Research Paper Volume 15, Issue 22 pp 12817—12851

Figure 2. Establishment and refinement of DUXAP8/CBX2/CEP55-centered prognostic model. (A) Shared model candidate survival-related RNAs from which was selected with Lasso Cox and RSF from the refined regulatory networks. (B) Forest plots of multivariate Cox showed the hazard ratio (HR), 95% confidence interval (CI), and corresponding P-values of model-used CPEB3, ANXA10, CEP55, CBX2, and DUXAP8 (C) Time-dependent AUC within 5 years of a prognostic model related to 5-survival genes using two-fold cross validation with 50 randomly repeated replications. (D) The examples of time-dependent ROC at 1-, 3-, and 5-year corresponding to the AUC in (C). (E) Kaplan-Meier plots of the risk score predicted with the prognostic model in TCGA-LIHC. The high and low risk group was determined with the median of risk score. P value was calculated by log-rank test. (F) Forest plots of multivariate Cox showed HR and p value of TNM, age, gender and risk score. (G) Time dependent ROC and AUC at 1-, 3-, and 5-year predicted with external validation using ICGC-LIRI-JP by 5-survival-gene prognostic model. RSF, Random Survival Forest; AUC, area under the curve; ROC, receiver operating characteristic curve; HR, hazard ratio.