Identification of the molecular subtypes and construction of risk models in neuroblastoma

The heterogeneity of neuroblastoma directly affects the prognosis of patients. Individualization of patient treatment to improve prognosis is a clinical challenge at this stage and the aim of this study is to characterize different patient populations. To achieve this, immune-related cell cycle genes, identified in the GSE45547 dataset using WGCNA, were used to classify cases from multiple datasets (GSE45547, GSE49710, GSE73517, GES120559, E-MTAB-8248, and TARGET) into subgroups by consensus clustering. ESTIMATES, CIBERSORT and ssGSEA were used to assess the immune status of the patients. And a 7-gene risk model was constructed based on differentially expressed genes between subtypes using randomForestSRC and LASSO. Enrichment analysis was used to demonstrate the biological characteristics between different groups. Key genes were screened using randomForest to construct neural network and validated. Finally, drug sensitivity was assessed in the GSCA and CellMiner databases. We classified the 1811 patients into two subtypes based on immune-related cell cycle genes. The two subtypes (Cluster1 and Cluster2) exhibited distinct clinical features, immune levels, chromosomal instability and prognosis. The same significant differences were demonstrated between the high-risk and low-risk groups. Through our analysis, we identified neuroblastoma subtypes with unique characteristics and established risk models which will improve our understanding of neuroblastoma heterogeneity.