Molecular classification of human diffuse gliomas by multidimensional scaling analysis of gene expression profiles parallels morphology-based classification, correlates with survival, and reveals clinically-relevant novel glioma subsets.

There are several currently employed classification systems for diffuse gliomas that sort tumors based on histological features. Contemporary molecular techniques, however, offer the promise of improved tumor classification and resultant patient stratification for treatment and prognosis. In particular, gene expression profiling has shown exceptional promise for providing an alternative and more objective molecular approach to glioma classification. In this study, we used cDNA array technology to profile the gene expression of 30 primary human glioma tissue samples comprising 4 different glioma subtypes as defined by current World Health Organization (WHO 2000) criteria: glioblastoma (GM, WHO grade IV), anaplastic astrocytoma (AA, WHO grade III), anaplastic oligodendroglioma (AO, WHO grade III), and oligodendroglioma (OL, WHO grade II). Gene expression data alone were used to group the tumors using multidimensional scaling, which is an unsupervised statistical method. Results show that impressive separation of the 4 glioma subtypes can be achieved solely on the basis of molecular data. In addition, a subcluster of 3 glioblastomas was identified as distinct from other GMs and from the oligodendroglial tumors. These 3 patients have shown extended survival compared to other GMs in the study. Survival analysis of the full data set revealed a good correlation with the molecular classification. Results of this proof-of-principle study demonstrate that molecular profiling alone can recapitulate conventional histologic classification and grading with high fidelity. In addition, results show that the molecular approach to tumor classification can generate clinically meaningful patient stratification, and, more importantly, is an efficient class-discovery tool for human gliomas, permitting the identification of previously unrecognized, clinically relevant tumor subsets.