Abstract
BACKGROUND: Gliomas traditionally have been sub-classified based on histopathological observations. However, this approach is subject to inter-observer variability, and histopathological features may not reflect the biological mechanisms that drive tumor growth. High-throughput transcriptional profiling has shown promise in objectively and reproducibly identifying glioma subtypes. Most prior studies have typically used only modest sample sizes and have sometimes overlooked important data-processing steps to ensure sample quality and to evaluate the robustness of quantitative findings. The purpose of our study was to define robust glioma subtypes by applying rigorous preprocessing and validation steps to 1,952 microarray samples aggregated from 16 prior studies. This data set is the most comprehensive collection of glioma microarray samples compiled to date. METHODS AND RESULTS: We evaluated each sample for quality-control issues, corrected for probe-composition biases, and adjusted for intra- and inter-study batch effects. Using a training/testing validation design that simulates a "bench-to-bedside process," we identified six transcriptional subtypes that contained a heterogeneous mix of histopathological subtypes and tumor grades. Similar to prior studies, age, survival and treatment patterns differed significantly across the transcriptional subtypes. However, due to our large sample size, we also observed that within a given histopathological subtype, our transcriptional subtypes provided additional prognostic value. Lastly, we used a pathway-based approach to elucidate the biological mechanisms associated with each subtype. CONCLUSIONS: Our findings provide clinical and biological insights that may not be apparent with alternative approaches or smaller data sets, and our approach serves as an example for meta-analyses that can be applied to other complex diseases.