Molecular classification of human endometrial cancer based on gene expression profiles from specialized microarrays

Abstract

Objective To investigate whether the molecular classification of endometrial cancer based on gene expression profiles can predict the biological behavior of the tumors and inform prognosis. Methods An array containing 492 genes was used to generate gene expression profiles from 35 tumor samples. A hierarchical cluster algorithm was used to compare gene expression patterns among the tumor samples. Results A cluster analysis revealed 3 distinct tumor clusters. A comparative analysis of tumor type, grade, FIGO stage, and depth of myometrial invasion revealed significant differences in grade and stage among the clusters, which appear to group tumors with specific clinical behaviors. Moreover, the cluster analysis initially revealed 2 clusters of differentially expressed genes. One contained 38 genes that were upregulated in most samples of the cluster representing the most advanced disease, and the other contained 27 genes that were upregulated in most samples of the cluster representing the least advanced disease. Conclusion Molecular classification of endometrial cancer based on gene expression profiles obtained by designing specialized microarrays indicated a marked correspondence with the histologic features and clinical behavior of endometrial cancer tumors.