Scientists Find Cancer's Genetic Core: 67 Genes Universally Activated In Human Cancer

These 67 genes constitute what U-M scientists call cancer’s meta-signature – a core set of essential genes, which somehow triggers the transformation from normal cells to cells that are neoplastic, or growing abnormally. A list of these genes, and a description of a new database used to identify them, will be published June 7 in the online edition of the Proceedings of the National Academy of Sciences (PNAS).

“We used a statistical analysis method called comparative meta-profiling to examine 40 datasets from other investigators showing specific patterns of genetic activity associated with one of 12 types of cancer,” says Daniel Rhodes, the paper’s first author and a student in the U-M Medical School’s Medical Scientist Training Program and Bioinformatics Graduate Program. “The result was a common transcriptional profile with 67 activated genes present in nearly all cancerous tissue samples in the study, regardless of their tissue of origin.”

“This paper is the first to report common patterns of genetic activity across multiple cancer microarray datasets from different investigators,” says Arul M. Chinnaiyan, M.D., Ph.D., an associate professor of pathology and urology in the U-M Medical School, who directed the study. “These genes and their associated proteins are the molecular components that make different types of cancer more similar to each other than to normal tissue. Since these genes appear to be involved in so many types of cancer, they are prime targets for new cancer-fighting drugs.”

Chinnaiyan, Rhodes and colleagues also identified another 69-gene meta-signature showing a common pattern of genetic activity in aggressive, invasive undifferentiated cancers. The pattern is different from the one found in well-differentiated cancers that grow slowly and are easier to control.

To test the validity of the meta-signatures, U-M researchers looked for the same patterns of genetic activity in 12 other gene expression databases, which were not available when the study began. “In seven of nine datasets, including three from new types of cancer, our first meta-signature significantly discriminated between cancer and respective normal tissue,” Rhodes says. “In three of five datasets, the second meta-signature significantly discriminated between samples from high- and low-grade cancers.”

ONCOMINE, the cancer microarray database used in the study, was developed by Chinnaiyan, Rhodes and colleagues in the U-M Medical School, Johns Hopkins University School of Medicine, and the Institute of Bioinformatics in Bangalore, India.

“Finding the genetic interactions that promote or inhibit cancer’s development is like looking for a needle in a haystack,” Chinnaiyan says. “There are millions of possible combinations. Since the development of microarray technology, more than 200 studies of gene expression signatures in human cancers have been published. They include millions of data points from thousands of experiments stored on different technology platforms. Most cancer biologists don’t have the bioinformatics expertise to access the data, so they have been unable to take advantage of it.”

Microarrays are microchips containing thousands DNA probes that are used to profile tissue samples from malignant tumors. Scientists wash tissue samples on the microarray with a solution containing fluorescent probes that glow when bound to a specific gene or gene combination. Tissue samples containing more of a specific gene glow more intensely than samples with little or no expression of the gene. The procedure is repeated by microarray analysis of normal tissue. A computer then analyzes the level of fluorescence to prepare a visual image comparing patterns of genetic activity in malignant and normal tissue.

“When we started the ONCOMINE project, our goal was to bring all the publicly available cancer microarray data together in an integrated, unified format, making it accessible to the entire cancer research community,” Chinnaiyan explains. “The database makes it possible to focus on the activity of one gene in many types of cancer, or look for patterns of gene activity in one type of cancer. It is a great time-saver, because it lets you do the initial experiments in silico without having to generate all the primary data yourself.”

While he believes the results of his study proves the value of the ONCOMINE database, Chinnaiyan emphasizes that it will never replace the need for careful laboratory analyses to validate initial research findings.

“Our challenge now is to incorporate the growing mass of cancer microarray data,” Rhodes says. “If we are to keep up with this rapidly advancing field, it’s crucial for us to streamline our entire process from data collection and normalization to statistical analysis and gene annotation.”

ONCOMINE is available at no charge to university scientists at http://www.oncomine.org. More than 800 investigators from 20 countries are registered users. Researchers from private corporations can purchase licensing rights. A description of the database was published previously in the January/February 2004 issue of Neoplasia.

Pilot funding for ONCOMINE was provided by the U-M Medical School and the Medical School’s Department of Pathology. Research funding was provided by the American Cancer Society, the U-M Comprehensive Cancer Center Support Grant, the U-M Specialized Program of Research in Prostate Cancer, and the U-M Bioinformatics Program.

Additional U-M collaborators in the study include: Jianjun Yu, bioinformatics graduate student; Radhika Varambally, computer systems analyst; Debashis Ghosh, Ph.D., assistant professor of biostatistics in the School of Public Health; and Terrence Barrette, research associate.

###

Citation: Proceedings of the National Academy of Sciences, article #04-01994