A New Molecular Prognostic Model Of Pancreatic Cancer Based On Endoplasmic Reticulum Stress Genes

Objective:We constructed a prognostic model to clarify the role of endoplasmic reticulum stress gene in the prognosis of pancreatic cancer,to search for prognostic markers of pancreatic cancer.Method:Part ⅠIn the first part,transcriptome sequencing data and clinical data of 177 pancreatic cancer samples were obtained from The Human Cancer Genome Atlas(TCGA).The transcriptome sequencing data and clinical data of 166 normal pancreatic samples were obtained from the Genotype-Tissue Expression(GTEX)database,and the expression matrices of tumor and normal groups were obtained by normalization.A total of 260 endoplasmic reticulum stress genes were obtained from The Molecular Signatures Database(Msig DB)website,and differential analysis of endoplasmic reticulum stress genes between tumor and normal groups was performed by Univariate Cox regression analysis,to obtain endoplasmic reticulum stress genes related to the prognosis of pancreatic cancer,molecular typing of endoplasmic reticulum stress genes of pancreatic cancer was constructed based on the prognostic genes by concordance cluster analysis,kaplan-meier survival curves between molecular subtypes were plotted.Part ⅡWe analyzed the different gene expression of endoplasmic reticulum stress gene molecular typing using Gene Ontology(GO)and KEGG analysis(KEGG).The differential genes related to prognosis of pancreatic cancer were obtained by univariate Cox regression,and the core genes were obtained by the least selection operator(Lasso)regression analysis,which was used to construct the prognostic prediction model of pancreatic cancer and calculate the risk score.Risk scoring was performed on 177 samples from the TCGA pancreatic cancer cohort,which was divided into high-risk and low-risk groups based on median risk values,the 177 sets of data were randomly divided into the training set(n = 125)and the test set(n = 52)in a 7:3 ratio,and ROC curves were plotted to assess the predictive performance of the model.GSE57495 data set was obtained from high-throughput Gene Expression Omnibus data base(GEO)to validate the predictive performance of the model.Finally,the immune infiltrates between the high-risk and low-risk groups in the TCGA pancreatic cancer cohort were analyzed.Results:Part One1.15 endoplasmic reticulum stress genes were differentially expressed in pancreatic cancer and normal tissues,and PIK3R2,NUPR1,RNF186,FBXO2 were low expressed in pancreatic cancer tissues.UBQLN2,UBXN10,PMAIP1,MARCKS,RNF139,CCND1,ANKS4 B,PPP1R15A,CEBPB,THBS1 were highly expressed in tumor tissues.2.The higher expression of MARCKS was associated with better prognosis of pancreatic cancer,while the higher expression of CEBPB,PMAIP1,UBXN10 were associated with worse prognosis of pancreatic cancer.3.Based on the expression characteristics of CEBPB,PMAIP1,UBXN10 and MARCKS,the TCGA pancreatic cancer cohort could be divided into two subgroups,cluster A and cluster B.The overall survival of patients in cluster A was shorter than that of Cluster B.Part Two1.A total of 51 differentially expressed genes were identified between cluster A and cluster B.11 genes were highly expressed in cluster A,40 genes were highly expressed in cluster B,and GO and KEGG analysis suggested that the differentially expressed genes clustered on the apoptotic pathway and phagosomes formation.2.Among the 51 genes,14 genes were associated with the prognosis of pancreatic cancer,which were DDIT4,GPR87,LY6 D,RHOV,SFTA2,SPRR1 B,CEBPB,CLIC3,CDA,GNA15,KRT6 B,PTGES,BHLHE40,AHNAK2.All of them were unfavorable factors of overall survival of pancreatic cancer.3.Lasso regression analysis screened out 5 core genes from 14 endoplasmic reticulum stress-related genes.Risk score = 0.156 * CDA + 0.135 * AHNAK2 + 0.020* RHOV + 0.095 * LY6 D + 0.054 * SPRR1 B.ROC curve showed that the model had excellent overall predictive performance,the AUC was 0.731 within 1 year,0.712 with3 years,and 0.686 within 5 years.The overall survival time of the low-risk group was longer than that of the high-risk group.4.The validation results of the external data set GSE57495 showed that the prediction performance of the model was good,the AUC was 0.674 within 1 year,0.680 within 3 years,and 0.608 within 5 years..5.The risk score was positively correlated with the infiltration of M0 macrophages,M2 macrophages and memory B cells,and was negatively correlated with the infiltration of monocytes,activated NK cells,primary B cells and plasma cells.Conclusion:1.The expression of endoplasmic reticulum stress gene was correlated with the prognosis of pancreatic cancer.CEBPB,MARCKS,PMIP1,UBXN10 may be the independent prognostic factors of pancreatic cancer,and MARCKS may be the protective factor,the high expression of CEBPB,PMIP1 and UBXN10 was associated with poor prognosis.2.The whole expression of endoplasmic reticulum stress gene affected the overall survival of pancreatic cancer patients.According to the expression of CEBPB,MARCKS,PMAIP1,UBXN10,177 pancreatic cancer patients in TCGA cohort could be divided into two subtypes,clusters A and B,and cluster B patients had longer overall survival.3.Endoplasmic reticulum stress-related genes DDIT4,GPR87,LY6 D,RHOV,SFTA2,SPRR1 B,CEBPB,CLIC3,CDA,GNA15,KRT6 B,PTGES,BHLHE40,AHNAK2 may be independent prognostic factors of pancreatic cancer,and were all associated with poor prognosis of pancreatic cancer.4.The expression of CDA,AHNAK2,RHOV,LY6 D and SPRR1 B could predict the prognosis of pancreatic cancer.The high expression of this group of genes was associated with shorter overall survival,which could be used as prognostic markers of pancreatic cancer.5.The high-risk group characterized by high expression of CDA,AHNAK2,RHOV,LY6 D and SPRR1 B had more M2 macrophage infiltration,and the tumor microenvironment was in an immunosuppressive state.In the low-risk group,activated NK cells,naive B cells and plasma cells infiltrated well and led to better anti-tumor immunity.