Comprehensive machine learning analysis of PANoptosis signatures in multiple myeloma identifies prognostic and...

PANoptosis is closely associated with tumorigenesis and therapeutic response, yet its role in multiple myeloma (MM) remains unclear. This study analyzed bulk transcriptomic and clinical data from the TCGA and GEO databases to identify seven PANoptosis-related genes (PRGs) using machine learning (LASSO regression and random forest models) and univariate Cox analysis, and constructed a prognostic risk model. The model demonstrated robust predictive performance across three external validation cohorts. High-risk patients exhibited higher tumor purity, increased tumor mutational burden, and distinct immune cell infiltration patterns. Drug sensitivity analysis revealed heightened sensitivity to cyclophosphamide, Sinularin, Wee1 inhibitor, osimertinib, JQ1, VE-822, and AZD6738 in high-risk patients. Single-cell transcriptomic analysis revealed significant enrichment of PARP1, ZBP1, LY96, and CASP3 in plasma cells. Quantitative PCR (qPCR) further validated differential expression patterns of the seven core PRGs between MM patients and healthy controls. Immunohistochemical analysis demonstrated distinct expression profiles of PARP1, ZBP1, LY96, and CASP3 in high-risk versus standard-risk MM patients. Furthermore, CCK-8 assays and Wright-Giemsa staining confirmed the crucial role of PARP1 in regulating MM cell viability. This PANoptosis-based prognostic model provides a valuable tool for predicting MM prognosis and guiding personalized treatment.