While dysregulation of multiple genes plays a role in many complex diseases, clinicians are still relying on a single gene or protein to make the call when it comes to diagnosis. Trey Ideker's lab at the University of California, San Diego, is working out this problem by using bioinformatic analysis and the mountains of gene and protein expression data available. Moving from a single-gene to a "network-based approach to biomarker discovery [and] disease prediction," he says, is an emerging field that holds much promise as clinicians looks for more effective ways to prevent the spread of diseases like cancer by detecting them earlier.
In a collaboration with the UCSD cancer clinic, Ideker is one of a handful of scientists working on refining biomarker classification to a pathway-based approach. "Biomarkers are typically thought of as individual molecules, like genes and proteins, and so clinicians are constantly arguing over what's the right set of biomarkers that can be used to diagnose a disease or predict an outcome," Ideker says. "The main advance here is we're looking at biomarkers as whole pathways, not as individual genes and proteins."
To work on breast cancer, he teamed up with scientists at the Korean Advanced Institute of Science and Technology in Daejeon, South Korea, to analyze gene expression profiles from female breast cancer cohorts to come up with a set of genes that could be classified into specific subnetworks. The total gene expression pattern of these networks could more precisely distinguish patients whose tumors had metastasized from those that had not.
He's doing the same kind of work on chronic lymphocytic leukemia in collaboration with Thomas Kipps at the UCSD cancer clinic. "The problem in CLL is predicting the outcome," he says. "The key isn't diagnosing CLL versus normal, it's in being able to predict when this person's disease is going to become severe." And while biomarkers do exist — most people are diagnosed without symptoms as the result of a routine blood test that comes back with a high white blood cell count — they're often only somewhat informative, Ideker says.
To that end, he's moving from the current methods to a pathway biomarker approach. Essentially, his pathway signature is a defined subset of genes that form a gene expression network. Ideker assesses the expression of genes in a pathway to come up with a pathway activity value. "One can think of a pathway, in this case, as essentially a set of genes, but instead of treating the genes individually, like is done now, you group the genes together and take something very close to their average," he says. "Whether or not a pathway is active is really what's going to set the state of your disease. And you need to somehow divine from the individual genes and proteins in that pathway whether or not it's active."
Ideker has filed a patent on the set of genes he's using for CLL, and in the future he hopes to improve upon the bioinformatic analysis of the test. He's considering different functions that would more effectively combine individual biomarkers as well as thinking about ways of crunching together more than one pathway, or a "set of [a] set of genes."
While promising, pathway activity signatures will need a few years before making a real mark in the clinic. "We still have to make it work," Ideker says. "At the moment, pathway-based biomarkers do a little bit better than single-gene biomarkers. We'd like to widen that gap, basically."