Identification of novel markers for neuroblastoma immunoclustering using machine learning.

Background: Due to the unique heterogeneity of neuroblastoma, its treatment and prognosis are closely related to the biological behavior of the tumor. However, the effect of the tumor immune microenvironment on neuroblastoma needs to be investigated, and there is a lack of biomarkers to reflect the condition of the tumor immune microenvironment. Methods: The GEO Database was used to download transcriptome data (both training dataset and test dataset) on neuroblastoma. Immunity scores were calculated for each sample using ssGSEA, and hierarchical clustering was used to categorize the samples into high and low immunity groups. Subsequently, the differences in clinicopathological characteristics and treatment between the different groups were examined. Three machine learning algorithms (LASSO, SVM-RFE, and Random Forest) were used to screen biomarkers and synthesize their function in neuroblastoma. Results: In the training set, there were 362 samples in the immunity_L group and 136 samples in the immunity_H group, with differences in age, MYCN status, etc. Additionally, the tumor microenvironment can also affect the therapeutic response of neuroblastoma. Six characteristic genes (BATF, CXCR3, GIMAP5, GPR18, ISG20, and IGHM) were identified by machine learning, and these genes are associated with multiple immune-related pathways and immune cells in neuroblastoma. Conclusions: BATF, CXCR3, GIMAP5, GPR18, ISG20, and IGHM may serve as biomarkers that reflect the conditions of the immune microenvironment of neuroblastoma and hold promise in guiding neuroblastoma treatment.