After reading Verhaak et al. 2010 in Cancer Cell, and I was impressed by this very good study analyzing data from an important resource for genomics research.
The authors were able to define gene signatures to define 4 subtypes of glioblastoma. The experimental design was pretty straightforward, and the results were quite clear. Most importantly, their predictive model was trained an a relatively large set of 173 patient samples and validated on an even larger set of 260 patient samples (from 5 independent studies).
The study focused mostly on data provided by The Cancer Genome Atlas (TCGA). TCGA is a database that contains various types of genomic data (gene/miRNA expression, gene/miRNA copy number, DNA sequence/polymorphism, and DNA methylation), and most or all types of genomic data are available for each patient in the database. This provides a unique opportunity to integrate many different types of data, usually for a large number of clinical samples. For anyone not aware of this resource, I would strongly recommend checking out the links provides above as well as the original TCGA paper (also on glioblastoma) published in Nature.