Related Experiment Video
Updated: Mar 2, 2026
Single-cell Analysis of Immunophenotype and Cytokine Production in Peripheral Whole Blood via Mass Cytometry
Published on: June 26, 2018
Characterization of immune features and discovery of potential biomarkers for ankylosing spondylitis using deep
Xiaohan Xu1, Wanlin Liu2, Bo Pang3
1Department of Rheumatology, Guang'anmen Hospital, China Academy of Chinese Medical Sciences, Beijing 100053, China.
Introduction:
Ankylosing spondylitis (AS) is a systemic inflammatory disorder that predominantly involves the axial skeleton, often leading to irreversible structural damage and disability. Although several therapeutic measurements are available, limitations in efficacy and long-term outcomes remain significant. Therefore, identifying novel biomarkers and therapeutic targets is of critical importance for optimizing clinical management and prognostic evaluation in AS patients.
Objectives:
This study aims to elucidate the immune features and discover potential biomarkers for AS by the integration of deep plasma proteomics and deep learning strategies.
Methods:
The deep quantitative proteomics was applied to analyze the plasma samples from 104 participants of AS patients with active and stable stages, along with healthy controls. The immune and functional features of AS patients in different stages were assessed. By integrating random forest (RF) with orthogonal partial least squares discriminant analysis (OPLS-DA), a machine learning model-based score matrix was constructed to identify biomarkers. ELISA experiments were performed on an independent cohort of 79 participants to confirm the potential biomarkers for AS.
Results:
Patients with AS exhibit significant dysregulation in the distributions and characteristics of immune cells. Several key proteins involved in integrin signaling pathway were significantly differentially expressed in patients with AS, highlighting the pathway's role in the pathogenesis of AS. Four proteins including SAA1, FERMT3, ILK, and TLN1, were identified as potential biomarkers for AS and further verified by ELISA experiments.
Conclusions:
By integrating the machine learning-based method with deep proteomics analysis, we explored the pathological mechanism and identified biomarkers for AS. Our study provides insights into the distinct protein expression patterns and pathogenesis of AS and may contribute to diagnosis, long-term monitoring, and therapy for this disease.
