Characterizing programmed cell death features in osteoarthritis through integrative multiomics and machine learning analysis

Background Programmed cell death (PCD) is an essential biological process in maintaining tissue homeostasis and eliminating damaged or unnecessary cells. Signaling molecules profoundly affect cellular metabolism and are crucial in various diseases; however, their role in osteoarthritis (OA) remains unclear. This study aimed to systematically evaluate the predictive value, genetic alterations, and therapeutic implications of PCD-associated genes in OA. Methods We performed multiomics analyses, integrating transcriptomic and single-cell transcriptome data. The biological importance of PCD genes was investigated using differential expression analysis, functional enrichment analysis, pathway analysis, weighted gene co-expression network analysis, and many machine learning models. Additionally, we evaluated diagnostic efficacy, immune infiltration, and competing endogenous RNA networks associated with these genes. We established an in vitro OA model using hypoxic treatment of ATDC5 chondrocyte cells and conducted extensive research on the expression and function of key PCD-related genes. Results The key PCD gene was identified as markedly dysregulated in OA. Elevated expression of S100A9, PMAIP1, and EDA2R was observed in OA samples, indicating these genes as potential risk factors for OA. However, FASN expression was reduced in OA samples compared to the normal group, indicating its potential role as a protective gene in OA. Furthermore, PCD emerged as a reliable diagnostic marker with improved predictive accuracy. Functional experimental studies demonstrated that S100A9, PMAIP1, and EDA2R downregulation through small interfering RNA, alongside FASN gene overexpression through plasmid transfection, significantly ameliorated hypoxia-induced reductions in cell viability, decreased hyaluronan secretion, and increased secretion of inflammatory cytokines (tumor necrosis factor-alpha and interleukin-6). Conclusion Utilizing a multi-model synergistic artificial intelligence framework, we demonstrated the remarkable potential of PCD to provide individualized vulnerability assessments and customized recommendations for metabolic and immunotherapeutic interventions in OA. We identified abnormal expression of four hub genes associated with PCD and examined their biological functions, thereby facilitating new avenues for research into the role of PCD in OA and other immune-mediated diseases.