Computational analysis of tumor ecosystems and their regulation and association with outcomes

Project Summary The cellular makeup of tumors can radically influence response to treatment, and survival outcomes. Biomarkers derived from tumor biopsies have had modest success in their clinical utility for prognosis or guiding treatment decisions, being confounded by factors such as cellular composition of tissues Moreover, different biomarkers may be needed in female vs male patients. In prior work we showed how meta-analysis of large clinically annotated public cancer datasets with clinical annotations can robustly identify specific genes and processes associated with survival for patients in both pan-cancer and cancer-specific ways. Here we still systematically investigate cancer-specific prognostic cell types through integration of single cell RNA-seq (scRNAseq) with bulk RNA-seq and methylation data. We will validate selected findings in tissue microarrays. First, we will identify cancer-specific cell transcriptional states and ecosystems associated with survival and treatment response, extending prior work that identified 10 different “ecotypes” of co-occurring cell states across carcinomas. Second, we will extend our framework to isolate cancer-specific cell-type-specific methylation profiles and their correlation with imputed gene expression across populations using paired bulk RNA-seq and methylation from TCGA. Third, we will validate survival associations of cancer- specific cell states by staining human tissue microarrays. We will focus on high grade serous ovarian cancer (HGSOC), which has dire prognosis, and non small-cell lung cancer (NSCLC) for which we have extensive information on immunotherapy response. We will use CODEX imaging on large tissue sections to assess the spatial organization of outcome-related cell states in NSCLC and HGSOC. Overall, we will comprehensively map cancer-specific cell states and ecotypes across malignancies, identifying potential biomarkers and possible new therapeutic targets.