Resistance to chemotherapeutics and metastasis are major hurdles in the development of drugs to eradicate cancers. Among these, drug resistance poses the most urgent challenge to solve. This is because whatever drugs we design to halt cancer progression, regardless of disease stage and mechanisms driving tumorigenesis, treatments always result in development of resistance. The more we understand the mechanisms leading to drug resistance at the systems level, the more challenging the problem becomes because so many molecular pathways are involved. However, the observation that a cancer's resistance to a given drug renders it sensitive to another drug poses the possibility that drug resistance can be overcome. Specifically, if we can reliably reverse engineer drug response networks for each drug and a corresponding to a given cancer state, and then dissect candidate genes in these networks for further experimental validation using both cancer cell lines and patient-derived tumor samples, we will provide a strong translational medicine platform. To achieve this goal, we will adjust the core engine from CellNet (a published platform I helped develop) to generate reverse engineered networks for 10 drugs, including inhibitors of tyrosine and serine- threonine kinases as well as cell cycle regulators, and cell toxicity in breas and ovarian cancers. We will work to reverse engineer networks corresponding to existing molecular settings that confer both sensitivity and resistance to drugs instead of post-drug treatment responses. Thus, basal transcriptome data (prior to drug treatment) from breast and ovarian cancer cell lines with known drug response phenotypes will be used. To the best of our knowledge, there is no such reverse engineered network resource available. To further characterize the reverse engineered networks, we will devise two additional novel network tools: (i) P-MAP (phenotype mapping) to dissect candidate genes that play pivotal roles in conferring drug response phenotype, i.e., where modulating candidate genes potentially allow us to reverse a drug response via engineered phenotype, from resistant to sensitive or vice versa; (ii) DDNA (differential dynamic network analysis) to characterize network connections that capture pathways crucial in shaping drug response phenotypes and informing prognosis. We will test, validate, and refine our computational predictions using breast and ovarian cancer cell lines. To integrate tumor cell heterogeneity and micro-environmental factors not captured in cell lines, we will examine ovarian cancer implants harvested from multiple sites in individual patients. To translate these computational models for clinical applications, we propose to apply our refined computational models using high-grade serous ovarian cancer samples (80 patients) with widely differing outcomes. The proposed work will transform our current understanding on cancer drug resistant and help devise strategies to overcome the problem. The proposed network platform together with its network resources and software will be made publicly available for users to upload their transcriptome data and obtain potential drug and gene candidates.
While traditional approaches aim to identify individual genetic mutations that cause drug resistance in cancers and develop drugs to combat these mutations, we seek to use novel computational tools to identify the cellular context - or big-picture view of the network of genes and molecular pathways - that renders a tumor resistant to one drug but sensitive to another. We will vigorously test, validate, and refine our computer-generated models by evaluating affected networks, first in cancer cell lines and then, using two clinically well characterized high grade serous ovarian cancer patient cohorts, where patient outcomes are available. The proposed network platform may facilitate the repurposing of existing approved drugs to treat ovarian cancers, and it has the potential to be used in the development of broader cancer treatment plans.
