RNA-Seq reveals an unprecedented complexity of the neuroblastoma transcriptome and is suitable for clinical endpoint prediction [microarray]

We generated gene expression profiles from 498 primary neuroblastomas using RNA-Seq and microarrays. We sought to systematically evaluate the capability of RNA deep-sequencing (RNA-Seq)-based classification for clinical endpoint prediction in comparison to microarray-based ones. The neuroblastoma cohort was randomly divided into training and validation sets, and 360 predictive models on six clinical endpoints were generated and evaluated. While prediction performances did not differ considerably between the two technical platforms, the RNA-Seq data processing pipelines, or feature levels (i.e., gene, transcript, and exon junction levels), RNA-Seq models based on the AceView database performed best on most endpoints. Collectively, our study reveals an unprecedented complexity of the neuroblastoma transcriptome, and provides guidelines for the development of gene expression-based predictive classifiers using high-throughput technologies. Sample clinical characteristics definitions: dataset: Expression data set used for classification training (1) or validation (2) Sex: M = male; F= female age at diagnosis: the age in days at diagnosis mycn status: Amplification status of the MYCN proto-oncogene (amplified = 1, no amplification = 0; no information = N/A) high risk: Clinically considered as high-risk neuroblastoma (yes=1, no= 0) INSS stage: disease stage according to International Neuroblastoma Staging System (INSS) (1, 2, 3, 4 and 4S) class label: Maximally divergent disease courses - unfavorable (= 1): patient died despite intensive chemotherapy, favorable (=0): patient survived without chemotharapy for at least 1000 days post diagnosis; not applicable (N/A) progression: Occurrence of a tumor progression event (yes=1; no=0) death from disease: Occurrence of death from the disease (yes=1; no=0)