Machine learning-based construction of Immunogenic cell death-related score for improving prognosis and personalized treatment in glioma.

Immunogenic cell death (ICD) is capable of activating both innate and adaptive immune responses. In this study, we aimed to develop an ICD-related signature in glioma patients and facilitate the assessment of their prognosis and drug sensitivity. Consensus clustering and non-negative matrix factorization (NMF) were performed to classify patients into subgroups. A least absolute shrinkage and selection operator (LASSO) logistic regression model was constructed to establish an ICD-related risk score (ICDS). CIBERSORT and ESTIMATE algorithms were employed to evaluate the infiltration of immune cells. Flow cytometry, CCK-8, EdU, and Transwell assays were used to detect cell proliferation and migration abilities. qPCR, Western blotting, immunohistochemistry and immunofluorescence were utilized to detect mRNA and protein expression levels. The ICDS proved effective in predicting the prognosis of glioma patients in both the training and two validating cohorts. The ICDS exhibited significant advantages when compared to the 71 previously published signatures. Patients with a high ICDS score demonstrated marked enhancement in immune cell infiltration and expression of immune checkpoint inhibitor-related genes. Furthermore, SERPINH1, one of the 14 key genes used to establish the ICDS, was abnormally overexpressed in gliomas and activate JAK/STAT signaling, thereby promoting glioma cell proliferation and migration. We developed an ICDS marker to evaluate the prognosis and drug response of glioma patients, and confirmed that SERPINH1 promotes the malignant phenotype of gliomas by modulating the JAK/STAT signaling pathway.