A novel molecular classifier enabling identification and prediction on immunotherapeutic response for MYCN-low pediatric neuroblastoma

PURPOSE: Neuroblastoma (NB) is a pediatric malignant solid tumor arising from peripheral neural crest cells, characterized by significant heterogeneity. Previous studies have stratified NB patients into risk groups based on pivotal genetic changes, but the suboptimal clinical outcomes of NB underscore the need for a more precise individualized grading system to guide the selection of novel therapeutic strategies. METHODS: In this study, we developed a dense neural network molecular classifier utilizing bulk transcriptomics data from the UCSC Treehouse database and applied it to a single-center cohort. RESULTS: The neural network molecular classifier on bulk transcriptomics refined the classification in both high-risk and low-risk groups by the traditional classification method. The classifier identified four molecular subtypes: High-risk MYCN-high NB (HR1), High-risk MYCN-low NB (HR2), Low/Intermediate-risk NB (LR1), and Low/Intermediate-risk GNB (LR2). By applying the new classifier, we identified factors such as PIK3R1, GATA2, and EYA1 that may be associated with a low-risk, differentiated ADRN-subtype NB, in addition to the classical adrenergic fate-determining factors PHOX2B, ASCL1, ALK, and GATA3. Additionally, we observed elevated GD2 and CTLA-4 expression in the High-risk MYCN-low NB group, which may serve as a potential clue for the development of personalized immunotherapeutic strategies. CONCLUSION: Our model details the current NB risk stratification at the transcriptomic level with a molecular classifier and offers potentially more personalized immunotherapeutic strategies for High-risk MYCN-low NBs.