Abstract
Purpose: Pancreatic cancer (PC) is an inflammatory tumor. Tumor microenvironment (TME) plays an important role in the development of PC. This study aims to explore hub genes of TME and establish a prognostic prediction system for PC.Methods: High throughput RNA-sequencing and clinical data of PC were downloaded from TCGA and ICGC database, respectively. PC Patients were divided into High- and low-score group by using stromal, immune scores system based on ESTIMATE. Differentially expressed genes (DEGs) between High- and low-score patients were screened and survival related DEGs were identified as candidate genes by univariate COX regression analysis. Final variables for establishment of the prognostic prediction system were determined by LASSO analysis and multivariate COX regression analysis. The predictive power of the prognostic system was evaluated by internal and external validation. Results: A total of 210 candidate genes were identified by stromal, immune scores system and survival analyses. Finally, the prognostic risk score system was constructed by the following genes: FAM57B, HTRA3, CXCL10, GABRP, SPRR1B, FAM83A, LY6D. In process of internal validation, Harrell's C-index of this prognostic risk score system was 0.73, and the area under the receiver operating characteristic curve (AUC) value of 1-year, 2-year and 3-year OS period was 0.67, 0.76 and 0.86, respectively. In the external validation set, the survival prediction C-index was 0.71, and the AUC was 0.81, 0.72, 0.78 at 1-year, 2-year and 3-year, respectively.Conclusion: This prognostic risk score system based on TME demonstrated a good predictive capacity to the prognosis of PC.