Background: One of the hallmarks of neuroblastoma is molecular heterogeneity, which leads to spontaneous remission in some patients and aggressive, resistant disease in others. Patients with low or intermediate risk have a high 5-year survival rate (90-95%), but high-risk neuroblastoma patients have a survival rate of 40% despite receiving high intensity therapies. Methods: Genome-wide transcriptome profiling can be used to stratify high-risk neuroblastoma patients and identify opportunities for molecularly targeted therapies. We developed a gene expression analysis for identifying molecular subtypes in large cancer gene expression cohorts (n > 100). Our method does not rely on a reference normal sample and thus can be applied to pediatric cancers where adequate normal samples are lacking. We applied the method to the TARGET NBL gene expression dataset (N = 162). Results: Our analysis identified 6,736 differentially expressed genes. We recapitulated amplification status of the transcription factor MYCN as originally determined in DNA-based assays by TARGET investigators (F1 score = 91%). We also identified two NTRK1 expression subtypes. The first was characterized by over-expression of MYCN, which downregulates expression of NTRK1. Tumors of the second subtype have lower activity for transcription factors that activate NTRK1 expression. Samples with the second subtype also have worse event-free survival than samples with higher NTRK1 expression (Log-rank test: p = 0.013). Thus, the low NTRK1/non-MYCN amplified subtype may benefit from an alternative treatment strategy. Finally, we used our gene expression approach to characterize the immune infiltrate of neuroblastoma patients and identified a subset of samples (~20%) with elevated expression of cytotoxic and regulatory T cell markers, including expression of inhibitory receptors and ligands. Conclusions: Our gene expression subtyping approach can tbe used to stratify high-risk neuroblastoma patients and identify opportunities for immunotherapy. Validation of this approach in preclinical models may lead to novel therapeutic strategies for high-risk neuroblastoma.