This application is prepared in response to the funding opportunity: Pediatric Preclinical Testing Consortium Research Programs (U01), RFA-CA-14-018. Specifically, we propose for Type B: Research Program for tumors of central nervous system (CNS) in vivo testing. Brain tumors are the leading cause of cancer-related death in children. One of the challenges in clinical drug development is how to effectively prioritize drug candidates to ensure clinical success in cancer patients. As pediatric drug candidates are much more numerous than the small number of patients, it is essential to perform comprehensive pre-clinical testing to identify the investigational agents that are most likely to be effective in the clinic. However, such effort has been blocked for many years due to the lack of clinically relevant and molecularly accurate model systems. Fortunately, we have established a large panel (>70) of patient derived orthotopic xenograft (PDOX) models of pediatric brain tumors through direct injection of patient tumor specimens into the brains of SCID mice. These PDOX models are shown to have replicated the histopathological features, invasive phenotypes and major genetic abnormalities (gene expression, DNA copy number and gene mutations) of the original primary tumors even during serial sub-transplantations in vivo in mouse brains. The xenograft tumor cells can also be cryopreserved for sustained and on-demand supply of tumorigenic PDOX cells. This capacity combined with our optimized surgical procedure, with which we can implant up to 260 mice per day, makes it possible for us to test multiple (e.g., 6-10) drugs per year for every tumor type. Our objective is therefore to make use of this unique panel of PDOX models to examine therapeutic efficacy of new agents and to analyze mechanisms of action and therapy resistance in high grade glioma, medulloblastoma and ependymoma. Our hypothesis is that these patient-specific PDOX tumors will respond to anti-cancer therapies similarly to the corresponding human primary tumors, and the effective agents identified through this system would have better chances of clinical success. To test this hypothesis, we will perform a series of in vitro and in vivo assays to achieve the following aims: 1) to identify genetically accurate candidate PDOX models that bear the therapeutic target(s) of new investigational drugs through data mining of our mouse model molecular characterization databases; 2) to select the most responsive models through functional in vitro screening to determine time- and dose- responses; 3) to demonstrate therapeutic efficacy of new investigational drugs in multiple target-bearing PDOX models; and 4) to perform detailed analysis of cellular and molecular mechanisms of cell killing as well as the causes of therapy resistance both in vitro and in vivo. Our novel panel of PDOX mouse models represents a broad spectrum of genetic abnormalities of pediatric CNS tumors. All the assays are well established and routinely performed in our laboratory; we are uniquely positioned to accomplish the proposed drug studies in vivo. Our findings should provide strong preclinical evidence to support the initiation of clinical trials.