Construction Of A Prognostic Model For Cholangiocarcinoma Based On MRNA And LncRNA Transcriptome Levels

Background: Morbidity and mortality rate of cholangiocarcinoma(CCA)have been rising globally.Most patients with CCA have extremely poor prognosis,partly due to the silent clinical character and the lack of the method for early diagnosis.The purpose of this study was to develop a molecular prognostic signature by combining transcriptome profiles of both mRNAs and lncRNAs to improve the ability of prognostic prediction in CCA patients.Methods: We collected a training dataset of 45 samples from The Cancer Genome Atlas(TCGA)database and a validation cohort(GSE107943)of 57 samples from Gene Expression Omnibus(GEO)database.An integrated mRNA-lncRNA risk score was established by univariate and multivariate Cox regression analyses.Time-dependent receiver operating characteristic(ROC)analysis was used to evaluate prognostic performance of the integrated mRNA-lncRNA signature.Moreover,we conducted a correlation analysis between the risk score model and different clinical characteristics,and preformed gene set enrichment analysis(GSEA)to investigate underlying biological function of the integrated mRNA-lncRNA signature.Weighted gene coexpression network analysis(WGCNA)was used to identify key genes associated with the integrated mRNA-lncRNA risk score.Results: A total of two mRNAs(CFHR3 and TROAP)and two lncRNAs(AC007285.1and AC134682.1)were identified to construct the integrated signature through univariate Cox regression and multivariable Cox analysis.The ROC curve suggested the integrated mRNA-lncRNA signature possessed a high specificity and sensitivity of prognostic prediction with an area under the curve(AUC)of 0.813 and 0.731 at 1-year and 3-years,respectively.Subsequently,the signature was validated in GSE107943 cohort and combined dataset,and the AUC reached up to 0.735 and 0.746 for survival prediction at 3-years.The signature was not only independent from different clinical features(HR=12.60,P=1.51E-02),but also outperformed other clinical characteristics as prognostic biomarkers with AUC of 0.746 at 3 years.GSEA method suggested that genes highly expressed in high risk score group were enriched in complement and coagulation cascades,PPAR signaling pathway,biomacromolecules synthesis and metabolism related pathways,which were highly involved in CCA carcinogenesis.A total of 13 key genes was identified as associated with the mRNA-lncRNA signature by WGCNA method.Conclusion: In this study,we constructed an mRNA-lncRNA prognostic model based on mRNA and lncRNA transcriptome data.The prognosis model may be able to identify the high-risk patients with poor prognosis and serves as an independent predictor of prognosis for CCA patients.