Abstract
Psoriatic arthritis (PsA) lacks reliable biomarkers to support early diagnosis and disease stratification. This study aimed to identify protein signatures associated with PsA and develop a clinically relevant model integrating molecular and clinical features. We conducted a cross-sectional study including 143 patients with PsA and 101 controls across three cohorts. The exploratory cohort included 77 patients with PsA and 50 symptomatic controls without inflammatory disease. Two independent cohorts were used for external validation: validation cohort 1 comprised PsA patients receiving conventional DMARDs, and validation cohort 2 included PsA patients receiving biological treatment. Proteomic profiling of 384 inflammation-related proteins was performed in PBMCs using Olink technology. Machine learning-based approaches were applied to identify and optimize a diagnostic biomarker model integrating proteomic and clinical variables. Functional in vitro assays were conducted to evaluate the mechanistic regulation of biomarker-associated proteins. Sixty-eight proteins were significantly dysregulated in PsA, highlighting activation of immune-inflammatory pathways. Several proteins correlated with arthritis severity, psoriasis burden, and CRP. A five-parameter diagnostic model (ESR, plaque psoriasis, LILRB4, ADGRE2, SIGLEC10) showed excellent discrimination across all three cohorts. The upregulation of LILRB4, ADGRE2, and SIGLEC10 in PBMCs exposed to PsA serum supported their mechanistic relevance. Distinct proteomic signatures differentiated early from established PsA, revealing molecular patterns linked to disease progression. This study identifies a potential protein signature associated with PsA and provides a validated diagnostic model with high accuracy. These findings advance molecularly guided diagnosis and highlight biomarkers with both mechanistic and translational relevance.