A novel machine learning-driven immunogenic cell death signature for predicting ovarian cancer prognosis
Wang Y., Zhao P., Sun X., Batalini F., Levin G., Soleymani majd H., Chen H., Gao T.
Background: Ovarian cancer (OC) is one of the most lethal malignancies in women, primarily due to the absence of reliable predictive biomarkers and effective therapies. The complex role of immunogenic cell death (ICD) in OC remains poorly understood, despite its critical implications for enhancing immune responses against tumors. We are committed to developing and validating a novel ICD-related gene signature and producing certain guiding value for the clinical treatment of OC. Methods: We employed single-sample gene set enrichment analysis (ssGSEA) and weighted gene coexpression network analysis (WGCNA) on The Cancer Genome Atlas (TCGA)-ovarian carcinoma dataset to identify ICD-associated genes. A combination of 10 different machine learning approaches was used to construct an ICD-related signature (ICDRS), which was then validated across multiple datasets. The model’s predictive power was integrated into a clinical nomogram to predict patient outcomes. Ultimately, we assessed the reaction of various risk subgroups to screen pharmaceuticals designed to address specific risk factors in the context of personalized medicine. Results: We identified 72 prognostic genes related to ICD. An unanimous ICDRS was developed using a 101-combination machine learning computational structure, demonstrating outstanding predictive accuracy for prognosis and clinical use. Patients categorized as low ICDRS varied from those of high ICDRS in terms of biological processes, mutation profiles, and immune cell penetration in the tumor microenvironment. In addition, potential medications that target specific subgroups at risk were identified. Conclusions: The ICDRS presents a significant advancement for prognostication of patients with OC, facilitating refined predictions and the exploration of personalized treatment pathways. Prospective clinical trials are necessary to validate its clinical utility and expand the application of this model to other cancer types.