Multiomics Identifies Potential Biomarkers in Ankylosing Spondylitis Bone Formation.

Ankylosing spondylitis (AS) is a long-term inflammatory condition characterized by intricate pathogenesis and significant genetic predisposition. Current treatment methods cannot completely halt the progression of the disease. The purpose of this research is to discover possible therapeutic targets for AS by integrating Mendelian Randomization (MR), transcriptomics analysis, and machine learning, providing new options for the clinical treatment of AS. In this study, we initially pinpointed differentially expressed genes (DEGs) linked to AS from the GEO database and acquired cis-eQTL data for these genes from the eQTLGen Consortium. Using MR and summary data-based Mendelian randomization (SMR) analyses, we screened for DEGs with causal relationships to AS. Subsequently, we analyzed the correlation between these causal genes and immune cell expression, constructed a risk prediction model, and identified key feature genes for AS. Next, we conducted phenome-wide association studies (PheWASs) on the identified AS feature genes to predict their potential adverse effects as therapeutic targets. We obtained AS-related therapeutic drugs from the DrugBank database and performed molecular docking analysis with AS feature genes. We used the CAIA collagen-induced AS mouse model; we measured joint swelling and employed microCT, H&E, and Safranin O-Fast Green staining to assess pathological changes in bone tissue. Additionally, we employed Western blot and RT-qPCR to analyze the expression levels of genes associated with bone mineralization and AS feature genes in joint tissues. A total of 1607 DEGs were obtained from the GEO database. After MR analysis and correction, 33 positive DEGs that have a causal relationship with AS were determined. Through the correlation analysis between these genes and the expressions of immune cells, it was found that 28 genes had significant regulatory relationships with 19 kinds of immune cells, with 55 pairs of negative regulatory relationships and 49 pairs of positive regulatory relationships, respectively. Four machine learning model algorithms determined the Top 5 genes (RIOK1, FUCA2, COL9A2, USP16, and TTC16) with the highest importance scores and constructed a nomogram to evaluate the risk probability. The results of the PheWAS showed that the five characteristic genes of AS had harmful or beneficial effects on numerous disease phenotypes of multiple types of diseases. Molecular docking indicated that 14 known AS treatment drugs had potential interactions with related genes. Using RT-qPCR, we evaluated the expression levels of five key genes in the joint tissue of the CAIA collagen-induced AS mouse model. Compared to the normal control group, we found that the levels of and were significantly elevated, while the levels of were significantly reduced. In contrast, the expression of and mRNA showed no significant difference. Our research findings demonstrate that FUCA2, USP16, and TTC16 may serve as biomarkers for AS.