Abstract
BACKGROUND: Osteomyelitis (OM) is a debilitating infectious disease characterized by inflammation of the bone and bone marrow. Emerging evidence suggests that multiple forms of programmed cell death (PCD) contribute to its pathogenesis. However, the specific roles and interactions of these PCD types in OM remain largely undefined. METHODS: Microarray-based transcriptome datasets related to OM were retrieved from the Gene Expression Omnibus (GEO) database. Thirteen PCD modalities were defined from the literature and specialized databases, including classical forms (e.g., apoptosis, autophagy) and non-classical forms (e.g., cuproptosis, entosis, ferroptosis). Gene Set Variation Analysis (GSVA) was used to evaluate pathway activities in OM, and their associations with immune infiltration, inflammation-related gene expression, and diagnostic value were systematically assessed. Weighted gene co-expression network analysis (WGCNA) was performed to identify essential modules and hub genes. A diagnostic model was constructed using machine learning with SHapley Additive exPlanations (SHAP), and candidate genes were validated in clinical peripheral blood samples using polymerase chain reaction (PCR). RESULTS: Eight core PCD pathways were significantly associated with OM, mainly represented by apoptosis, autophagy, and non-classical forms such as cuproptosis and entosis. By integrating WGCNA with SHAP analysis, five hub genes (SORT1, KIF1B, TMEM106B, NPC1, and ATP6V0B) were identified as key diagnostic candidates. qPCR validation confirmed their significantly different expression between OM patients and healthy controls, supporting their utility as diagnostic biomarkers for early detection and treatment stratification. CONCLUSIONS: This study provides a comprehensive landscape of PCD involvement in OM, identifies novel diagnostic biomarkers, and highlights potential therapeutic targets for clinical intervention.