Table 2.

Predictive performance for DFS and OS using various features based on C-index and IBS.

Features	C-index		IBS
	OS	DFS	OS	DFS
Clinical	0.7955±0.04	0.7894±0.04	0.338	0.319
KG+Clinical	0.8019±0.04	0.7937±0.04	0.349	0.317
KG+Sub+Clinical	0.8057±0.04	0.8363±0.03	0.326	0.283
HR+Clinical	0.8152±0.04	0.8125±0.04	0.329	0.308
Sub+Clinical	0.8026±0.04	0.8361±0.03	0.329	0.286
HR+Sub+Clinical	$0.8157\pm 0.04$	$0.8388\pm 0.03$	0.318	0.265

Logistic regression was used as a prediction model. Clinical, nine clinical data; KG, known cancer driver genes (KRAS, CDKN2A, TP53, and SMAD4); HR, a high-risk group of patients harboring mutations in five genes with high CP values; Sub, a feature representing subgroups generated by integrating mRNA, miRNA, and DNA methylation subtypes.