LAUSR

608 Background: Stromal elements in the tumor microenvironment (TME) impact prognosis and response to therapy. Advances in mRNA-seq improved understanding of gene expressivity, but few models exist to model prognosis in association with mRNA expression. Methods: Clinical data and mRNA-seq of 1,715 patients (pts) – pancreatic adenocarcinoma (PAAD), colorectal adenocarcinoma (CRC), hepatocellular carcinoma (HCC), gastric adenocarcinoma (GAAD), esophageal adenocarcinoma (EsoAd), and esophageal squamous cell carcinoma (EsoSCC) – were obtained from TCGA. The expressivity of 191 genes enriched in cellular and structural components of the TME and clinical data were analyzed using machine learning, multivariable COX model, and Kaplan-Meier (KM) analysis to model risk score (RS) to predict prognosis. Results: Genes associated with good and poor prognosis were identified via machine learning and statistic methods. Higher RS represents worse prognosis with max RS = 1 (Table). In all 6 cancers, high P/G (the expression ratio of genes associated with poor to good prognosis) and old age are related to worse survival except EsoAd with younger pts having worse prognosis. The location of tumors in CRC and sex in HCC impact RS. When pts are grouped into 3 pt groups in each cancer, KM curves in pts with low, intermediate, and high RS are statistically different (p < 0.0001) with high hazard ratio (HR > 2). Conclusions: Analysis of large data was assisted by machine learning and statistics, identifying genes associated with survival and creating RS as a tool to predict prognosis. This provides valuable information about prognosis for pts encountered in the clinic when genomic profiles are given. Computational modeling to predict response to chemotherapy and immunotherapy is underway. [Table: see text]