coordinate data of resveratrol were retrieved from the ZINC database. We used Open Babel and Auto Dock Tools software to preprocess ligand (resveratrol) and receptor (target protein) files, including water molecule removal and hydrogen atom addition. Binding affinities were calculated with Auto Dock Vina 1.5.6, and the docked poses were visualized using Py MOL software. Results Functional Enrichment, Interaction Network, and Disease Relevance of T Cells in Rheumatoid Arthritis Synovial Tissue T lymphocytes occupy a central position in adaptive immune responses and represent key drivers of the autoimmune pathological processes underlying rheumatoid arthritis (RA). Previous research has demonstrated that hyperactive T cells promote disease progression via multiple mechanisms, including pro-inflammatory cytokine secretion, functional crosstalk with B lymphocytes, and direct cytotoxicity against articular joint tissues. Against this background, we set out to investigate the specific functional roles and underlying regulatory networks of distinct T-cell subsets within the RA synovial microenvironment. Single-cell RNA sequencing (scRNA-seq) was employed to dissect the cellular composition of RA synovial tissue samples. Through unsupervised clustering analysis, we successfully identified seven distinct cell lineages: T lymphocytes, B lymphocytes, macrophages, fibroblasts, neutrophils, natural killer (NK) cells, and dendritic cells (Figure 1A). Each cell subset was further annotated based on well-established lineage-specific marker genes (Figure 1B). Notably, the relative proportion of each cell type exhibited substantial inter-patient variability (Figure 1C), which highlights the inherent cellular heterogeneity characteristic of RA pathogenesis. A heatmap illustrating the expression of key lineage marker genes (Figure 1D) revealed that IL7R, FOXP3, and CXCL13 were predominantly expressed in T-cell populations. Subsequent KEGG pathway enrichment analysis (Figure 1E) further linked T cells to multiple RA-associated biological pathways, such as Th17, Th1, and Th2 subset differentiation, hematopoietic lineage development, cytokine-receptor interactions, and cell adhesion molecule signaling pathways. We constructed a protein-protein interaction network from these KEGG pathways (Figure 1F) to map how T-cell molecules interact in the RA environment. Six functional modules emerged: hematopoietic lineage, Th17 differentiation, RA pathogenesis, Th1/Th2 differentiation, cytokine signaling, and cell adhesion molecule interactions. Several hub genes— IL7R , CXCL13 , IL2RA , CD40 , ICAM1 , and PDCD1 —stood out due to their extensive connections. Their central positions in the network point to key roles in T-cell differentiation, inflammatory signaling, and cell adhesion, making them attractive targets for therapeutic intervention. Transcriptional Dysregulation and Target Intersection We conducted differential expression analysis on the GSE89408 dataset to delineate the molecular expression profile of RA synovial tissue at the bulk transcriptomic level. Unsupervised hierarchical clustering of genome-wide expression profiles revealed a pronounced separation between RA patients and healthy controls, as shown in the heatmap (Figure 2A). The clustering dendrogram exhibited distinct disease-associated grouping, where RA samples and normal control samples formed independent branches respectively. This finding confirms that RA synovial tissue carries a distinct transcriptional signature that can be reliably distinguished from normal tissue. Using stringent statistical cutoffs (adjusted P 1), we identified 9,448 DEGs. Of these, 1,254 were upregulated and 119 were downregulated in RA samples compared to healthy controls (Figure 2B). The volcano plot showed that most statistically significant transcripts clustered in the upregulated region, indicating that transcriptional activation dominates the RA synovial profile. To pinpoint potential pharmacological targets of resveratrol, we compiled resveratrol-associated genes from four established databases: the CTD, Gene Cards, Super-PRED, and Swiss Target Prediction. CTD returned the largest set of resveratrol-related genes (n = 7,280), substantially more than Gene Cards (n = 1,254), Super-PRED (n = 119), or Swiss Target Prediction (n = 100) (Figure 2C). We then intersected three gene lists: T cell-specific markers (n = 230), bulk RNA-seq DEGs (n = 9,448), and resveratrol-related targets (n = 7,713). This analysis yielded 68 genes common to all three sets (Figure 2D). Functional Characterization and Protein-Protein Interaction (PPI) Network GO enrichment analysis showed that the 68-gene set is functionally concentrated in immune regulation and inflammatory response processes central to RA. At the biological process level, genes involved in T cell activation, lymphocyte proliferation, immune synapse assembly, and cytokine production were enriched. Cellular component analysis revealed localization to immune synapses and T cell receptor complexes. Molecular function enrichment highlighted immune receptor binding and protein kinase binding activities (Figure 3A and 3B). KEGG pathway analysis demonstrated that these core genes cluster in classical inflammatory signaling cascades—TNF, NF-κB, and JAK–STAT pathways—as well as T cell receptor signaling (Figure 3D and 3E). This enrichment pattern suggests the 68 genes may collectively drive the onset and persistence of synovial inflammation in RA by modulating these signaling networks. PPI network analysis revealed complex regulatory interactions among the core genes, with CD3E and ZAP70 emerging as hub genes with substantially higher connectivity than other network members (Figure 3E). Establishment of a Robust 5–Gene Diagnostic Model Comprehensive performance benchmarking was carried out on 120 classifier combinations to screen the optimal diagnostic model, and the LASSO + PLS combined algorithm achieved the highest average diagnostic accuracy (Figure 4A). This model identified a 5-gene signature: PTTG1 , NUSAP1 , PDE4D , ITGA4 , and BIRC3 . In the training set, the model achieved an AUC of 0.983 (Figure 4B). External validation in cohorts GSE55457 and GSE77298 returned AUCs of 0.908 and 0.902, respectively (Figure 4C–D). Decision curve analysis and calibration curve analysis were performed to verify the clinical applicability of the 5-gene diagnostic model, and the results confirmed its favorable clinical utility (Figure 4E–F). Correlation Between the 5–Gene Diagnostic Signature and Synovial Immune Microenvironment in Rheumatoid Arthritis Using CIBERSORT, we quantified the infiltration of 22 immune cell subsets in RA and normal synovial tissues from the GSE89408 dataset. Pro-inflammatory immune cells were significantly more abundant in RA synovium compared to normal tissue, with activated memory CD4 + T cells and M1 macrophages showing the most pronounced enrichment (Figure 5A). Analysis of immune cell crosstalk revealed that cells within the RA synovium form a complex regulatory network, characterized by enhanced interactions between adaptive and innate immune compartments (Figure 5B). Correlation analysis showed a significant positive association between the 5-gene signature and pro-inflammatory immune cell subsets in RA synovium (Figure 5C). Among the signature genes, ITGA4 and BIRC3 exhibited the strongest correlations with activated memory CD4 + T cells and M1 macrophages. GSEA of Pathways Associated with the Five Signature Genes To clarify the biological functions and regulatory mechanisms of the five signature genes, we performed single-gene GSEA based on KEGG pathway annotations. Each gene displayed a distinct enrichment pattern (Figure 6). BIRC3 (baculoviral IAP repeat containing 3) was significantly enriched in the antigen processing and presentation pathway (KEGG: hsa04612), suggesting its involvement in MHC-mediated immune recognition (Figure 6A–B). Three genes— ITGA4 (integrin subunit alpha 4), NUSAP1 (nucleolar and spindle-associated protein 1), and PTTG1 (pituitary tumor-transforming gene 1)—showed convergent enrichment in the T cell receptor signaling pathway (KEGG: hsa04660), indicating their coordinated roles in T cell activation and differentiation (Figure 6C–D). PDE4D (phosphodiesterase 4D) was notably enriched in the autophagy regulation pathway (KEGG: hsa04140), pointing to its function in maintaining cellular homeostasis and modulating inflammatory responses (Figure 6E). Molecular Docking Validation Results from molecular docking simulations verified that resveratrol can bind efficiently to all five target proteins encoded by the diagnostic signature, forming stable hydrogen bond interactions with key amino acid residues within the binding pocket of each target protein. Specifically, resveratrol established interactions with Asp201, His200, and Gln369 of phosphodiesterase 4D ( PDE4D ); Gln328 and Arg332 of baculoviral IAP repeat-containing protein 3 ( BIRC3 ); Arg383 and Asn329 of integrin α4 ( ITGA4 ); Leu2034 and Lys1901 of pituitary tumor-transforming gene 1 ( PTTG1 ); and His407 of nucleolar and spindle-associated protein 1 ( NUSAP1 ) (Figure 7A–E). Binding affinity measurements revealed variation in the strength of these receptor-ligand complexes. The binding energies of resveratrol with PDE4D , PTTG1 , ITGA4 , NUSAP1 , and BIRC3 were -8.0 kcal/mol, -6.8 kcal/mol, -6.3 kcal/mol, -5.6 kcal/mol, and -5.5 kcal/mol, respectively (Figure 7F). All binding free energy values were within the low-energy range, which is conducive to the formation of stable resveratrol-target protein complexes. Discussion We integrated single-cell transcriptomics, bulk RNA sequencing, machine learning, and computational pharmacology technologies into a unified analytical pipeline to address the unsolved issues in current RA biomarker research. The findings of this study mainly revealed five key research outcomes. T-cell profiling in RA synovium uncovered subpopulations with markedly different activation states and effector functions—a level of heterogeneity not fully appreciated in earlier bulk-tissue studies. Cross-referencing T-cell marker genes, RA-linked differentially expressed genes, and predicted resveratrol targets yielded 68 genes shared across all three lists. Running 120 machine learning combinations against our gene pool ultimately narrowed the field to five candidates— PTTG1 , NUSAP1 , PDE4D , ITGA4 , and BIRC3 —whose joint performance proved consistently strong. This signature demonstrated strong diagnostic performance (AUC > 0.90) across multiple validation cohorts. In RA synovium, higher expression of these five genes went hand-in-hand with greater M1 macrophage and CD4⁺ T cell infiltration, hinting at a shared regulatory axis. Docking runs placed resveratrol at favorable binding poses on all five proteins, which, while preliminary, gives a structural rationale for treating these targets pharmacologically. Comparison with Existing RA Biomarkers and Diagnostic Signatures Compared with the RA diagnostic biomarkers reported in previous studies, the 5-gene signature constructed in this study has prominent advantages in both clinical application and biological relevance. At present, clinical serological diagnosis of RA mainly depends on RF and ACPA, but both of these two serological markers have the defect of unstable detection sensitivity, and cannot accurately reflect the dynamic changes of RA disease activity in clinical follow-up [ 40 , 41 ]. Reading gene expression directly from the inflamed joint cuts out the noise introduced when blood is used as a proxy tissue. Blood-based panels—Raterman et al. [ 42 ] and Tao et al. [ 43 ] being notable examples—pick up whole-body immune changes and may miss what is happening specifically inside the joint. Our synovial-based approach avoids this drawback. Each gene in the panel also has a paper trail: prior studies tie them individually to T-cell activation, mitotic regulation, or the inflammatory signaling that erodes cartilage and bone in RA. Each of the five signature genes has documented links to immune function and inflammatory processes. PTTG1 encodes securin; beyond its canonical role in sister-chromatid separation, securin appears to fuel T-cell expansion and is overexpressed in a range of autoimmune diseases [ 44 ]. The protein encoded by NUSAP1 organizes microtubules during cell division; work in lymphocytes suggests it also plays a role when these cells receive activation signals [ 45 ]. ITGA4 encodes integrin α4, which has already proven itself as a therapeutic target—natalizumab, an antibody blocking α4 integrin, works effectively in multiple sclerosis and Crohn’s disease [ 46 ]. cIAP2 , the product of BIRC3 , sits upstream of NF-κB and appears to shield autoreactive T cells from apoptosis—a property with obvious relevance to chronic inflammation [ 47 ]. PDE4D encodes phosphodiesterase 4D, an enzyme controlling c AMP levels that serves as an anti-inflammatory target; apremilast, which inhibits PDE4 , has received approval for treating psoriatic arthritis [ 48 ]. Implications for RA Prognosis and Treatment Response Prediction Diagnostic performance aside, the signature may say something about where a patient’s disease is headed and how it will respond to treatment. The tight link we found between signature expression and infiltrating immune cells raises the possibility that these five genes track disease severity in real time. Patients with the highest ITGA4 and BIRC3 readings also had the densest M1 macrophage infiltrates, a pattern that has historically predicted faster radiographic deterioration [ 49 ]. Whether PDE4D expression marks out patients who stand to gain most from PDE4 inhibition is worth testing prospectively [ 50 ]. Treatment response prediction is a natural next question. Sorting patients by molecular profile before prescribing has gained traction as a strategy for improving RA outcomes [ 51 , 52 ]. The five genes in our signature operate within pathways that conventional and biologic DMARDs target. BIRC3 , for example, affects NF-κB signaling—a pathway that TNF inhibitors shut down [ 53 ]—while ITGA4 is directly targeted by integrin-blocking drugs [ 54 ]. A prospective study measuring these genes at baseline—before any drug is started—would tell us whether the signature has genuine predictive value for treatment selection. Research Limitations and Future Directions Our findings come with caveats worth stating plainly. Working from archived public datasets means batch effects and platform differences could have crept into the expression measurements in ways we cannot fully correct [ 55 ]. All three cohorts skewed toward established, classified RA; the signature has not yet faced the harder test of early or undifferentiated arthritis [ 56 ]. It should be noted that the binding affinity scores obtained from molecular docking simulations are only theoretical computational results, and the actual binding ability between resveratrol and these target proteins needs to be further verified by in vitro cell-free binding experiments and protein-ligand co-crystallography technology before drawing definite conclusions [ 57 ]. The source datasets drew predominantly from East Asian and European populations, so performance in other groups is an open question [ 58 ]. The most pressing next step is enrolling patients with early or undifferentiated arthritis to see whether the signature flags future RA before classification criteria are met [ 56 , 59 ]. Knocking out or silencing each gene individually in T-cell or macrophage models—using CRISPR/Cas9—would pin down which steps in the inflammatory cascade each one controls [ 60 ]. mRNA levels do not always predict protein abundance; pairing this transcriptomic dataset with proteomics and metabolomics would expose post-transcriptional controls that the current analysis misses [ 61 , 62 ]. Resveratrol’s notoriously poor bioavailability has hampered earlier trials; formulation work—nanoparticle encapsulation or phospholipid complexes—should precede any definitive randomized trial in RA patients [ 63 , 64 ]. Conclusions Taken together, our data support a five-gene panel— PTTG1 , NUSAP1 , PDE4D , ITGA4 , and BIRC3 —as a synovial tissue-based diagnostic tool for RA, one that held up across two independent cohorts tested here. The same genes track with the immune cell infiltrates that drive synovitis, suggesting they do more than discriminate cases from controls; they may index disease activity. Docking analysis adds a structural argument that resveratrol could modulate all five targets, though this remains a hypothesis until tested in the laboratory. Moving from computational prediction to bedside application will require prospective trials and mechanistic experiments, but the convergent evidence assembled here gives those efforts a defined molecular starting point. Declarations Acknowledgments : Not applicable. Funding: This study was supported by Ningxia Medical University School-level Project (Grant No.: XY2025046). Author Contributions : Conception and design: Nan Yu; Methodology: Yan Zhou, Jia Guo; Formal analysis: Yan Zhou, Xiaoqing Bao; Investigation: Yan Zhou, Jia Guo; Data curation: Xiaoqing Bao; Writing – original draft: Yan Zhou; Writing – review and editing: Nan Yu; Supervision: Nan Yu; Project administration: Nan Yu; Funding acquisition: Nan Yu. Compliance with Ethical Standards Declarations : In accordance with the Declaration of Helsinki, this research is based on open-source data, so ethics approval was waived. Competing Interests : The authors have no conflicts of interest to declare. References Smolen JS, Aletaha D, Barton A, Burmester GR, Emery P, Firestein GS, Kavanaugh A, McInnes IB, Solomon DH, Strand V, et al (2018) Rheumatoid arthritis. Nat Rev Dis Primers 4:18001. https://doi.org/10.1038/nrdp.2018.1 Firestein GS, McInnes IB (2017) Immunopathogenesis of rheumatoid arthritis. Immunity 46:183–196. https://doi.org/10.1016/j.immuni.2017.02.006 McInnes IB, Schett G (2011) The pathogenesis of rheumatoid arthritis. N Engl J Med 365:2205–2219. https://doi.org/10.1056/NEJMra1004965 Smolen JS, Aletaha D, McInnes IB (2016) Rheumatoid arthritis. Lancet 388:2023–2038. https://doi.org/10.1016/S0140-6736(16)30173-8 Smolen JS, Landewé RBM, Bijlsma JWJ, Burmester GR, Dougados M, Kerschbaumer A, McInnes IB, Sepriano A, van Vollenhoven RF, de Wit M, et al (2020) EULAR recommendations for the management of rheumatoid arthritis with synthetic and biological disease-modifying antirheumatic drugs: 2019 update. Ann Rheum Dis 79:685–699. https://doi.org/10.1136/annrheumdis-2019-216655 Burmester GR, Pope JE (2017) Novel treatment strategies in rheumatoid arthritis. Lancet 389:2338–2348. https://doi.org/10.1016/S0140-6736(17)31491-5 Weyand CM, Goronzy JJ (2021) The immunology of rheumatoid arthritis. Nat Immunol 22:10–18. https://doi.org/10.1038/s41590-020-00816-x Cope AP, Schulze-Koops H, Aringer M (2007) The central role of T cells in rheumatoid arthritis. Clin Exp Rheumatol 25(Suppl 46):S4–S11 Chemin K, Gerstner C, Malmström V (2019) Effector functions of CD4 + T cells at the site of local autoimmune inflammation—Lessons from rheumatoid arthritis. Front Immunol 10:353. https://doi.org/10.3389/fimmu.2019.00353 Miossec P, Korn T, Kuchroo VK (2009) Interleukin-17 and type 17 helper T cells. N Engl J Med 361:888–898. https://doi.org/10.1056/NEJMra0707449 Lubberts E (2015) The IL-23–IL-17 axis in inflammatory arthritis. Nat Rev Rheumatol 11:415–429. https://doi.org/10.1038/nrrheum.2015.53 Nie H, Zheng Y, Li R, Guo TB, He D, Fang L, Liu X, Xiao L, Chen X, Wan B, et al (2013) Phosphorylation of FOXP3 controls regulatory T cell function and is inhibited by TNF-α in rheumatoid arthritis. Nat Med 19:322–328. https://doi.org/10.1038/nm.3085 Komatsu N, Okamoto K, Sawa S, Nakashima T, Oh-hora M, Kodama T, Tanaka S, Bluestone JA, Takayanagi H (2014) Pathogenic conversion of Foxp3+ T cells into TH17 cells in autoimmune arthritis. Nat Med 20:62–68. https://doi.org/10.1038/nm.3432 Aletaha D, Neogi T, Silman AJ, Funovits J, Felson DT, Bingham CO, Birnbaum NS, Burmester GR, Bykerk VP, Cohen MD, et al (2010) 2010 Rheumatoid arthritis classification criteria: An American College of Rheumatology/European League Against Rheumatism collaborative initiative. Arthritis Rheum 62:2569–2581. https://doi.org/10.1002/art.27584 Nishimura K, Sugiyama D, Kogata Y, Tsuji G, Nakazawa T, Kawano S, Saigo K, Morinobu A, Koshiba M, Kuntz KM, et al (2007) Meta-analysis: Diagnostic accuracy of anti-cyclic citrullinated peptide antibody and rheumatoid factor for rheumatoid arthritis. Ann Intern Med 146:797–808. https://doi.org/10.7326/0003-4819-146-11-200706050-00008 Steiner G, Tesar V (2009) Diagnostic value of anti-citrullinated protein antibodies in rheumatoid arthritis. Ann N Y Acad Sci 1173:439–447. https://doi.org/10.1111/j.1749-6632.2009.04863.x van der Helm-van Mil AHM, Huizinga TWJ (2008) Advances in the genetics of rheumatoid arthritis point to subclassification into distinct disease subsets. Arthritis Res Ther 10:205. https://doi.org/10.1186/ar2384 Smolen JS, Aletaha D (2004) Patients with rheumatoid arthritis in clinical care. Ann Rheum Dis 63:221–225. https://doi.org/10.1136/ard.2003.012252 Strimbu K, Tavel JA (2010) What are biomarkers? Curr Opin HIV AIDS 5:463–466. https://doi.org/10.1097/COH.0b013e32833ed177 Brazma A, Hingamp P, Quackenbush J, Sherlock G, Spellman P, Stoeckert C, Aach J, Ansorge W, Ball CA, Causton HC, et al (2001) Minimum information about a microarray experiment (MIAME)—Toward standards for microarray data. Nat Genet 29:365–371. https://doi.org/10.1038/ng1201-365 Wang Z, Gerstein M, Snyder M (2009) RNA-Seq: A revolutionary tool for transcriptomics. Nat Rev Genet 10:57–63. https://doi.org/10.1038/nrg2484 van der Pouw Kraan TC, Wijbrandts CA, Drijfhout JW, Baggen JM, Verweij CL, Tak PP (2008) Responsiveness to anti-tumour necrosis factor alpha therapy is related to pre-treatment tissue inflammation levels in rheumatoid arthritis patients. Ann Rheum Dis 67:563–566. https://doi.org/10.1136/ard.2007.081950 Dennis G Jr, Holweg CT, Kummerfeld SK, Choy DF, Setiadi AF, Hackney JA, Haverty PM, Gilbert H, Lin WY, Diehl L, et al (2014) Synovial phenotypes in rheumatoid arthritis correlate with response to biologic therapeutics. Arthritis Res Ther 16: R90. https://doi.org/10.1186/ar4555 Baur JA, Sinclair DA (2006) Therapeutic potential of resveratrol: The in vivo evidence. Nat Rev Drug Discov 5:493–506. https://doi.org/10.1038/nrd2060 Malaguarnera L (2019) Influence of resveratrol on the immune response. Nutrients 11:946. https://doi.org/10.3390/nu11050946 Shakibaei M, Buhrmann C, Mobasheri A (2011) Resveratrol-mediated SIRT-1 interactions with p300 modulate receptor activator of NF-κB ligand (RANKL) activation of NF-κB signaling and inhibit osteoclastogenesis in bone-derived cells. J Biol Chem 286:11492–11505. https://doi.org/10.1074/jbc.M110.198713 Xuzhu G, Komai-Koma M, Leung BP, Howe HS, McSharry C, McInnes IB, Xu D (2012) Resveratrol modulates murine collagen-induced arthritis by inhibiting Th17 and B-cell function. Ann Rheum Dis 71:129–135. https://doi.org/10.1136/ard.2011.149831 Hopkins AL (2008) Network pharmacology: The next paradigm in drug discovery. Nat Chem Biol 4:682–690. https://doi.org/10.1038/nchembio.118 Greener JG, Kandathil SM, Moffat L, Jones DT (2022) A guide to machine learning for biologists. Nat Rev Mol Cell Biol 23:40–55. https://doi.org/10.1038/s41580-021-00407-0 Chicco D (2017) Ten quick tips for machine learning in computational biology. Bio Data Min 10:35. https://doi.org/10.1186/s13040-017-0155-3 Walsh I, Fishman D, Garcia-Gasulla D, Titber T, Harber G, Lordge J, Tosatto SCE, Capriotti E (2021) DOME: Recommendations for supervised machine learning validation in biology. Nat Methods 18:1122–1127. https://doi.org/10.1038/s41592-021-01205-4 Liu Z, Liu L, Weng S, Guo C, Dang Q, Xu H, Wang L, Lu T, Zhang Y, Sun Z, et al (2022) Machine learning-based integration develops an immune-derived lncRNA signature for improving outcomes in colorectal cancer. Nat Commun 13:816. https://doi.org/10.1038/s41467-022-28421-6 Friedman J, Hastie T, Tibshirani R (2010) Regularization paths for generalized linear models via coordinate descent. J Stat Softw 33:1–22. https://doi.org/10.18637/jss.v033.i01 Chen T, Guestrin C (2016) XGBoost: A scalable tree boosting system. In: Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, San Francisco, CA, USA, 13–17 August 2016, pp 785–794. https://doi.org/10.1145/2939672.2939785 Stuart T, Butler A, Hoffman P, Hafemeister C, Papalexi E, Mauck WM, Hao Y, Stoeckius M, Smibert P, Satija R (2019) Comprehensive integration of single-cell data. Cell 177:1888–1902.e21. https://doi.org/10.1016/j.cell.2019.05.031 Newman AM, Liu CL, Green MR, Gentles AJ, Feng W, Xu Y, Hoang CD, Diehn M, Alizadeh AA (2015) Robust enumeration of cell subsets from tissue expression profiles. Nat Methods 12:453–457. https://doi.org/10.1038/nmeth.3337 Wu T, Hu E, Xu S, Chen M, Guo P, Dai Z, Feng T, Zhou L, Tang W, Zhan L, et al (2021) clusterProfiler 4.0: A universal enrichment tool for interpreting omics data. Innovation 2:100141. https://doi.org/10.1016/j.xinn.2021.100141 Ritchie ME, Phipson B, Wu D, Hu Y, Law CW, Shi W, Smyth GK (2015) limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res 43: e47. https://doi.org/10.1093/nar/gkv007 Morris GM, Huey R, Lindstrom W, Sanner MF, Belew RK, Goodsell DS, Olson AJ (2009) AutoDock4 and AutoDockTools4: Automated docking with selective receptor flexibility. J Comput Chem 30:2785–2791. https://doi.org/10.1002/jcc.21256 Trott O, Olson AJ (2010) AutoDock Vina: Improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading. J Comput Chem 31:455–461. https://doi.org/10.1002/jcc.21334 Vlotides G, Eigler T, Melmed S (2007) Pituitary tumor-transforming gene: Physiology and implications for tumorigenesis. Endocr Rev 28:165–186. https://doi.org/10.1210/er.2006-0042 Raemaekers T, Ribbeck K, Beaudouin J, Annaert W, Van Camp M, Stockmans I, Smets N, Bouillon R, Ellenberg J, Carmeliet G (2003) NuSAP, a novel microtubule-associated protein involved in mitotic spindle organization. J Cell Biol 162:1017–1029. https://doi.org/10.1083/jcb.200302129 Houslay MD, Schafer P, Zhang KY (2005) Keynote review: Phosphodiesterase-4 as a therapeutic target. Drug Discov Today 10:1503–1519. https://doi.org/10.1016/S1359-6446(05)03622-6 Jin SL, Ding SL, Lin SC (2012) Phosphodiesterase 4 and its inhibitors in inflammatory diseases. Chang Gung Med J 35:197–210 Yednock TA, Cannon C, Fritz LC, Sanchez-Madrid F, Steinman L, Karin N (1992) Prevention of experimental autoimmune encephalomyelitis by antibodies against alpha4beta1 integrin. Nature 356:63–66. https://doi.org/10.1038/356063a0 Varfolomeev E, Vucic D (2011) Inhibitor of apoptosis proteins: Fascinating biology leads to attractive tumor therapeutic targets. Future Oncol 7:633–648. https://doi.org/10.2217/fon.11.40 Raterman HG, Vosslamber S, de Ridder S, Nurmohamed MT, Lems WF, Boers M, van de Wiel M, Dijkmans BA, Verweij CL, Voskuyl AE (2012) The interferon type I signature towards prediction of non-response to rituximab in rheumatoid arthritis patients. Arthritis Res Ther 14: R95. https://doi.org/10.1186/ar3819 Lewis MJ, Barnes MR, Blighe K, Goldmann K, Rana S, Hackney JA, Ramamoorthi N, John CR, Watson DS, Kummerfeld SK, et al (2019) Molecular portraits of early rheumatoid arthritis identify clinical and treatment response phenotypes. Cell Rep 28:2455–2470.e5. https://doi.org/10.1016/j.celrep.2019.07.091 Schafer PH, Parton A, Capone L, Cedzik D, Brady H, Evans JF, Man HW, Muller GW, Stirling DI, Chopra R (2014) Apremilast is a selective PDE4 inhibitor with regulatory effects on innate immunity. Cell Signal 26:2016–2029. https://doi.org/10.1016/j.cellsig.2014.05.014 Riveiro-Naveira RR, Valcárcel-Ares MN, Almonte-Becerril M, Vaamonde-García C, Loureiro J, Hermida-Carballo L, López-Peláez E, Blanco FJ, López-Armada MJ (2016) Resveratrol lowers synovial hyperplasia, inflammatory markers and oxidative damage in an acute antigen-induced arthritis model. Rheumatology 55:1889–1900. https://doi.org/10.1093/rheumatology/kew255 Walle T, Hsieh F, De Legge MH, Oatis JE Jr, Walle UK (2004) High absorption but very low bioavailability of oral resveratrol in humans. Drug Metab Dispos 32:1377–1382. https://doi.org/10.1124/dmd.104.000885 Hayden MS, Ghosh S (2008) Shared principles in NF-κB signaling. Cell 132:344–362. https://doi.org/10.1016/j.cell.2008.01.020 Pitzalis C, Kelly S, Humby F (2013) New learnings on the pathophysiology of RA from synovial biopsies. Curr Opin Rheumatol 25:334–344. https://doi.org/10.1097/BOR.0b013e32835fd8eb McCann FE, Palfreeman AC, Andrews M, Sherlock JP, Williams RO, Brennan FM (2010) Apremilast, a novel PDE4 inhibitor, inhibits spontaneous production of tumour necrosis factor-alpha from human rheumatoid synovial cells and ameliorates experimental arthritis. Arthritis Res Ther 12: R107. https://doi.org/10.1186/ar3041 Leek JT, Scharpf RB, Bravo HC, Simcha D, Langmead B, Johnson WE, Geman D, Baggerly K, Irizarry RA (2010) Tackling the widespread and critical impact of batch effects in high-throughput data. Nat Rev Genet 11:733–739. https://doi.org/10.1038/nrg2825 van der Helm-van Mil AHM, le Cessie S, van Dongen H, Brinkman J, de Craen AJM, Huizinga TWJ (2007) A prediction rule for disease outcome in patients with recent-onset undifferentiated arthritis: How to guide individual treatment decisions. Arthritis Rheum 56:433–440. https://doi.org/10.1002/art.22380 Copeland RA, Pompliano DL, Meek TD (2006) Drug-target residence time and its implications for lead optimization. Nat Rev Drug Discov 5:730–739. https://doi.org/10.1038/nrd2082 Popejoy AB, Fullerton SM (2016) Genomics is failing on diversity. Nature 538:161–164. https://doi.org/10.1038/538161a Gerlag DM, Raza K, van Baarsen LGM, Brouwer E, Tak PP (2012) EULAR recommendations for terminology and research in individuals at risk of rheumatoid arthritis: Report from the Study Group for Risk Factors for Rheumatoid Arthritis. Ann Rheum Dis 71:638–641. https://doi.org/10.1136/annrheumdis-2011-200990 Cong L, Ran FA, Cox D, Lin S, Barretto R, Habib N, Hsu PD, Wu X, Jiang W, Marraffini LA, et al (2013) Multiplex genome engineering using CRISPR/Cas systems. Science 339:819–823. https://doi.org/10.1126/science.1231143 Hasin Y, Seldin M, Lusis A (2017) Multi-omics approaches to disease. Genome Biol 18:83. https://doi.org/10.1186/s13059-017-1215-1 Karczewski KJ, Snyder MP (2018) Integrative omics for health and disease. Nat Rev Genet 19:299–310. https://doi.org/10.1038/nrg.2018.4 Pangeni R, Sahni JK, Ali J, Sharma S, Baboota S (2014) Resveratrol: Review on therapeutic potential and recent advances in drug delivery. Expert Opin Drug Deliv 11:1285–1298. https://doi.org/10.1517/17425247.2014.919253 Khojah HM, Ahmed S, Abdel-Rahman MS, Elhakeim EH (2018) Resveratrol as an effective adjuvant therapy in the management of rheumatoid arthritis: A clinical study. Clin Rheumatol 37:2035–2042. https://doi.org/10.1007/s10067-018-4080-8 Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9098862","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":609126093,"identity":"94337083-d918-4cf8-84e2-8e0fad30861d","order_by":0,"name":"Yan Zhou¹","email":"","orcid":"","institution":"Ningxia Medical University General Hospital","correspondingAuthor":false,"prefix":"","firstName":"Yan","middleName":"","lastName":"Zhou¹","suffix":""},{"id":609126096,"identity":"e1bc6644-b113-4533-b608-ae31a356fd91","order_by":1,"name":"Jia Guo²","email":"","orcid":"","institution":"People's Hospital of Ningxia Hui Autonomous Region","correspondingAuthor":false,"prefix":"","firstName":"Jia","middleName":"","lastName":"Guo²","suffix":""},{"id":609126099,"identity":"122bc8aa-bfa6-472f-8cf9-e339461ceec1","order_by":2,"name":"Xiaoqing Bao¹","email":"","orcid":"","institution":"Ningxia Medical University General Hospital","correspondingAuthor":false,"prefix":"","firstName":"Xiaoqing","middleName":"","lastName":"Bao¹","suffix":""},{"id":609126102,"identity":"11091012-be6e-42a0-8d2c-6e0d0716dbda","order_by":3,"name":"Nan Yu¹","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA5UlEQVRIie3PLwvCQBjH8RsHZ3lkwXKizGY+GVj3VnYIJgXjBcNkchf8V30ZNrU5hLOc3TjZG9Bm03Vl02a4b/59eHgQstn+sCB2JtlDPCFQcZKGYlxOHIVjHxnsMdA9lhr9BVlVZM2R2Gd00K5fp7icYJzvR4bwLRoQwSOCXDULCwnJSbYWwPfRWV/4romoOW8KCWBHMTCUR8myf+GGIEaHxYTmV2hVMh4doTviEpcTlpNGVYY+09BF35K4szYHrz4nPRoaDaW/BKvTNb2JA7itLLk/xNhz1aKYvAW/zW02m832sReUAUwKRK/nYgAAAABJRU5ErkJggg==","orcid":"","institution":"Ningxia Medical University General Hospital","correspondingAuthor":true,"prefix":"","firstName":"Nan","middleName":"","lastName":"Yu¹","suffix":""}],"badges":[],"createdAt":"2026-03-12 01:23:46","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9098862/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9098862/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":105084185,"identity":"d116465e-65b0-4569-ae75-3267f627f5d1","added_by":"auto","created_at":"2026-03-20 19:04:53","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":1263264,"visible":true,"origin":"","legend":"\u003cp\u003eCellular Heterogeneity and T–Cell Functional Characteristics in RA Synovial Tissue\u003c/p\u003e\n\u003cp\u003e(A) Uniform Manifold Approximation and Projection (UMAP) visualization of scRNA-seq data from RA synovial tissue, identifying seven major cell lineages: T cells, B cells, macrophages, fibroblasts, neutrophils, NK cells, and DCs. (B) Dot plot showing the expression level (color) and proportion (dot size) of canonical marker genes across different cell subsets. (C) Stacked bar chart illustrating the proportional distribution of cell types in synovial tissue samples from individual RA patients. (D) Heatmap displaying the Z-score of top marker gene expression in each cell cluster, with red indicating high expression and blue indicating low expression. (E) Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis showing the six core significantly enriched pathways in T cells, with the x-axis representing -log10(p.adjust). (F) Protein – protein interaction (PPI) network constructed based on T-cell-enriched pathways. Nodes of different colors represent distinct KEGG pathway modules, and hub genes are highlighted at the center of the network.\u003c/p\u003e","description":"","filename":"image1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9098862/v1/740e6548cb6f5fb582e832f1.jpg"},{"id":105727976,"identity":"bbe55fe7-72f3-4e65-a37f-aa6f8f981b38","added_by":"auto","created_at":"2026-03-30 11:07:07","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":1536131,"visible":true,"origin":"","legend":"\u003cp\u003eBulk transcriptomic profiling of RA synovial tissue and resveratrol target acquisition\u003c/p\u003e\n\u003cp\u003e(A) Hierarchical clustering heatmap depicting genome-wide expression patterns across RA and normal control synovial samples from GSE89408. (B) Volcano plot illustrating differential gene expression between RA and control groups. (C) Bar chart summarizing resveratrol-associated molecular targets retrieved from four pharmacological prediction platforms: CTD, Gene Cards, Super-PRED, and Swiss Target Prediction. (D) The intersection among three distinct gene sets in the Venn diagram.\u003c/p\u003e","description":"","filename":"image2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9098862/v1/9c551fdab012f3e36ff4038e.jpg"},{"id":105562971,"identity":"9073b307-60bd-4c3c-80f2-74d1aa9d24a7","added_by":"auto","created_at":"2026-03-27 12:45:28","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":1734375,"visible":true,"origin":"","legend":"\u003cp\u003eFunctional Enrichment and PPI Network Analysis of Resveratrol Targeting Rheumatoid Arthritis-Related T Cells\u003c/p\u003e\n\u003cp\u003e(A and B) Bar chart of GO functional enrichment analysis of 68 candidate genes (the y-axis represents GO functional terms, and the x-axis represents enrichment significance). (C and D) Bubble chart of KEGG pathway enrichment analysis of 68 candidate genes (the size of bubbles represents gene count, and the color represents enrichment significance). (E) PPI network analysis diagram of 68 candidate genes (the size of nodes indicates connectivity, the color of lines indicates the type of interaction).\u003c/p\u003e","description":"","filename":"image3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9098862/v1/ed5662aef96f127b63d1cff0.jpg"},{"id":105084182,"identity":"8348b668-c4db-492a-a484-6219aa1851c4","added_by":"auto","created_at":"2026-03-20 19:04:53","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":1739181,"visible":true,"origin":"","legend":"\u003cp\u003eConstruction and validation of the diagnostic model\u003c/p\u003e\n\u003cp\u003e(A) Heatmap of AUC values for 120 algorithm combinations; LASSO+PLS was optimal. (B) ROC curve in the training set (AUC = 0.983). (C–D) ROC curves in external validation sets GSE55457 (AUC = 0.908) and GSE77298 (AUC = 0.902). (E) DCA. (F) Calibration curve.\u003c/p\u003e","description":"","filename":"image4.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9098862/v1/eeda9bfc3e2dbcdb92c285c6.jpg"},{"id":105084179,"identity":"53d3afd5-6739-4de4-b0ae-dba808651c3a","added_by":"auto","created_at":"2026-03-20 19:04:53","extension":"jpg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":1130708,"visible":true,"origin":"","legend":"\u003cp\u003eImmune infiltration landscape and correlation analysis in RA synovial tissue\u003c/p\u003e\n\u003cp\u003e(A) Box plots comparing the relative abundance of 22 immune cell types between in RA and Normal groups based on CIBERSORT algorithm analysis. Statistical significance was determined using the Wilcoxon rank-sum test (*\u003cem\u003ep\u003c/em\u003e \u0026lt; 0.05, **\u003cem\u003ep\u003c/em\u003e \u0026lt; 0.01, ***\u003cem\u003ep\u003c/em\u003e \u0026lt; 0.001; ns, not significant). (B) Correlation heatmap displaying Spearman’s correlation coefficients among the 22 immune cell types. (C) Bubble plot illustrating correlations between the five diagnostic signature genes (\u003cem\u003ePTTG1\u003c/em\u003e, \u003cem\u003eUSAP1\u003c/em\u003e, \u003cem\u003eDE4D\u003c/em\u003e, \u003cem\u003eTGA4\u003c/em\u003e, and \u003cem\u003eIRC3\u003c/em\u003e) and immune cell infiltration levels. Statistically significant correlations (\u003cem\u003ep\u003c/em\u003e \u0026lt; 0.05) are displayed.\u003c/p\u003e","description":"","filename":"image5.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9098862/v1/29d9ad4b4d71d6d646d35fab.jpg"},{"id":105084180,"identity":"f4971b29-0d3a-42d5-8b61-8fe960ec1995","added_by":"auto","created_at":"2026-03-20 19:04:53","extension":"jpg","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":1216739,"visible":true,"origin":"","legend":"\u003cp\u003eGSEA of individual core genes from the five-gene diagnostic signature in RA\u003c/p\u003e\n\u003cp\u003e(A) GSEA enrichment plot for \u003cem\u003eBIRC3\u003c/em\u003e in the antigen processing and presentation pathway. (B) GSEA enrichment plot for \u003cem\u003eITGA4\u003c/em\u003e in the T cell receptor signaling pathway. (C) GSEA enrichment plot for \u003cem\u003eNUSAP1\u003c/em\u003e in the T cell receptor signaling pathway. (D) GSEA enrichment plot for \u003cem\u003ePDE4D\u003c/em\u003e in the regulation of autophagy pathway. (E) GSEA enrichment plot for \u003cem\u003ePTTG1\u003c/em\u003e in the T cell receptor signaling pathway.\u003c/p\u003e","description":"","filename":"image6.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9098862/v1/4afc195e12026685d7ffd17b.jpg"},{"id":105903917,"identity":"af9cfa25-59a6-4ed6-befb-5092ffc10d6c","added_by":"auto","created_at":"2026-04-01 09:58:09","extension":"jpg","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":750419,"visible":true,"origin":"","legend":"\u003cp\u003eMolecular docking of Resveratrol with signature targets\u003c/p\u003e\n\u003cp\u003e3D visualization of Resveratrol binding to (A) \u003cem\u003ePDE4D\u003c/em\u003e, (B) \u003cem\u003eBIRC3\u003c/em\u003e, (C)\u003cem\u003eITGA4\u003c/em\u003e, (D) \u003cem\u003ePTTG1\u003c/em\u003e, (E) \u003cem\u003eNUSAP1\u003c/em\u003e. (F) Heatmap of binding energies (kcal/mol).\u003c/p\u003e","description":"","filename":"image7.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9098862/v1/215314248a7bd1a94218bd82.jpg"},{"id":105906320,"identity":"d82e0b7f-27e7-4935-85c8-9e284c5a46fc","added_by":"auto","created_at":"2026-04-01 10:19:22","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":10348460,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9098862/v1/774b29f0-4686-4cc7-bda1-c0c0996687e9.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"A Five-Gene Signature Identified by Integrating Single-Cell Analysis and Machine Learning Benchmarking Enables Rheumatoid Arthritis Diagnosis and Resveratrol Target Discovery","fulltext":[{"header":"Highlight Box","content":"\u003cp\u003e\u003cstrong\u003e\u003cem\u003eKey findings\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul start=\"50\"\u003e\n \u003cli\u003eFive immune-related genes (\u003cem\u003eBIRC3\u003c/em\u003e, \u003cem\u003eITGA4\u003c/em\u003e, \u003cem\u003eNUSAP1\u003c/em\u003e, \u003cem\u003ePTTG1\u003c/em\u003e, \u003cem\u003ePDE4D\u003c/em\u003e) were identified as a robust diagnostic signature for rheumatoid arthritis (RA), and a multi-algorithm machine learning model demonstrated high and stable diagnostic performance across training and external validation cohorts.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eWhat is known and what is new?\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul start=\"50\"\u003e\n \u003cli\u003eT cells and inflammatory signaling are central to the pathogenesis of rheumatoid arthritis.\u003c/li\u003e\n \u003cli\u003eThis study integrates single-cell analysis, machine learning benchmarks, and computational pharmacology to identify a five-gene signature for RA diagnosis and uncover potential resveratrol targets.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eWhat is the implication, and what should change now?\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul start=\"50\"\u003e\n \u003cli\u003eThe five-gene signature could serve as a reliable tool for early and accurate diagnosis of rheumatoid arthritis.\u003c/li\u003e\n \u003cli\u003eThe identified signature and resveratrol targets offer new insights for mechanistic research and personalized treatment of RA.\u003c/li\u003e\n\u003c/ul\u003e"},{"header":"Introduction","content":"\u003cp\u003eRheumatoid arthritis has a global prevalence of approximately 0.5% to 1% in the adult population, and its typical pathological feature is persistent synovial inflammation that can induce cartilage erosion and irreversible articular damage if not effectively intervened in the early stage [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. The pathogenesis of RA is a complex process jointly mediated by genetic susceptibility, environmental inducing factors and immune system disorder, with all these pathological factors converging and exerting effects in the synovial tissue microenvironment [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. Although conventional disease-modifying antirheumatic drugs (DMARDs) and biological agents have greatly enriched the clinical therapeutic strategies for RA, a considerable proportion of patients still cannot obtain satisfactory disease control effects in clinical practice [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. T lymphocytes act as the core regulatory component of the inflammatory network in RA, exerting their pathological effects through antigen recognition, pro-inflammatory cytokine secretion, and functional crosstalk with other immune cells in the synovial microenvironment [\u003cspan additionalcitationids=\"CR8\" citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. The Th1 and Th17 subsets of CD4⁺ helper T cells promote the development and persistence of synovial inflammation by secreting pro-inflammatory cytokines such as IFN-γ, IL\u0026ndash;17 and TNF-α [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. In contrast, regulatory T cells (Tregs) in RA patients exhibit impaired immune suppressive function, which fails to effectively inhibit excessive inflammatory responses and leads to persistent synovial inflammation [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eAlthough our grasp of RA pathogenesis has deepened considerably, molecular biomarkers suitable for routine diagnosis remain elusive [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. Clinicians currently diagnose RA using ACR/EULAR criteria that blend symptoms, imaging, and serology\u0026mdash;chiefly rheumatoid factor (RF) and anti-citrullinated protein antibodies (ACPA) [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]. Neither marker is ideal: RF specificity hovers around 60\u0026ndash;80% because titers rise in other autoimmune states and even in healthy older adults; ACPA, although more specific, is undetectable in ~\u0026thinsp;30% of patients (seronegative RA) [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. Worse still, serological titers seldom mirror real-time disease activity or forecast therapeutic outcomes [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eHigh-throughput transcriptomics now permits discovery of gene-expression signatures that may aid diagnosis and prognosis [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. Several research teams have proposed interferon response modules and inflammation-related pathway gene panels as potential diagnostic biomarkers for RA [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]. However, these gene signatures have obvious limitations in clinical application: they are mostly derived from peripheral blood samples rather than the pathogenic site of RA (synovial tissue), their detection results cannot be stably replicated in different patient cohorts, and their intrinsic biological correlation with the pathophysiological process of RA remains unclear [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]. A robust, synovium-derived signature with clear biological underpinnings is therefore still needed.\u003c/p\u003e \u003cp\u003eThe stilbene resveratrol (trans-3,5,4\u0026prime;-trihydroxystilbene), found in grapes, berries, and peanuts, has attracted interest owing to reported anti-inflammatory and immunomodulatory effects [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]. In preclinical RA models, resveratrol has been shown to alleviate synovial inflammation through multiple mechanisms, including inhibiting the activation of NF-κB signaling pathway, reducing the secretion of pro-inflammatory cytokines, and promoting the differentiation and functional enhancement of Tregs [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e, \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e]. Nevertheless, the specific molecular targets of resveratrol that mediate these anti-inflammatory and immunomodulatory effects in the RA synovial microenvironment have not been clearly identified [\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e, \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e]. Network pharmacology wedded to transcriptomics provides a rational route to map resveratrol targets onto RA-dysregulated genes [\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eMachine learning now underpins many efforts to distill high-dimensional omics data into clinically useful classifiers [\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e, \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e]. Early studies on RA gene biomarkers mostly constructed diagnostic models based on a single machine learning algorithm, which easily leads to overfitting of the discovery cohort data and thus the model cannot be effectively validated in external independent cohorts [\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e, \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e]. In contrast, the simultaneous benchmarking of multiple machine learning algorithm combinations can effectively avoid overfitting and construct a diagnostic model with better generalization ability in unknown patient cohorts [\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e, \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eTo address the above-mentioned deficiencies in current RA biomarker research and resveratrol target exploration, we designed an integrated multi-omics analysis pipeline for this study. Our specific research objectives are as follows: (i) to analyze the heterogeneity of T cell subsets in RA synovial tissue by scRNA-seq technology; (ii) to perform intersection analysis of T cell-specific marker genes, RA-related dysregulated genes and known resveratrol molecular targets; (iii) to screen a stable RA diagnostic gene signature by benchmarking multiple machine learning algorithms; (iv) to explore the clinical relevance of this gene signature through immune cell infiltration profiling and molecular docking simulation.\u003c/p\u003e"},{"header":"Materials and Methods","content":"\u003cp\u003e\u003cstrong\u003eData Acquisition and Single-Cell RNA Sequencing Analysis\u003cbr\u003e\u003c/strong\u003eThe scRNA-seq dataset (GSE296117) derived from RA synovial biopsy samples was retrieved from the Gene Expression Omnibus (GEO) database, and we implemented all data preprocessing procedures with the Seurat R software package. We applied strict quality control to exclude low-quality cells, with the filtering criteria defined as 200 \u0026lt; n Feature RNA \u0026lt; 4000 and mitochondrial gene expression proportion \u0026lt; 10%. Following data normalization, unsupervised clustering was performed to distinguish distinct cell populations, and each population was annotated based on canonical marker genes compiled in the Cell Marker2 database.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDifferential Expression Analysis and Drug Target Prediction\u003cbr\u003e\u003c/strong\u003eWe designated the bulk transcriptomic dataset GSE89408 as the training cohort for differential expression analysis, and the limma software package was utilized to compare gene expression differences between RA patients and healthy controls, with the threshold for differentially expressed genes set as adjusted \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05 and |log\u003csub\u003e2\u003c/sub\u003e fold change| \u0026gt; 1. In parallel, we compiled potential molecular targets of resveratrol from the Comparative Toxicogenomics Database (CTD) and Gene Cards, and leveraged the chemical structure of resveratrol to predict supplementary targets via the Super-PRED and Swiss Target Prediction online platforms.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eIdentification of\u003c/strong\u003e \u003cstrong\u003eKey Candidate Genes\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe Venn Diagram software package was adopted to compute the intersection of three gene sets, namely (1) T cell-specific marker genes identified from scRNA-seq data; (2) RA-related differentially expressed genes (DEGs) obtained from bulk transcriptomic analysis; and (3) predicted resveratrol target genes. The overlapping genes were designated as key therapeutic targets and subjected to Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis using \u003cem\u003ecluster Profiler\u003c/em\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMachine Learning Benchmarking and Model Construction\u003cbr\u003e\u003c/strong\u003eFor the purpose of constructing a robust and reliable RA diagnostic gene signature, we adopted an integrated machine learning analytical framework with the GSE89408 dataset as the training cohort. A total of 19 machine learning algorithms covering regularization-based, tree-based, linear, non-linear and ensemble categories were combined into 120 distinct classifier models with optimized hyperparameter combinations using the caret and glmnet packages, including lasso, Stepglm, glm Boost, Enet, rf, svm Radial, svm Linear, kknn, gbm, pls Rglm, lda, knn, nnet, naive_bayes, pls, step QDA, glm Step AIC, step LDA, and Logit Boost. To minimize the interference of random sampling bias on model performance evaluation, each constructed classifier model was subjected to strict 10-fold cross-validation with three repeated validation cycles. The model with the highest average value of the area under the receiver operating characteristic (ROC) curve in all cross-validation experiments was defined as the optimal diagnostic model for subsequent research. We conducted further validation on the diagnostic efficacy and stability of the identified optimal model using two independent external datasets, GSE55457 and GSE77298.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eImmune Infiltration and Single-Gene Gene Set Enrichment Analysis (GSEA)\u0026nbsp;\u003cbr\u003e\u003c/strong\u003eWe quantified the infiltration levels of 22 immune cell subsets in the GSE89408 dataset via the CIBERSORT algorithm, with parameters set as permutation number = 1000, confidence interval = 95%, and deconvolution results with \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05 deemed statistically significant and reliable for subsequent analyses. Pearson coefficients linked z-scored signature expression to cell-type fractions (two-tailed \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05). Single-gene gene set enrichment analysis (GSEA) was performed with the cluster Profiler package (KEGG database v7.5.1; 1000 permutations) to explore the biological pathways associated with each gene in the diagnostic signature.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMolecular Docking Simulation\u003cbr\u003e\u003c/strong\u003eThe three-dimensional structures of target proteins were obtained from the RCSB PDB database, while the molecular coordinate data of resveratrol were retrieved from the ZINC database. We used Open Babel and Auto Dock Tools software to preprocess ligand (resveratrol) and receptor (target protein) files, including water molecule removal and hydrogen atom addition. Binding affinities were calculated with Auto Dock Vina 1.5.6, and the docked poses were visualized using Py MOL software.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003eFunctional Enrichment, Interaction Network, and Disease Relevance of T Cells in Rheumatoid Arthritis Synovial Tissue\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eT lymphocytes occupy a central position in adaptive immune responses and represent key drivers of the autoimmune pathological processes underlying rheumatoid arthritis (RA). Previous research has demonstrated that hyperactive T cells promote disease progression via multiple mechanisms, including pro-inflammatory cytokine secretion, functional crosstalk with B lymphocytes, and direct cytotoxicity against articular joint tissues. Against this background, we set out to investigate the specific functional roles and underlying regulatory networks of distinct T-cell subsets within the RA synovial microenvironment.\u003c/p\u003e\n\u003cp\u003eSingle-cell RNA sequencing (scRNA-seq) was employed to dissect the cellular composition of RA synovial tissue samples. Through unsupervised clustering analysis, we successfully identified seven distinct cell lineages: T lymphocytes, B lymphocytes, macrophages, fibroblasts, neutrophils, natural killer (NK) cells, and dendritic cells (Figure 1A). Each cell subset was further annotated based on well-established lineage-specific marker genes (Figure 1B). Notably, the relative proportion of each cell type exhibited substantial inter-patient variability (Figure 1C), which highlights the inherent cellular heterogeneity characteristic of RA pathogenesis. A heatmap illustrating the expression of key lineage marker genes (Figure 1D) revealed that IL7R, FOXP3, and CXCL13 were predominantly expressed in T-cell populations. Subsequent KEGG pathway enrichment analysis (Figure 1E) further linked T cells to multiple RA-associated biological pathways, such as Th17, Th1, and Th2 subset differentiation, hematopoietic lineage development, cytokine-receptor interactions, and cell adhesion molecule signaling pathways.\u003c/p\u003e\n\u003cp\u003eWe constructed a protein-protein interaction network from these KEGG pathways (Figure 1F) to map how T-cell molecules interact in the RA environment. Six functional modules emerged: hematopoietic lineage, Th17 differentiation, RA pathogenesis, Th1/Th2 differentiation, cytokine signaling, and cell adhesion molecule interactions. Several hub genes\u0026mdash;\u003cem\u003eIL7R\u003c/em\u003e, \u003cem\u003eCXCL13\u003c/em\u003e, \u003cem\u003eIL2RA\u003c/em\u003e, \u003cem\u003eCD40\u003c/em\u003e, \u003cem\u003eICAM1\u003c/em\u003e, and \u003cem\u003ePDCD1\u003c/em\u003e\u0026mdash;stood out due to their extensive connections. Their central positions in the network point to key roles in T-cell differentiation, inflammatory signaling, and cell adhesion, making them attractive targets for therapeutic intervention.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTranscriptional Dysregulation and Target Intersection\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe conducted differential expression analysis on the GSE89408 dataset to delineate the molecular expression profile of RA synovial tissue at the bulk transcriptomic level. Unsupervised hierarchical clustering of genome-wide expression profiles revealed a pronounced separation between RA patients and healthy controls, as shown in the heatmap (Figure 2A). The clustering dendrogram exhibited distinct disease-associated grouping, where RA samples and normal control samples formed independent branches respectively. This finding confirms that RA synovial tissue carries a distinct transcriptional signature that can be reliably distinguished from normal tissue.\u003c/p\u003e\n\u003cp\u003eUsing stringent statistical cutoffs (adjusted \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05 and |log₂ fold change| \u0026gt; 1), we identified 9,448 DEGs. Of these, 1,254 were upregulated and 119 were downregulated in RA samples compared to healthy controls (Figure 2B). The volcano plot showed that most statistically significant transcripts clustered in the upregulated region, indicating that transcriptional activation dominates the RA synovial profile.\u003c/p\u003e\n\u003cp\u003eTo pinpoint potential pharmacological targets of resveratrol, we compiled resveratrol-associated genes from four established databases: the CTD, Gene Cards, Super-PRED, and Swiss Target Prediction. CTD returned the largest set of resveratrol-related genes (n = 7,280), substantially more than Gene Cards (n = 1,254), Super-PRED (n = 119), or Swiss Target Prediction (n = 100) (Figure 2C).\u003c/p\u003e\n\u003cp\u003eWe then intersected three gene lists: T cell-specific markers (n = 230), bulk RNA-seq DEGs (n = 9,448), and resveratrol-related targets (n = 7,713). This analysis yielded 68 genes common to all three sets (Figure 2D).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunctional Characterization and Protein-Protein Interaction (PPI) Network\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eGO enrichment analysis showed that the 68-gene set is functionally concentrated in immune regulation and inflammatory response processes central to RA. At the biological process level, genes involved in T cell activation, lymphocyte proliferation, immune synapse assembly, and cytokine production were enriched. Cellular component analysis revealed localization to immune synapses and T cell receptor complexes. Molecular function enrichment highlighted immune receptor binding and protein kinase binding activities (Figure 3A and 3B).\u003c/p\u003e\n\u003cp\u003eKEGG pathway analysis demonstrated that these core genes cluster in classical inflammatory signaling cascades\u0026mdash;TNF, NF-\u0026kappa;B, and JAK\u0026ndash;STAT pathways\u0026mdash;as well as T cell receptor signaling (Figure 3D and 3E). This enrichment pattern suggests the 68 genes may collectively drive the onset and persistence of synovial inflammation in RA by modulating these signaling networks. PPI network analysis revealed complex regulatory interactions among the core genes, with CD3E and ZAP70 emerging as hub genes with substantially higher connectivity than other network members (Figure 3E).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEstablishment of a Robust 5\u0026ndash;Gene Diagnostic Model\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eComprehensive performance benchmarking was carried out on 120 classifier combinations to screen the optimal diagnostic model, and the LASSO + PLS combined algorithm achieved the highest average diagnostic accuracy (Figure 4A). This model identified a 5-gene signature:\u003cem\u003e\u0026nbsp;PTTG1\u003c/em\u003e, \u003cem\u003eNUSAP1\u003c/em\u003e, \u003cem\u003ePDE4D\u003c/em\u003e, \u003cem\u003eITGA4\u003c/em\u003e, and \u003cem\u003eBIRC3\u003c/em\u003e. In the training set, the model achieved an AUC of 0.983 (Figure 4B). External validation in cohorts GSE55457 and GSE77298 returned AUCs of 0.908 and 0.902, respectively (Figure 4C\u0026ndash;D). Decision curve analysis and calibration curve analysis were performed to verify the clinical applicability of the 5-gene diagnostic model, and the results confirmed its favorable clinical utility (Figure 4E\u0026ndash;F).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCorrelation Between the 5\u0026ndash;Gene Diagnostic Signature and Synovial Immune Microenvironment in Rheumatoid Arthritis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eUsing CIBERSORT, we quantified the infiltration of 22 immune cell subsets in RA and normal synovial tissues from the GSE89408 dataset. Pro-inflammatory immune cells were significantly more abundant in RA synovium compared to normal tissue, with activated memory CD4\u003csup\u003e+\u003c/sup\u003e T cells and M1 macrophages showing the most pronounced enrichment (Figure 5A). Analysis of immune cell crosstalk revealed that cells within the RA synovium form a complex regulatory network, characterized by enhanced interactions between adaptive and innate immune compartments (Figure 5B). Correlation analysis showed a significant positive association between the 5-gene signature and pro-inflammatory immune cell subsets in RA synovium (Figure 5C). Among the signature genes, \u003cem\u003eITGA4\u003c/em\u003e and \u003cem\u003eBIRC3\u003c/em\u003e exhibited the strongest correlations with activated memory CD4\u003csup\u003e+\u003c/sup\u003e T cells and M1 macrophages.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eGSEA of Pathways Associated with the Five Signature Genes\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo clarify the biological functions and regulatory mechanisms of the five signature genes, we performed single-gene GSEA based on KEGG pathway annotations. Each gene displayed a distinct enrichment pattern (Figure 6).\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eBIRC3\u003c/em\u003e (baculoviral IAP repeat containing 3) was significantly enriched in the antigen processing and presentation pathway (KEGG: hsa04612), suggesting its involvement in MHC-mediated immune recognition (Figure 6A\u0026ndash;B). Three genes\u0026mdash;\u003cem\u003eITGA4\u003c/em\u003e (integrin subunit alpha 4), \u003cem\u003eNUSAP1\u003c/em\u003e (nucleolar and spindle-associated protein 1), and \u003cem\u003ePTTG1\u003c/em\u003e (pituitary tumor-transforming gene 1)\u0026mdash;showed convergent enrichment in the T cell receptor signaling pathway (KEGG: hsa04660), indicating their coordinated roles in T cell activation and differentiation (Figure 6C\u0026ndash;D). \u003cem\u003ePDE4D\u003c/em\u003e (phosphodiesterase 4D) was notably enriched in the autophagy regulation pathway (KEGG: hsa04140), pointing to its function in maintaining cellular homeostasis and modulating inflammatory responses (Figure 6E).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMolecular Docking Validation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eResults from molecular docking simulations verified that resveratrol can bind efficiently to all five target proteins encoded by the diagnostic signature, forming stable hydrogen bond interactions with key amino acid residues within the binding pocket of each target protein. Specifically, resveratrol established interactions with Asp201, His200, and Gln369 of phosphodiesterase 4D (\u003cem\u003ePDE4D\u003c/em\u003e); Gln328 and Arg332 of baculoviral IAP repeat-containing protein 3 (\u003cem\u003eBIRC3\u003c/em\u003e); Arg383 and Asn329 of integrin \u0026alpha;4 (\u003cem\u003eITGA4\u003c/em\u003e); Leu2034 and Lys1901 of pituitary tumor-transforming gene 1 (\u003cem\u003ePTTG1\u003c/em\u003e); and His407 of nucleolar and spindle-associated protein 1 (\u003cem\u003eNUSAP1\u003c/em\u003e) (Figure 7A\u0026ndash;E).\u003c/p\u003e\n\u003cp\u003eBinding affinity measurements revealed variation in the strength of these receptor-ligand complexes. The binding energies of resveratrol with \u003cem\u003ePDE4D\u003c/em\u003e, \u003cem\u003ePTTG1\u003c/em\u003e,\u003cem\u003e\u0026nbsp;ITGA4\u003c/em\u003e, \u003cem\u003eNUSAP1\u003c/em\u003e, and \u003cem\u003eBIRC3\u003c/em\u003e were -8.0 kcal/mol, -6.8 kcal/mol, -6.3 kcal/mol, -5.6 kcal/mol, and -5.5 kcal/mol, respectively (Figure 7F). All binding free energy values were within the low-energy range, which is conducive to the formation of stable resveratrol-target protein complexes.\u0026nbsp;\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eWe integrated single-cell transcriptomics, bulk RNA sequencing, machine learning, and computational pharmacology technologies into a unified analytical pipeline to address the unsolved issues in current RA biomarker research. The findings of this study mainly revealed five key research outcomes. T-cell profiling in RA synovium uncovered subpopulations with markedly different activation states and effector functions\u0026mdash;a level of heterogeneity not fully appreciated in earlier bulk-tissue studies. Cross-referencing T-cell marker genes, RA-linked differentially expressed genes, and predicted resveratrol targets yielded 68 genes shared across all three lists. Running 120 machine learning combinations against our gene pool ultimately narrowed the field to five candidates\u0026mdash;\u003cem\u003ePTTG1\u003c/em\u003e, \u003cem\u003eNUSAP1\u003c/em\u003e, \u003cem\u003ePDE4D\u003c/em\u003e, \u003cem\u003eITGA4\u003c/em\u003e, and \u003cem\u003eBIRC3\u003c/em\u003e\u0026mdash;whose joint performance proved consistently strong. This signature demonstrated strong diagnostic performance (AUC\u0026thinsp;\u0026gt;\u0026thinsp;0.90) across multiple validation cohorts. In RA synovium, higher expression of these five genes went hand-in-hand with greater M1 macrophage and CD4⁺ T cell infiltration, hinting at a shared regulatory axis. Docking runs placed resveratrol at favorable binding poses on all five proteins, which, while preliminary, gives a structural rationale for treating these targets pharmacologically.\u003c/p\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003eComparison with Existing RA Biomarkers and Diagnostic Signatures\u003c/h2\u003e \u003cp\u003eCompared with the RA diagnostic biomarkers reported in previous studies, the 5-gene signature constructed in this study has prominent advantages in both clinical application and biological relevance. At present, clinical serological diagnosis of RA mainly depends on RF and ACPA, but both of these two serological markers have the defect of unstable detection sensitivity, and cannot accurately reflect the dynamic changes of RA disease activity in clinical follow-up [\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e, \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e]. Reading gene expression directly from the inflamed joint cuts out the noise introduced when blood is used as a proxy tissue. Blood-based panels\u0026mdash;Raterman et al. [\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e] and Tao et al. [\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e] being notable examples\u0026mdash;pick up whole-body immune changes and may miss what is happening specifically inside the joint. Our synovial-based approach avoids this drawback. Each gene in the panel also has a paper trail: prior studies tie them individually to T-cell activation, mitotic regulation, or the inflammatory signaling that erodes cartilage and bone in RA.\u003c/p\u003e \u003cp\u003eEach of the five signature genes has documented links to immune function and inflammatory processes. \u003cem\u003ePTTG1\u003c/em\u003e encodes securin; beyond its canonical role in sister-chromatid separation, securin appears to fuel T-cell expansion and is overexpressed in a range of autoimmune diseases [\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e]. The protein encoded by \u003cem\u003eNUSAP1\u003c/em\u003e organizes microtubules during cell division; work in lymphocytes suggests it also plays a role when these cells receive activation signals [\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e]. \u003cem\u003eITGA4\u003c/em\u003e encodes integrin α4, which has already proven itself as a therapeutic target\u0026mdash;natalizumab, an antibody blocking α4 integrin, works effectively in multiple sclerosis and Crohn\u0026rsquo;s disease [\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e]. \u003cem\u003ecIAP2\u003c/em\u003e, the product of \u003cem\u003eBIRC3\u003c/em\u003e, sits upstream of NF-κB and appears to shield autoreactive T cells from apoptosis\u0026mdash;a property with obvious relevance to chronic inflammation [\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e]. \u003cem\u003ePDE4D\u003c/em\u003e encodes phosphodiesterase 4D, an enzyme controlling c AMP levels that serves as an anti-inflammatory target; apremilast, which inhibits \u003cem\u003ePDE4\u003c/em\u003e, has received approval for treating psoriatic arthritis [\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e].\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003eImplications for RA Prognosis and Treatment Response Prediction\u003c/h2\u003e \u003cp\u003eDiagnostic performance aside, the signature may say something about where a patient\u0026rsquo;s disease is headed and how it will respond to treatment. The tight link we found between signature expression and infiltrating immune cells raises the possibility that these five genes track disease severity in real time. Patients with the highest \u003cem\u003eITGA4\u003c/em\u003e and \u003cem\u003eBIRC3\u003c/em\u003e readings also had the densest M1 macrophage infiltrates, a pattern that has historically predicted faster radiographic deterioration [\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e]. Whether \u003cem\u003ePDE4D\u003c/em\u003e expression marks out patients who stand to gain most from \u003cem\u003ePDE4\u003c/em\u003e inhibition is worth testing prospectively [\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eTreatment response prediction is a natural next question. Sorting patients by molecular profile before prescribing has gained traction as a strategy for improving RA outcomes [\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e, \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e]. The five genes in our signature operate within pathways that conventional and biologic DMARDs target. \u003cem\u003eBIRC3\u003c/em\u003e, for example, affects NF-κB signaling\u0026mdash;a pathway that TNF inhibitors shut down [\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e]\u0026mdash;while \u003cem\u003eITGA4\u003c/em\u003e is directly targeted by integrin-blocking drugs [\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e]. A prospective study measuring these genes at baseline\u0026mdash;before any drug is started\u0026mdash;would tell us whether the signature has genuine predictive value for treatment selection.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003eResearch Limitations and Future Directions\u003c/h2\u003e \u003cp\u003eOur findings come with caveats worth stating plainly. Working from archived public datasets means batch effects and platform differences could have crept into the expression measurements in ways we cannot fully correct [\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e]. All three cohorts skewed toward established, classified RA; the signature has not yet faced the harder test of early or undifferentiated arthritis [\u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e]. It should be noted that the binding affinity scores obtained from molecular docking simulations are only theoretical computational results, and the actual binding ability between resveratrol and these target proteins needs to be further verified by in vitro cell-free binding experiments and protein-ligand co-crystallography technology before drawing definite conclusions [\u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e]. The source datasets drew predominantly from East Asian and European populations, so performance in other groups is an open question [\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThe most pressing next step is enrolling patients with early or undifferentiated arthritis to see whether the signature flags future RA before classification criteria are met [\u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e, \u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e]. Knocking out or silencing each gene individually in T-cell or macrophage models\u0026mdash;using CRISPR/Cas9\u0026mdash;would pin down which steps in the inflammatory cascade each one controls [\u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e60\u003c/span\u003e]. mRNA levels do not always predict protein abundance; pairing this transcriptomic dataset with proteomics and metabolomics would expose post-transcriptional controls that the current analysis misses [\u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e61\u003c/span\u003e, \u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e62\u003c/span\u003e]. Resveratrol\u0026rsquo;s notoriously poor bioavailability has hampered earlier trials; formulation work\u0026mdash;nanoparticle encapsulation or phospholipid complexes\u0026mdash;should precede any definitive randomized trial in RA patients [\u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e63\u003c/span\u003e, \u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e64\u003c/span\u003e].\u003c/p\u003e \u003c/div\u003e"},{"header":"Conclusions","content":"\u003cp\u003eTaken together, our data support a five-gene panel\u0026mdash;\u003cem\u003ePTTG1\u003c/em\u003e, \u003cem\u003eNUSAP1\u003c/em\u003e, \u003cem\u003ePDE4D\u003c/em\u003e, \u003cem\u003eITGA4\u003c/em\u003e, and \u003cem\u003eBIRC3\u003c/em\u003e\u0026mdash;as a synovial tissue-based diagnostic tool for RA, one that held up across two independent cohorts tested here. The same genes track with the immune cell infiltrates that drive synovitis, suggesting they do more than discriminate cases from controls; they may index disease activity. Docking analysis adds a structural argument that resveratrol could modulate all five targets, though this remains a hypothesis until tested in the laboratory. Moving from computational prediction to bedside application will require prospective trials and mechanistic experiments, but the convergent evidence assembled here gives those efforts a defined molecular starting point.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgments\u003c/strong\u003e: Not applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding:\u003c/strong\u003e This study was supported by Ningxia Medical University School-level Project (Grant No.: XY2025046).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor Contributions\u003c/strong\u003e: Conception and design: Nan Yu; Methodology: Yan Zhou, Jia Guo; Formal analysis: Yan Zhou, Xiaoqing Bao; Investigation: Yan Zhou, Jia Guo; Data curation: Xiaoqing Bao; Writing \u0026ndash; original draft: Yan Zhou; Writing \u0026ndash; review and editing: Nan Yu; Supervision: Nan Yu; Project administration: Nan Yu; Funding acquisition: Nan Yu.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompliance with Ethical Standards Declarations\u003c/strong\u003e: In accordance with the Declaration of Helsinki, this research is based on open-source data, so ethics approval was waived.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting Interests\u003c/strong\u003e: The authors have no conflicts of interest to declare.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eSmolen JS, Aletaha D, Barton A, Burmester GR, Emery P, Firestein GS, Kavanaugh A, McInnes IB, Solomon DH, Strand V, et al (2018) Rheumatoid arthritis. Nat Rev Dis Primers 4:18001. https://doi.org/10.1038/nrdp.2018.1\u003c/li\u003e\n\u003cli\u003eFirestein GS, McInnes IB (2017) Immunopathogenesis of rheumatoid arthritis. Immunity 46:183\u0026ndash;196. https://doi.org/10.1016/j.immuni.2017.02.006\u003c/li\u003e\n\u003cli\u003eMcInnes IB, Schett G (2011) The pathogenesis of rheumatoid arthritis. N Engl J Med 365:2205\u0026ndash;2219. https://doi.org/10.1056/NEJMra1004965\u003c/li\u003e\n\u003cli\u003eSmolen JS, Aletaha D, McInnes IB (2016) Rheumatoid arthritis. Lancet 388:2023\u0026ndash;2038. https://doi.org/10.1016/S0140-6736(16)30173-8\u003c/li\u003e\n\u003cli\u003eSmolen JS, Landew\u0026eacute; RBM, Bijlsma JWJ, Burmester GR, Dougados M, Kerschbaumer A, McInnes IB, Sepriano A, van Vollenhoven RF, de Wit M, et al (2020) EULAR recommendations for the management of rheumatoid arthritis with synthetic and biological disease-modifying antirheumatic drugs: 2019 update. Ann Rheum Dis 79:685\u0026ndash;699. https://doi.org/10.1136/annrheumdis-2019-216655\u003c/li\u003e\n\u003cli\u003eBurmester GR, Pope JE (2017) Novel treatment strategies in rheumatoid arthritis. Lancet 389:2338\u0026ndash;2348. https://doi.org/10.1016/S0140-6736(17)31491-5\u003c/li\u003e\n\u003cli\u003eWeyand CM, Goronzy JJ (2021) The immunology of rheumatoid arthritis. Nat Immunol 22:10\u0026ndash;18. https://doi.org/10.1038/s41590-020-00816-x\u003c/li\u003e\n\u003cli\u003eCope AP, Schulze-Koops H, Aringer M (2007) The central role of T cells in rheumatoid arthritis. Clin Exp Rheumatol 25(Suppl 46):S4\u0026ndash;S11\u003c/li\u003e\n\u003cli\u003eChemin K, Gerstner C, Malmstr\u0026ouml;m V (2019) Effector functions of CD4\u003csup\u003e+\u003c/sup\u003e T cells at the site of local autoimmune inflammation\u0026mdash;Lessons from rheumatoid arthritis. Front Immunol 10:353. https://doi.org/10.3389/fimmu.2019.00353\u003c/li\u003e\n\u003cli\u003eMiossec P, Korn T, Kuchroo VK (2009) Interleukin-17 and type 17 helper T cells. N Engl J Med 361:888\u0026ndash;898. https://doi.org/10.1056/NEJMra0707449\u003c/li\u003e\n\u003cli\u003eLubberts E (2015) The IL-23\u0026ndash;IL-17 axis in inflammatory arthritis. Nat Rev Rheumatol 11:415\u0026ndash;429. https://doi.org/10.1038/nrrheum.2015.53\u003c/li\u003e\n\u003cli\u003eNie H, Zheng Y, Li R, Guo TB, He D, Fang L, Liu X, Xiao L, Chen X, Wan B, et al (2013) Phosphorylation of FOXP3 controls regulatory T cell function and is inhibited by TNF-\u0026alpha; in rheumatoid arthritis. Nat Med 19:322\u0026ndash;328. https://doi.org/10.1038/nm.3085\u003c/li\u003e\n\u003cli\u003eKomatsu N, Okamoto K, Sawa S, Nakashima T, Oh-hora M, Kodama T, Tanaka S, Bluestone JA, Takayanagi H (2014) Pathogenic conversion of Foxp3+ T cells into TH17 cells in autoimmune arthritis. Nat Med 20:62\u0026ndash;68. https://doi.org/10.1038/nm.3432\u003c/li\u003e\n\u003cli\u003eAletaha D, Neogi T, Silman AJ, Funovits J, Felson DT, Bingham CO, Birnbaum NS, Burmester GR, Bykerk VP, Cohen MD, et al (2010) 2010 Rheumatoid arthritis classification criteria: An American College of Rheumatology/European League Against Rheumatism collaborative initiative. Arthritis Rheum 62:2569\u0026ndash;2581. https://doi.org/10.1002/art.27584\u003c/li\u003e\n\u003cli\u003eNishimura K, Sugiyama D, Kogata Y, Tsuji G, Nakazawa T, Kawano S, Saigo K, Morinobu A, Koshiba M, Kuntz KM, et al (2007) Meta-analysis: Diagnostic accuracy of anti-cyclic citrullinated peptide antibody and rheumatoid factor for rheumatoid arthritis. Ann Intern Med 146:797\u0026ndash;808. https://doi.org/10.7326/0003-4819-146-11-200706050-00008\u003c/li\u003e\n\u003cli\u003eSteiner G, Tesar V (2009) Diagnostic value of anti-citrullinated protein antibodies in rheumatoid arthritis. Ann N Y Acad Sci 1173:439\u0026ndash;447. https://doi.org/10.1111/j.1749-6632.2009.04863.x\u003c/li\u003e\n\u003cli\u003evan der Helm-van Mil AHM, Huizinga TWJ (2008) Advances in the genetics of rheumatoid arthritis point to subclassification into distinct disease subsets. Arthritis Res Ther 10:205. https://doi.org/10.1186/ar2384\u003c/li\u003e\n\u003cli\u003eSmolen JS, Aletaha D (2004) Patients with rheumatoid arthritis in clinical care. Ann Rheum Dis 63:221\u0026ndash;225. https://doi.org/10.1136/ard.2003.012252\u003c/li\u003e\n\u003cli\u003eStrimbu K, Tavel JA (2010) What are biomarkers? Curr Opin HIV AIDS 5:463\u0026ndash;466. https://doi.org/10.1097/COH.0b013e32833ed177\u003c/li\u003e\n\u003cli\u003eBrazma A, Hingamp P, Quackenbush J, Sherlock G, Spellman P, Stoeckert C, Aach J, Ansorge W, Ball CA, Causton HC, et al (2001) Minimum information about a microarray experiment (MIAME)\u0026mdash;Toward standards for microarray data. Nat Genet 29:365\u0026ndash;371. https://doi.org/10.1038/ng1201-365\u003c/li\u003e\n\u003cli\u003eWang Z, Gerstein M, Snyder M (2009) RNA-Seq: A revolutionary tool for transcriptomics. Nat Rev Genet 10:57\u0026ndash;63. https://doi.org/10.1038/nrg2484\u003c/li\u003e\n\u003cli\u003evan der Pouw Kraan TC, Wijbrandts CA, Drijfhout JW, Baggen JM, Verweij CL, Tak PP (2008) Responsiveness to anti-tumour necrosis factor alpha therapy is related to pre-treatment tissue inflammation levels in rheumatoid arthritis patients. Ann Rheum Dis 67:563\u0026ndash;566. https://doi.org/10.1136/ard.2007.081950\u003c/li\u003e\n\u003cli\u003eDennis G Jr, Holweg CT, Kummerfeld SK, Choy DF, Setiadi AF, Hackney JA, Haverty PM, Gilbert H, Lin WY, Diehl L, et al (2014) Synovial phenotypes in rheumatoid arthritis correlate with response to biologic therapeutics. Arthritis Res Ther 16: R90. https://doi.org/10.1186/ar4555\u003c/li\u003e\n\u003cli\u003eBaur JA, Sinclair DA (2006) Therapeutic potential of resveratrol: The in vivo evidence. Nat Rev Drug Discov 5:493\u0026ndash;506. https://doi.org/10.1038/nrd2060\u003c/li\u003e\n\u003cli\u003eMalaguarnera L (2019) Influence of resveratrol on the immune response. Nutrients 11:946. https://doi.org/10.3390/nu11050946\u003c/li\u003e\n\u003cli\u003eShakibaei M, Buhrmann C, Mobasheri A (2011) Resveratrol-mediated SIRT-1 interactions with p300 modulate receptor activator of NF-\u0026kappa;B ligand (RANKL) activation of NF-\u0026kappa;B signaling and inhibit osteoclastogenesis in bone-derived cells. J Biol Chem 286:11492\u0026ndash;11505. https://doi.org/10.1074/jbc.M110.198713\u003c/li\u003e\n\u003cli\u003eXuzhu G, Komai-Koma M, Leung BP, Howe HS, McSharry C, McInnes IB, Xu D (2012) Resveratrol modulates murine collagen-induced arthritis by inhibiting Th17 and B-cell function. Ann Rheum Dis 71:129\u0026ndash;135. https://doi.org/10.1136/ard.2011.149831\u003c/li\u003e\n\u003cli\u003eHopkins AL (2008) Network pharmacology: The next paradigm in drug discovery. Nat Chem Biol 4:682\u0026ndash;690. https://doi.org/10.1038/nchembio.118\u003c/li\u003e\n\u003cli\u003eGreener JG, Kandathil SM, Moffat L, Jones DT (2022) A guide to machine learning for biologists. Nat Rev Mol Cell Biol 23:40\u0026ndash;55. https://doi.org/10.1038/s41580-021-00407-0\u003c/li\u003e\n\u003cli\u003eChicco D (2017) Ten quick tips for machine learning in computational biology. Bio Data Min 10:35. https://doi.org/10.1186/s13040-017-0155-3\u003c/li\u003e\n\u003cli\u003eWalsh I, Fishman D, Garcia-Gasulla D, Titber T, Harber G, Lordge J, Tosatto SCE, Capriotti E (2021) DOME: Recommendations for supervised machine learning validation in biology. Nat Methods 18:1122\u0026ndash;1127. https://doi.org/10.1038/s41592-021-01205-4\u003c/li\u003e\n\u003cli\u003eLiu Z, Liu L, Weng S, Guo C, Dang Q, Xu H, Wang L, Lu T, Zhang Y, Sun Z, et al (2022) Machine learning-based integration develops an immune-derived lncRNA signature for improving outcomes in colorectal cancer. Nat Commun 13:816. https://doi.org/10.1038/s41467-022-28421-6\u003c/li\u003e\n\u003cli\u003eFriedman J, Hastie T, Tibshirani R (2010) Regularization paths for generalized linear models via coordinate descent. J Stat Softw 33:1\u0026ndash;22. https://doi.org/10.18637/jss.v033.i01\u003c/li\u003e\n\u003cli\u003eChen T, Guestrin C (2016) XGBoost: A scalable tree boosting system. In: Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, San Francisco, CA, USA, 13\u0026ndash;17 August 2016, pp 785\u0026ndash;794. https://doi.org/10.1145/2939672.2939785\u003c/li\u003e\n\u003cli\u003eStuart T, Butler A, Hoffman P, Hafemeister C, Papalexi E, Mauck WM, Hao Y, Stoeckius M, Smibert P, Satija R (2019) Comprehensive integration of single-cell data. Cell 177:1888\u0026ndash;1902.e21. https://doi.org/10.1016/j.cell.2019.05.031\u003c/li\u003e\n\u003cli\u003eNewman AM, Liu CL, Green MR, Gentles AJ, Feng W, Xu Y, Hoang CD, Diehn M, Alizadeh AA (2015) Robust enumeration of cell subsets from tissue expression profiles. Nat Methods 12:453\u0026ndash;457. https://doi.org/10.1038/nmeth.3337\u003c/li\u003e\n\u003cli\u003eWu T, Hu E, Xu S, Chen M, Guo P, Dai Z, Feng T, Zhou L, Tang W, Zhan L, et al (2021) clusterProfiler 4.0: A universal enrichment tool for interpreting omics data. Innovation 2:100141. https://doi.org/10.1016/j.xinn.2021.100141\u003c/li\u003e\n\u003cli\u003eRitchie ME, Phipson B, Wu D, Hu Y, Law CW, Shi W, Smyth GK (2015) limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res 43: e47. https://doi.org/10.1093/nar/gkv007\u003c/li\u003e\n\u003cli\u003eMorris GM, Huey R, Lindstrom W, Sanner MF, Belew RK, Goodsell DS, Olson AJ (2009) AutoDock4 and AutoDockTools4: Automated docking with selective receptor flexibility. J Comput Chem 30:2785\u0026ndash;2791. https://doi.org/10.1002/jcc.21256\u003c/li\u003e\n\u003cli\u003eTrott O, Olson AJ (2010) AutoDock Vina: Improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading. J Comput Chem 31:455\u0026ndash;461. https://doi.org/10.1002/jcc.21334\u003c/li\u003e\n\u003cli\u003eVlotides G, Eigler T, Melmed S (2007) Pituitary tumor-transforming gene: Physiology and implications for tumorigenesis. Endocr Rev 28:165\u0026ndash;186. https://doi.org/10.1210/er.2006-0042\u003c/li\u003e\n\u003cli\u003eRaemaekers T, Ribbeck K, Beaudouin J, Annaert W, Van Camp M, Stockmans I, Smets N, Bouillon R, Ellenberg J, Carmeliet G (2003) NuSAP, a novel microtubule-associated protein involved in mitotic spindle organization. J Cell Biol 162:1017\u0026ndash;1029. https://doi.org/10.1083/jcb.200302129\u003c/li\u003e\n\u003cli\u003eHouslay MD, Schafer P, Zhang KY (2005) Keynote review: Phosphodiesterase-4 as a therapeutic target. Drug Discov Today 10:1503\u0026ndash;1519. https://doi.org/10.1016/S1359-6446(05)03622-6\u003c/li\u003e\n\u003cli\u003eJin SL, Ding SL, Lin SC (2012) Phosphodiesterase 4 and its inhibitors in inflammatory diseases. Chang Gung Med J 35:197\u0026ndash;210\u003c/li\u003e\n\u003cli\u003eYednock TA, Cannon C, Fritz LC, Sanchez-Madrid F, Steinman L, Karin N (1992) Prevention of experimental autoimmune encephalomyelitis by antibodies against alpha4beta1 integrin. Nature 356:63\u0026ndash;66. https://doi.org/10.1038/356063a0\u003c/li\u003e\n\u003cli\u003eVarfolomeev E, Vucic D (2011) Inhibitor of apoptosis proteins: Fascinating biology leads to attractive tumor therapeutic targets. Future Oncol 7:633\u0026ndash;648. https://doi.org/10.2217/fon.11.40\u003c/li\u003e\n\u003cli\u003eRaterman HG, Vosslamber S, de Ridder S, Nurmohamed MT, Lems WF, Boers M, van de Wiel M, Dijkmans BA, Verweij CL, Voskuyl AE (2012) The interferon type I signature towards prediction of non-response to rituximab in rheumatoid arthritis patients. Arthritis Res Ther 14: R95. https://doi.org/10.1186/ar3819\u003c/li\u003e\n\u003cli\u003eLewis MJ, Barnes MR, Blighe K, Goldmann K, Rana S, Hackney JA, Ramamoorthi N, John CR, Watson DS, Kummerfeld SK, et al (2019) Molecular portraits of early rheumatoid arthritis identify clinical and treatment response phenotypes. Cell Rep 28:2455\u0026ndash;2470.e5. https://doi.org/10.1016/j.celrep.2019.07.091\u003c/li\u003e\n\u003cli\u003eSchafer PH, Parton A, Capone L, Cedzik D, Brady H, Evans JF, Man HW, Muller GW, Stirling DI, Chopra R (2014) Apremilast is a selective PDE4 inhibitor with regulatory effects on innate immunity. Cell Signal 26:2016\u0026ndash;2029. https://doi.org/10.1016/j.cellsig.2014.05.014\u003c/li\u003e\n\u003cli\u003eRiveiro-Naveira RR, Valc\u0026aacute;rcel-Ares MN, Almonte-Becerril M, Vaamonde-Garc\u0026iacute;a C, Loureiro J, Hermida-Carballo L, L\u0026oacute;pez-Pel\u0026aacute;ez E, Blanco FJ, L\u0026oacute;pez-Armada MJ (2016) Resveratrol lowers synovial hyperplasia, inflammatory markers and oxidative damage in an acute antigen-induced arthritis model. Rheumatology 55:1889\u0026ndash;1900. https://doi.org/10.1093/rheumatology/kew255\u003c/li\u003e\n\u003cli\u003eWalle T, Hsieh F, De Legge MH, Oatis JE Jr, Walle UK (2004) High absorption but very low bioavailability of oral resveratrol in humans. Drug Metab Dispos 32:1377\u0026ndash;1382. https://doi.org/10.1124/dmd.104.000885\u003c/li\u003e\n\u003cli\u003eHayden MS, Ghosh S (2008) Shared principles in NF-\u0026kappa;B signaling. Cell 132:344\u0026ndash;362. https://doi.org/10.1016/j.cell.2008.01.020\u003c/li\u003e\n\u003cli\u003ePitzalis C, Kelly S, Humby F (2013) New learnings on the pathophysiology of RA from synovial biopsies. Curr Opin Rheumatol 25:334\u0026ndash;344. https://doi.org/10.1097/BOR.0b013e32835fd8eb\u003c/li\u003e\n\u003cli\u003eMcCann FE, Palfreeman AC, Andrews M, Sherlock JP, Williams RO, Brennan FM (2010) Apremilast, a novel PDE4 inhibitor, inhibits spontaneous production of tumour necrosis factor-alpha from human rheumatoid synovial cells and ameliorates experimental arthritis. Arthritis Res Ther 12: R107. https://doi.org/10.1186/ar3041\u003c/li\u003e\n\u003cli\u003eLeek JT, Scharpf RB, Bravo HC, Simcha D, Langmead B, Johnson WE, Geman D, Baggerly K, Irizarry RA (2010) Tackling the widespread and critical impact of batch effects in high-throughput data. Nat Rev Genet 11:733\u0026ndash;739. https://doi.org/10.1038/nrg2825\u003c/li\u003e\n\u003cli\u003evan der Helm-van Mil AHM, le Cessie S, van Dongen H, Brinkman J, de Craen AJM, Huizinga TWJ (2007) A prediction rule for disease outcome in patients with recent-onset undifferentiated arthritis: How to guide individual treatment decisions. Arthritis Rheum 56:433\u0026ndash;440. https://doi.org/10.1002/art.22380\u003c/li\u003e\n\u003cli\u003eCopeland RA, Pompliano DL, Meek TD (2006) Drug-target residence time and its implications for lead optimization. Nat Rev Drug Discov 5:730\u0026ndash;739. https://doi.org/10.1038/nrd2082\u003c/li\u003e\n\u003cli\u003ePopejoy AB, Fullerton SM (2016) Genomics is failing on diversity. Nature 538:161\u0026ndash;164. https://doi.org/10.1038/538161a\u003c/li\u003e\n\u003cli\u003eGerlag DM, Raza K, van Baarsen LGM, Brouwer E, Tak PP (2012) EULAR recommendations for terminology and research in individuals at risk of rheumatoid arthritis: Report from the Study Group for Risk Factors for Rheumatoid Arthritis. Ann Rheum Dis 71:638\u0026ndash;641. https://doi.org/10.1136/annrheumdis-2011-200990\u003c/li\u003e\n\u003cli\u003eCong L, Ran FA, Cox D, Lin S, Barretto R, Habib N, Hsu PD, Wu X, Jiang W, Marraffini LA, et al (2013) Multiplex genome engineering using CRISPR/Cas systems. Science 339:819\u0026ndash;823. https://doi.org/10.1126/science.1231143\u003c/li\u003e\n\u003cli\u003eHasin Y, Seldin M, Lusis A (2017) Multi-omics approaches to disease. Genome Biol 18:83. https://doi.org/10.1186/s13059-017-1215-1\u003c/li\u003e\n\u003cli\u003eKarczewski KJ, Snyder MP (2018) Integrative omics for health and disease. Nat Rev Genet 19:299\u0026ndash;310. https://doi.org/10.1038/nrg.2018.4\u003c/li\u003e\n\u003cli\u003ePangeni R, Sahni JK, Ali J, Sharma S, Baboota S (2014) Resveratrol: Review on therapeutic potential and recent advances in drug delivery. Expert Opin Drug Deliv 11:1285\u0026ndash;1298. https://doi.org/10.1517/17425247.2014.919253\u003c/li\u003e\n\u003cli\u003eKhojah HM, Ahmed S, Abdel-Rahman MS, Elhakeim EH (2018) Resveratrol as an effective adjuvant therapy in the management of rheumatoid arthritis: A clinical study. Clin Rheumatol 37:2035\u0026ndash;2042. https://doi.org/10.1007/s10067-018-4080-8\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Rheumatoid arthritis, Single-cell RNA sequencing, T cells, Machine learning benchmarking, Resveratrol, Molecular docking","lastPublishedDoi":"10.21203/rs.3.rs-9098862/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9098862/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eRheumatoid arthritis (RA) is a systemic autoimmune disease featured by persistent synovial inflammation and multi-organ injury, with T lymphocytes as the core driver of its pathological progression. However, the molecular mechanisms of abnormal T cell functional regulation and the exact biological targets of resveratrol (a natural immunomodulatory polyphenol) in RA synovial microenvironment remain unclear. This study aimed to construct a T cell-focused RA diagnostic classifier and identify resveratrol-modulated molecular targets via multi-omics integration.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eWe integrated scRNA-seq and bulk transcriptome data of RA synovial tissues to characterize T cell-specific gene expression, intersected resveratrol targets from pharmacological databases with RA-related differentially expressed genes, and performed systematic benchmarking of 10 machine learning algorithms to screen stable diagnostic gene signatures (validated in external cohorts). CIBERSORT quantified synovial immune cell infiltration, and molecular docking evaluated resveratrol-target protein binding interactions.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eDistinct T cell subsets with differential resveratrol target gene expression were identified in RA synovium. A 5-gene signature (\u003cem\u003ePTTG1\u003c/em\u003e, \u003cem\u003eNUSAP1\u003c/em\u003e, \u003cem\u003ePDE4D\u003c/em\u003e, \u003cem\u003eITGA4\u003c/em\u003e, \u003cem\u003eBIRC3\u003c/em\u003e) was screened, showing excellent diagnostic performance (training cohort AUC\u0026thinsp;=\u0026thinsp;0.983; external cohorts GSE55457 AUC\u0026thinsp;=\u0026thinsp;0.908, GSE77298 AUC\u0026thinsp;=\u0026thinsp;0.902). Its expression was positively correlated with activated T cell and pro-inflammatory macrophage infiltration. Molecular docking revealed stable binding of resveratrol to the five target proteins (binding free energy: -5.5 to -8.0 kcal/mol).\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e \u003cp\u003eWe constructed a robust T cell-centered 5-gene signature for RA diagnosis, and these five genes are potential molecular targets of resveratrol in RA immunotherapy. This study provides a novel molecular basis and research direction for developing synovitis-targeted therapeutic strategies for RA.\u003c/p\u003e","manuscriptTitle":"A Five-Gene Signature Identified by Integrating Single-Cell Analysis and Machine Learning Benchmarking Enables Rheumatoid Arthritis Diagnosis and Resveratrol Target Discovery","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-03-20 19:04:43","doi":"10.21203/rs.3.rs-9098862/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"5f5d9c2a-0de0-4ca4-93e2-65b0a89b0c6d","owner":[],"postedDate":"March 20th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-03-27T20:20:05+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-20 19:04:43","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9098862","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9098862","identity":"rs-9098862","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}