Researchers from MD Anderson Cancer Center and Rice University have developed a computational approach for identifying genes responsible for drug resistance in cancer, as well as therapies that could potentially overcome their effects.
The team performed a network-based analysis of gene expression data from breast cancer cell lines that were either sensitive to or resistant to lapatinib — a tyrosine-kinase inhibitor that is used to treat breast and other solid tumors. They found that disrupting glucose metabolism in the cells could be an effective strategy for treating drug-defying tumors.
This approach enabled the team to “look beyond changes in the immediate molecular signaling pathways of breast cancer cells and to consider the wider system of molecular networks within the cell,” Prahlad Ram, a professor at the University of Texas MD Anderson Cancer Center and senior author on a paper describing the technique, said in a statement
An added benefit is that it can reveal potential new indications for existing drugs, which might help oncologists personalize their patients’ treatments, Ram told BioInform.
The method was published in a recent issue of Molecular Systems Biology.
Lapatinib — marketed by GlaxoSmithKline as Tykerb — is used to treat patients with advanced or metastatic breast cancer in cases where the tumors overexpress the ErbB2 gene. This gene provides instructions for making a specific growth factor receptor that in excess can cause cells that grow and divide continuously.
In the MSB article, the scientists explain that they used microarrays to measure gene expression in breast cancer cells with and without treatment with lapatinib. They team then used the NetWalk algorithm — a random walk-based network scoring method that is part of an application dubbed NetWalker — to analyze the data. In total, they analyzed 240,000 physical and functional interactions among more then 15,000 genes to find specific network alterations that contribute to drug resistance.
A separate BMC Genomics paper that describes NetWalker explains that it provides "unique analysis capabilities to assess entire data distributions together with network connectivity to prioritize molecular and functional networks, respectively.”
The MSB paper explains that NetWalk provides “a distribution of network-wide scores for each interaction in the network based on the local connectivity as well as the supplied gene expression values.”
This lets users compare gene expression data "between different conditions at a network, rather than at a gene level,” the paper explains.
Using NetWalk, the team was able to identify clusters of genes that were regulated in significant ways in the resistant cells. Closer analysis revealed “increased expression of the glucose deprivation response network" in breast cancer cells that had acquired resistance to lapatinib.
Furthermore, when the researchers compared gene expression data from ErbB2-positive breast cancer patients to the survival rates of breast cancer patients, they discovered that, similarly, increased expression of the glucose deprivation network in these cells was linked to low survival rates among patients, the paper explains.
As a next step, the researchers mined the Broad Institute’s Connectivity Map (CMAP) database — which is comprised of genome-wide transcriptional expression data from human cells treated with small molecules and simple pattern-matching algorithms — to find drugs that could target the glucose response network.
Although the researchers focused on breast tumors in this case study, the method could be applied to other cancer types, Ram said.
He told BioInform that his team has begun working with researchers at MD Anderson to run similar analysis for ovarian and melanoma cancer cases.
“The NetWalker algorithm is definitely a step in [the] right direction,” Purvesh Khatri, a research associate at Stanford University, said in an e-mail.
Khatri, who did not participate in the MD Anderson/Rice study, told BioInform that NetWalker addresses one of the challenges mentioned in a review he and colleagues published earlier this year in PLoS Computational Biology that looked at pathway analysis challenges and approaches — “that of considering cross talk between different pathways instead of studying each pathway in isolation.”
He pointed out that the software only uses about half of the genome, “which means a lot of interactions and biology are missing.” However, he stressed that “this is not a criticism for NetWalker” because “only about half of the genome has high quality annotations. We still don't know much about [the other] half of the genome.”