Integrative analysis reveals shared biomarkers and pathways in osteoporosis and rheumatoid arthritis using machine learning and single-cell sequencing
Jin Zhao, Yixuan Chen, Fei Yang, Cheng Tian, Wenyan Luo, Zhou Su, Bohuan Yang, Wei Liu, Hongbo YangBackground
Osteoporosis (OP) and Rheumatoid Arthritis (RA) are prevalent chronic diseases characterized by bone loss and immune dysregulation. Evidence suggests that patients with RA are at increased risk of developing OP; however, the shared molecular mechanisms linking these conditions remain incompletely understood.
Methods
Publicly available bulk transcriptomic and single-cell RNA sequencing (scRNA-seq) datasets for OP and RA were obtained from the Gene Expression Omnibus (GEO) database. Weighted gene co-expression network analysis (WGCNA) was applied to identify shared gene modules associated with both diseases. Functional enrichment and protein–protein interaction analyses were performed to characterize common biological pathways. Machine learning algorithms, including Random Forest (RF) and Support Vector Machine (SVM), were used to identify and validate shared hub genes and to construct diagnostic models. Immune cell infiltration was evaluated using CIBERSORT, and the cellular localization of hub genes was further validated using scRNA-seq analysis.
Results
WGCNA identified 329 genes shared between OP- and RA-associated modules, which were significantly enriched in muscle-related processes, cytokine signaling, and immune–inflammatory pathways. RF analysis identified three shared candidate genes: ANO5, IL27RA, and CNTN3. Among these, ANO5 and IL27RA demonstrated consistent expression evidence and literature support, whereas CNTN3 emerged as a computationally nominated candidate requiring cautious interpretation and experimental validation. An SVM-based model incorporating these genes demonstrated robust predictive accuracy in both OP and RA validation cohorts. Immune infiltration analysis revealed distinct yet coordinated immune landscapes, with hub genes showing significant correlations with T cells, macrophages, B cells, and plasma cells. ScRNA-seq analysis revealed that ANO5 was predominantly expressed in macrophages and fibroblasts, IL27RA was mainly enriched in T cells, whereas CNTN3 showed minimal enrichment at the single-cell level in both OP bone and RA synovial tissues.
Conclusion
This integrative bioinformatics and machine learning study identifies shared molecular signatures and immune-related mechanisms underlying OP and RA. ANO5 and IL27RA emerge as robust candidate regulators linking immune dysregulation and bone remodeling, whereas CNTN3—although identified by machine learning—requires cautious interpretation and further investigation due to limited supporting evidence. These findings provide novel insights into the pathogenesis of RA-associated OP and highlight potential biomarkers and therapeutic targets for their comorbidity.