Table 4.

Cox p-values obtained from four clustering methods (consensus clustering, hierarchical clustering, local shrinkage, and cluster ensemble) using seven sets of genes: all genes and the signatures obtained from IA + addCLT, IA + Fisher, GSEA + addCLT, GSEA + Fisher, GSA + Fisher, and NBIA. Cells with emboldening text have p-values smaller than 5%. Using any of the clustering methods, NBIA has the most significant p-values. In addition, it is the only method that provides significant p-values across all four clustering methods.

	All genes	IA + addCLT	IA + Fisher	GSEA + addCLT	GSEA + Fisher	GSA + Fisher	NBIA
Consensus clustering	0.359	0.016	0.001	0.145	0.089	0.072	1e-04
Hierarchical clustering	0.742	0.004	0.002	0.009	0.896	0.345	3e-04
Local shrinkage	0.066	0.002	0.09	0.788	0.02	0.24	4e-05
Cluster ensemble	0.591	0.902	0.048	0.916	0.068	0.132	2e-06