Skip to main content
. 2023 Jul 3;39(7):btad422. doi: 10.1093/bioinformatics/btad422

Figure 4.

Figure 4.

Feature importance analysis. (A) The feature importance comparison of the 28 CNV features on five splits calculated by GradientShap. GradientShap values on five splits are presented as boxplot. (B) The feature importance visulization on the CNV-DAG, where higher feature importance is shown with higher opacity. For clarity, virtual nodes and virtual edges are not shown. (C and D) Kaplan–Meier curve of PFS and OS probability according to the clustering result in Section 3.3 (Cluster 1 and Cluster 3: the top two overlapped lines, Cluster 2: the lowest separate line). The P-values are calculated according to a two-sided log-rank test