Table 8.

The statistical information of GSAE outputs between the training and test TCGA data sets of four cancer types

			Two proportion z-test
TCGA data set	Superset Jaccard Indexa	Gene set Jaccard Indexb	Superset Proportionc	Gene set Proportiond	P-valuee
BRCA	0.344	0.124	11 / 24	31 / 197	0.0002
LUAD	0.182	0.113	6 / 12	32 / 145	0.0150
SKCM	0.179	0.069	5 / 19	17 / 139	0.0485
LGG	0.483	0.475	29 / 45	299 / 481	0.3821

Supersets/gene sets with log-rank P-value < 0.05 were selected as prognostic significant sets. ^aJaccard index of significant supersets between training and test data. ^bJaccard index of significant gene sets between training and test data. ^cSuperset proportion: (# of overlapped significant supersets between training and test data) over (# of significant supersets in training data). ^dGene set proportion: (# of overlapped significant gene sets between training and test data) over (# of significant gene sets in training data). ^eThe P-value of z-test comparing superset and gene set proportions