Research Paper Advance Articles

Figure 7. Development and validation of the dCAF-based prognostic signature. (A) Workflow for feature gene selection and HR values of the 8 feature genes with 95% CI. (B) Characteristics of coefficient changes of variables; selection of tuning parameter λ and the corresponding number of variables through cross-validation in LASSO model. (C) Depths of feature genes calculated by the RSF model. (D) Gain values of feature genes calculated by the XGBoost model. (E) KM survival analysis was performed in the low- and high-risk groups categorized by the median risk score in the training set (TCGA) and validation sets (GSE63885, GSE26712, ICGC_AU). (F) ROC and AUC values with 95% CI of 1-, 3-, and 5-year survival prediction by the dCAF-based risk score in the training set (TCGA) and testing sets (GSE63885, GSE26712, and ICGC_AU). (***p < 0.001, **p < 0.01, *p < 0.05).