Integration of machine learning to develop a disulfidptosis model for predicting glioma prognosis, immunotherapy response, and drug

Glioma prognosis is challenged by tumor heterogeneity and lack of biomarkers. Disulfidptosis, a novel cell death mechanism induced by disulfide stress, remains poorly understood in gliomas. This study analyzed eight glioma cohorts, identifying two disulfidptosis patterns with distinct genomic alterations, immune microenvironments, and clinical outcomes. A prognostic model—DisulfidpScore—was developed using machine learning, demonstrating robust predictive ability for survival. Crucially, single-cell profiling and virtual knockout analysis revealed elevated disulfidptosis in glioblastoma astrocytes and identified IQGAP1 as a key driver that modulates gene networks governing the cell cycle and neuron-glia interactions. High DisulfidpScore scores correlated with immunosuppressive microenvironments and poorer prognosis but increased chemotherapy sensitivity, whereas low scores indicated better survival and immunotherapy response. The model supports prognostic stratification and personalized treatment for glioma patients.