EPCO-14. QPC-GBM DECONVOLUTION REVEALS OPPOSING PROGNOSTIC ROLES OF GLIOMA-ASSOCIATED MYELOID AND ENDOTHELIAL CELLS IN GLIOBLASTOMA

Abstract Glioblastoma (GBM) is a highly malignant brain tumor that becomes resistance to current therapy partially due to interactions between tumor cells and tumor microenvironment (TME). To further understand these interactions and its patient prognosis association, we constructed a bulk RNA deconvolution method, the quantifying proportion of cells for GBM (QPC-GBM), to estimate cell compositions in GBM RNA samples. The QPC-GBM is a new combined methods with optimal results of each cell type from different deconvolution combinations of reference matrix and deconvolution approach with specific cell types and reference matrices from GBM single cell RNA seq datasets. By analyzing TCGA-GBM and CGGA-GBM datasets, we found the abundance of tumor and endothelial cells were positively correlated to each other, whereas their abundances were negatively correlated to GAMs and mural cells. Tumor and endothelial cells decreased while the oligodendrocytes increased in recurrent tumors. Endothelial cell abundance positively correlated to better prognosis whereas GAMs had opposite effect in newly diagnosed GBM. GAMs specifically enriched in TLR/LDL-related pathways were negatively correlated to survival and may serve as novel drug targets. This study provides a new method for studying GBM TME and links GBM TME cells to patient survival that drives new thoughts to future GBM therapy.