Gsslasso Cox: a Bayesian hierarchical model for predicting survival and detecting associated genes by incorporating pathway information

Table 6 The measures of optimal group spike-and-slab lasso (gsslasso) cox and the lasso cox models for TCGA ovarian cancer, lung adenocarcinoma (LUAD) and breast cancer dataset with pathway genes by 10 times 10-fold cross validation

	Pathway number	Genes included	Methods	CVPL	C-index	Number of non-zero gene
TCGA	271	4260	gsslasso	− 1041.218 (2.118)	0.577 (0.012)	33
ovarian			lasso	− 1042.905 (1.687)	0.533 (0.027)	15
cancer			grlasso	− 1044.110 (12.741)	0.504 (0.014)	24
N = 304			grMCP	− 1046.965 (8.604)	0.502 (0.007)	24
			grSCAD	−1042.349 (5.339)	0.503 (0.012)	24
			cMCP	− 1043.373 (2.215)	0.532 (0.019)	13
TCGA	274	4266	gsslasso	− 938.973 (1.675)	0.559 (0.010)	64
LUAD			lasso	− 941.383 (3.720)	0.545 (0.019)	13
N = 491			grlasso	− 945.605 (8.137)	0.547 (0.023)	111
			grMCP	− 1092.091 (30.477)	0.512 (0.015)	25
			grSCAD	− 940.358 (1.331)	0.538 (0.021)	123
			cMCP	−942.831 (3.301)	0.530 (0.022)	3
TCGA	275	4385	gsslasso	−996.491 (2.131)	0.640 (0.153)	86
Breast			lasso	−1002.046 (5.356)	0.523 (0.027)	2
cancer			grlasso	− 1001.073 (9.641)	0.590 (0.022)	93
N = 1082			grMCP	− 1016.864 (25.290)	0.520 (0.019)	12
			grSCAD	− 1005.299 (2.268)	0.522 (0.007)	24
			cMCP	− 1012.587 (44.339)	0.502 (0.012)	1

Note: Values in the parentheses are standard errors. For group spike-and-slab lasso model, the optimal s₀ = 0.03 for three data sets. In TCGA ovarian cancer, we mapped 4260 genes into 271 pathways. The analyses was performed on these genes including in these pathways. The same is true for other two datasets

ISSN: 1471-2105