With expression data gleaned from the Stanford Microarray Database and the Whitehead Institute Center for Genomic Research, researchers from Harvard, Stanford, and Hebrew Universities developed a “cancer module map” revealing gene expression patterns that may aid early detection, drug target identification, and cancer cell biology.
The data pulls together microarray information from more than 1,975 published microarrays spanning 22 tumor types and subtypes, said lead author Eran Segal. The group looked for similar “profile signatures” — sets of genes that seem to act together and how they behave depending on the kind of tumor, said Segal. Certain blocks of genes are activated and some are deactivated in specific types of cancer. “Then you can also ask what is shared between different types of tumors even,” he added.
The map, published in the October issue of Nature Genetics, is publicly available at http://robotics.stanford.edu/~erans/cancer/. Nir Friedman of Hebrew University said GeneXPress (http://genexpress.stanford.edu/), a publicly available visualization and analysis tool the group developed, was pivotal in the construction of the map. The tool allows researchers to “construct their own module maps, replace data or modules,” and otherwise modify gene expression data to analyze more than cancer, he added. “You can use it to study any phenomenon for any organism [having] publicly available data,” said Daphne Koller, a Stanford collaborator.
The module map also shows that certain modules are specific to particular tumor types, allowing drug researchers to relate the two, giving them an idea of which genes to target, said Segal. A drug target related to one gene in a module might suggest other targets within the same group, Friedman said.
Another advantage is the possibility of discovering previously unnoticed gene-expression correlations between tumor types. One bone-building module was active in many primary tumors, including primary breast cancers, he said. Since breast tumors are known to metastasize to bone, since not much was known about activation of this module in primary tumors, Segal said. “That might suggest a hypothesis that somehow it is to the advantage of the primary tumor to activate this bone-related module,” he added.
Knowing the activation characteristics of the gene sets across different tumors may help in identification as well. “There are a lot of predictions we studied in depth — it might be that by profiling individual breast tumors, you might be able to predict their ability to metastasize to bone,” Segal said.
The group has not been contacted by any drug companies, but they are working on devising computational experiments to provide additional targets for groups studying specific tumors, Segal said.