LA JOLLA, Calif.—Perhaps the most significant challenge in finding prognostic markers for human diseases is the fact that most diseases don't involve individual genes but rather complex networks of interacting proteins and other biomolecules. With this in mind, researchers at UCSD and the Korea Advanced Institute of Science and Technology recently correlated breast cancer gene expression patterns with protein interaction databases to identify metabolic networks of cancer metastasis.
As they describe in Molecular Systems Biology, the researchers took microarray data from two cohorts of breast cancer patients, dividing the expression profiles into "metastatic" and "non-metastatic" groups. They then overlaid this data onto a pooled data set of almost 60,000 interactions of more than 11,000 proteins derived from yeast two-hybrid experiments, predicted from orthology, or curated in the literature. In total, they identify almost 400 subnetworks, which they have made available online as the Cell Circuits database.
The researchers found there was significantly more overlap between the two data sets at the subnetworks level than the single genes level and were relatively predictive of metastatic prognosis both within and across breast cancer cohorts. Furthermore, using the subnetwork system, the researchers could identify well established prognostic markers that are not identified in conventional gene expression screens, including HER2/neu, myc, and cyclin D1.
Concluded the researchers: "At present, the success of network-based pathways identification and classification support the notion that cancer is indeed a 'disease of pathways' and that the keys for understanding at least some of these pathways are encoded in the protein network."