Normal cellular behavior can be described as a complex, regulated network of interaction between genes and proteins. Targeted cancer therapies aim to neutralize specific proteins that are necessary for the cancer cell to remain viable in vivo. Ideally, the proteins targeted should be such that their downregulation has a major impact on the survival/fitness of the tumor cells and, at the same time, has a smaller effect on normal cells. It is difficult to use standard analysis methods on gene or protein expression levels to identify these targets because the level thresholds for tumorigenic behavior are different for different genes/proteins. We have developed a novel methodology to identify therapeutic targets by using a new paradigm called "gene centrality." The main idea is that, in addition to being overexpressed, good therapeutic targets should have a high degree of connectivity in the tumor network because one expects that suppression of its expression would affect many other genes. We propose a mathematical quantity called "centrality," which measures the degree of connectivity of genes in a network in which each edge is weighted by the expression level of the target gene. Using our method, we found that several SRC proto-oncogenes LYN, YES1, HCK, FYN, and LCK have high centrality in identifiable subsets of basal-like and HER2+ breast cancers. To experimentally validate the clinical value of this finding, we evaluated the effect of YES1 knockdown in basal-like breast cancer cell lines that overexpress this gene. We found that YES1 downregulation has a significant effect on the survival of these cell lines. Our results identify YES1 as a target for therapeutics in a subset of basal-like breast cancers.
Keywords: breast cancer; eigenvector centrality; oncogene; therapeutic target.