Frontiers in Genetics (Sep 2022)
Machine learning identification of cuproptosis and necroptosis-associated molecular subtypes to aid in prognosis assessment and immunotherapy response prediction in low-grade glioma
Abstract
Both cuproptosis and necroptosis are typical cell death processes that serve essential regulatory roles in the onset and progression of malignancies, including low-grade glioma (LGG). Nonetheless, there remains a paucity of research on cuproptosis and necroptosis-related gene (CNRG) prognostic signature in patients with LGG. We acquired patient data from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) and captured CNRGs from the well-recognized literature. Firstly, we comprehensively summarized the pan-cancer landscape of CNRGs from the perspective of expression traits, prognostic values, mutation profiles, and pathway regulation. Then, we devised a technique for predicting the clinical efficacy of immunotherapy for LGG patients. Non-negative matrix factorization (NMF) defined by CNRGs with prognostic values was performed to generate molecular subtypes (i.e., C1 and C2). C1 subtype is characterized by poor prognosis in terms of disease-specific survival (DSS), progression-free survival (PFS), and overall survival (OS), more patients with G3 and tumour recurrence, high abundance of immunocyte infiltration, high expression of immune checkpoints, and poor response to immunotherapy. LASSO-SVM-random Forest analysis was performed to aid in developing a novel and robust CNRG-based prognostic signature. LGG patients in the TCGA and GEO databases were categorized into the training and test cohorts, respectively. A five-gene signature, including SQSTM1, ZBP1, PLK1, CFLAR, and FADD, for predicting OS of LGG patients was constructed and its predictive reliability was confirmed in both training and test cohorts. In both the training and the test datasets (cohorts), higher risk scores were linked to a lower OS rate. The time-dependent ROC curve proved that the risk score had outstanding prediction efficiency for LGG patients in the training and test cohorts. Univariate and multivariate Cox regression analyses showed the CNRG-based prognostic signature independently functioned as a risk factor for OS in LGG patients. Furthermore, we developed a highly reliable nomogram to facilitate the clinical practice of the CNRG-based prognostic signature (AUC > 0.9). Collectively, our results gave a promising understanding of cuproptosis and necroptosis in LGG, as well as a tailored prediction tool for prognosis and immunotherapeutic responses in patients.
Keywords
