This study utilizes a large-scale longitudinal cohort to compare the plasma proteomic profiles of healthy individuals, individuals at high risk for rheumatoid arthritis (RA), and RA patients. Through this, it identifies a set of key protein biomarkers closely associated with disease progression, disease activity, and Anti-Citrullinated Protein Antibody (ACPA) levels. Furthermore, the research unearths signature proteins linked to treatment responses for Methotrexate (MTX) combined with either Leflunomide (LEF) or Hydroxychloroquine (HCQ). By leveraging machine learning, the authors constructed a predictive model for treatment response, which demonstrated robust performance and clinical potential in an independent validation cohort. Overall, this study provides a crucial molecular foundation for early RA screening and personalized therapy, paving the way for the translational application of protein biomarkers in the precision management of RA.

This study utilized large-scale plasma proteomic profiling to analyze a longitudinal cohort encompassing 182 ACPA-positive RA patients, 67 ACPA-negative RA patients, 60 high-risk individuals, and 99 healthy controls. To characterize the molecular transitions from high-risk states to clinical onset, we followed 38 high-risk individuals over time, specifically identifying individuals who progressed to RA. Data analysis integrated stratified clustering and the DE-SWAN sliding window algorithm to identify age- and gender-specific protein variations, while multivariate linear models and LOESS smoothing were applied to map protein dynamics against disease activity (DAS28-CRP) and clinical parameters such as TJC, SJC, and VAS. Finally, the study leveraged a LASSO-based machine learning framework to develop and validate treatment response models for 206 patients receiving either Methotrexate (MTX) combined with Leflunomide (LEF) or MTX with Hydroxychloroquine (HCQ).

The proteomic analysis revealed distinct molecular signatures across the four cohorts, showing that neutrophil degranulation and antigen presentation are activated in high-risk individuals well before the appearance of clinical symptoms. In converters, a significant decline in complement components and an upregulation of the immunoproteasome (e.g., PSMB7) were observed, highlighting the roles of lipid metabolism, neutrophil extracellular traps (NETs), and iron homeostasis in disease progression. The study identified a critical molecular shift in women around age 45, where decreasing protein diversity coincided with increased inflammatory markers like CRP and SAA2. Furthermore, six protein clusters were identified that correlate with disease activity, pinpointing biological shift at DAS28-CRP levels of 3.1, 3.8, and 5.0, which correspond to transitions in complement activation and oxidative stress. The machine learning models demonstrated high predictive accuracy, with the MTX+LEF and MTX+HCQ models achieving AUC values of 0.90 and 0.86, respectively, in the independent validation cohort

This research establishes a comprehensive longitudinal proteomic map of RA, uncovering the fundamental molecular mechanisms that drive disease evolution from the pre-clinical phase to active inflammation. By identifying age-specific molecular windows and dynamic activity thresholds, the study provides a robust theoretical basis for more precise monitoring of disease progression. Ultimately, the high-performance machine learning models developed here offer powerful new tools for early screening and the realization of personalized, precision-guided therapeutic strategies in the management of Rheumatoid Arthritis.