Abstract
In this study, we conducted a predictive meta-analysis of multiple microarrays to identify a gene signature that can be potentially used to distinguish different grades of malignant gliomas (i.e., Grade III and IV). We showed our developed classification rule achieved an average accuracy and a J index of 75% and 52.6%, respectively, as measured by a dataset cross-validation strategy. Furthermore, we showed that clustering samples on the basis of similarity of expression profiles of the gene signature divides the data across available studies mainly into the two phenotypic groups regardless of the actual study. From the standpoint of data analytics, the results of this study confirm the utility of meta-analysis in integrating raw data from multiple studies into a predictive framework. From a biological perspective, the identified gene signature can be potentially used to shed light on the molecular mechanisms underlying the formation of malignant gliomas.