Abstract
This study explored the proteomic landscape of inflammatory protein dysregulation in ankylosing spondylitis (AS), a chronic inflammatory disorder primarily affecting the axial skeleton and strongly associated with the HLA-B27 allele, particularly the HLA-B2705 and HLA-B2704 subtypes prevalent in Chinese populations. Blood samples from HLA-B27-positive AS patients and normal controls (NC) were analyzed using the Olink Target 96 inflammation panel to profile 92 inflammatory proteins. HLA-B27 subtyping was performed via PCR-SSP. To identify key proteins and stratify AS subtypes, we employed machine learning classifiers, including LightGBM models coupled with SHAP value interpretation, alongside traditional statistical analyses. The proteomic analysis revealed significant dysregulation of pro-inflammatory cytokines, such as IL-6 and IL-17A, in AS patients compared to NC, with CXCL9 and NRTN identified as potential biomarkers associated with disease activity. The combination of LightGBM classifiers and traditional statistical methods demonstrated high accuracy in distinguishing AS from NC and effectively stratifying subtypes. These findings provide valuable insights into the inflammatory mechanisms underlying AS pathogenesis and highlight potential biomarkers and therapeutic targets for improving diagnosis and treatment strategies. Future studies with larger and more diverse cohorts, as well as longitudinal designs, are warranted to validate these biomarkers and elucidate their dynamic changes during disease progression.