CSIG-13. COMPUTATIONAL MODELING OF GLIOBLASTOMA STEM CELL SIGNALING NETWORKS. (5th November 2018)

• Record Type: Journal Article
• Authors: Holtzapple, Emilee; Miskov-Zivanov, Natasa; Jahan, Kenneth; Zhang, Yahan; Young, Steven; Cochran, Brent
• Is Part Of: Neuro-oncology. Volume 20(2018)Supplement 6
• Page Start: vi45
• Page End: vi45
• Publication Date: 2018-11-05
• Subjects: Brain Neoplasms -- Periodicals; Brain -- Tumors -- Periodicals; Brain -- Cancer -- Periodicals; Nervous system -- Cancer -- Periodicals; 616.99481
• DOI: 10.1093/neuonc/noy148.179
• Languages: English
• ISSNs: 1522-8517

Abstract

Abstract: Despite a tremendous increase in knowledge about glioblastoma in recent years, it has proven difficult to devise new effective therapies. It is likely that a major reason for the failure of new therapies is due to the molecular heterogeneity of GBM between tumors. We have found from RNAi screens that there is significant diversity in essential genes between the tumor stem cells of different patients with only about 50% of all inhibitory kinases being in common between any 2 stem cell lines. Thus, it is likely that a personalized therapeutic approach will be needed for effective treatment of brain tumors. We are building causal computer models of signaling pathways and networks in these tumor stem cells in order to predict the drug responsiveness of individual GBM stem cell lines. To do this, we are using a framework that assembles and tests element rule-based models in an automated manner, assuming a discrete modeling approach, capable of capturing causal relationships between model elements. The use of causal relationships (positive and negative regulation), in addition to mechanistic relationships (e.g., phosphorylation, binding), allows for capturing not only direct but also indirect interactions between elements. By including these indirect interactions on pathways, when there is no information available about exact mechanisms, we are able to account for known element relationships, and capture larger network motifs (e.g., intertwined feedback and feedforward loops), which are often critical in network response to interventions. Our current model with 141 elements and 298 edges can successfully model responses to CDK6 and GSK3-beta inhibition after initialization with RNA-Seq data. With our automated framework, we are now extending the model using text mining with human supervision, and we will test the extended model against chemical inhibitor and RNAi data from multiple GBM stem cell lines.