Molecular classification of human carcinomas by use of gene expression signatures.

Classification of human tumors according to their primary anatomical site of origin is fundamental for the optimal treatment of patients with cancer. Here we describe the use of large-scale RNA profiling and supervised machine learning algorithms to construct a first-generation molecular classification scheme for carcinomas of the prostate, breast, lung, ovary, colorectum, kidney, liver, pancreas, bladder/ureter, and gastroesophagus, which collectively account for approximately 70% of all cancer-related deaths in the United States. The classification scheme was based on identifying gene subsets whose expression typifies each cancer class, and we quantified the extent to which these genes are characteristic of a specific tumor type by accurately and confidently predicting the anatomical site of tumor origin for 90% of 175 carcinomas, including 9 of 12 metastatic lesions. The predictor gene subsets include those whose expression is typical of specific types of normal epithelial differentiation, as well as other genes whose expression is elevated in cancer. This study demonstrates the feasibility of predicting the tissue origin of a carcinoma in the context of multiple cancer classes.

Analysis of gene expression identifies candidate markers and pharmacological targets in prostate cancer.

Detection, treatment, and prediction of outcome for men with prostate cancer increasingly depend on a molecular understanding of tumor development and behavior. We characterized primary prostate cancer by monitoring expression levels of more than 8900 genes in normal and malignant tissues. Patterns of gene expression across tissues revealed a precise distinction between normal and tumor samples, and revealed a striking group of about 400 genes that were overexpressed in tumor tissues. We ranked these genes according to their differential expression in normal and cancer tissues by selecting for highly and specifically overexpressed genes in the majority of cancers with correspondingly low or absent expression in normal tissues. Several such genes were identified that act within a variety of biochemical pathways and encode secreted molecules with diagnostic potential, such as the secreted macrophage inhibitory cytokine, MIC-1. Other genes, such as fatty acid synthase, encode enzymes known as drug targets in other contexts, which suggests new therapeutic approaches.

Analysis of gene expression profiles in normal and neoplastic ovarian tissue samples identifies candidate molecular markers of epithelial ovarian cancer.

Epithelial ovarian cancer is the leading cause of death from gynecologic cancer, in part because of the lack of effective early detection methods. Although alterations of several genes, such as c-erb-B2, c-myc, and p53, have been identified in a significant fraction of ovarian cancers, none of these mutations are diagnostic of malignancy or predictive of tumor behavior over time. Here, we used oligonucleotide microarrays with probe sets complementary to >6,000 human genes to identify genes whose expression correlated with epithelial ovarian cancer. We extended current microarray technology by simultaneously hybridizing ovarian RNA samples in a highly parallel manner to a single glass wafer containing 49 individual oligonucleotide arrays separated by gaskets within a custom-built chamber (termed "array-of-arrays"). Hierarchical clustering of the expression data revealed distinct groups of samples. Normal tissues were readily distinguished from tumor tissues, and tumors could be further subdivided into major groupings that correlated both to histological and clinical observations, as well as cell type-specific gene expression. A metric was devised to identify genes whose expression could be considered ideal for molecular determination of epithelial ovarian malignancies. The list of genes generated by this method was highly enriched for known markers of several epithelial malignancies, including ovarian cancer. This study demonstrates the rapidity with which large amounts of expression data can be generated. The results highlight important molecular features of human ovarian cancer and identify new genes as candidate molecular markers.