Weighted-SAMGSR: combining significance analysis of microarray-gene set reduction algorithm with pathway topology-based weights to select relevant genes

Table 8 Performance statistics on the test set for the weighted-SAMGSR algorithm (PPI information retrieved from the STRING database)

	No.	Error (%)	GBS	BMC	AUPR	Rand (gene)	Rand (GS)
MS (b)	22	43.3	0.279	0.581	0.847	15.3 %	27.1 %
MS (c)	20	28.3	0.179	0.613	0.828	15.5 %	25.4 %
Stage for LC (b)	32	45.3	0.318	0.520	0.552	36.3 %	40.1 %
Stage for LC (c)	26	45.3	0.274	0.525	0.566	35.8 %	40.4 %
MC for LC (b)	22^a	47.3	0.337	0.411	0.510	--	--
MC for LC (c)	31^a	51.3	0.334	0.410	0.512	--	--

Note: (b): using the binary values indicating if two genes are connected or not; (c): using the confidence scores for the gene connectivity. MS: the multiple sclerosis application; Stage for LC: the NSCLC stage application trained on the RNA-Seq data; MC for LC: the NSCLC multiple-class application. Rand (gene): the rand index at the gene level, across the gene lists obtained from 10-fold cross-validation data; Rand (GS): the rand index at the gene set level
^ais the number of selected genes for the stage segmentation, the number of selected genes for the subtype segmentation > 300

ISSN: 1745-6150