Construction of genomic instability⁃associated LncRNA models to predict prognosis and cisplatin sensitivity in colon cancer patients / 安徽医科大学学报

ABSTRACT

Objective @#To explore whether the long⁃chain noncoding RNA associated with genomic instability in colon cancer can predict clinical prognosis and therapeutic.@*Methods @#The R package " limma" was used for differential analysis , and the prognostic risk model was constructed by univariate Cox analysis and multivariate Cox proportional risk regression analysis. The difference in prognosis was evaluated by Kaplan⁃Meier method , and the difference was significant by log⁃rank test. The efficiency of the prognostic model was evaluated using a time⁃dependent area under the subject operating characteristic curve (AUC) . The R package“ pRRophetic ”was used to predict the sensitivity of patients to anticancer drugs. R software package rms was used to build a line graph , and the consistency index of the line graph was calculated. Real⁃time quantitative PCR was used to detect the difference in the expression levels of prognostic protective factors. @*Results @#A total of 22 LncRNAs associated with genomic instabili⁃y in patients with prognosis were obtained , 2 were protective factors for prognosis in patients with colon cancer, and 20 were risk factors for prognosis. A prognosis model composed of LncRNAs associated with genomic instability was constructed , and patients with high risk scores had lower AUCs and shorter median survival. The five⁃year survival AUC predicted by the model was 0. 823 in the training set , 0. 722 in the validation set , and 0. 759 in the overall TCGA colon cancer patient population. Patients with low risk scores had lower half inhibitory concentration (IC50 ) of cisplatin and higher sensitivity (P < 0. 000 1) . The expression of prognostic protective factors in colon cancer tissues was significantly lower than that in adjacent colon cancer tissues. @*Conclusion @#A prognostic risk model composed of 8 LncRNAs associated with genomic instability was constructed and verified. In addition , the model can also predict cisplatin drug sensitivity. A histogram was constructed combining the tumor stage and the prognosis model. The predictive ability of this graph for five⁃year survival of colon cancer patients is better than that of traditional histopathological features and prognostic models constructed by predecessors.