Network-based biomarkers enhance classical approaches to prognostic gene expression signatures

BMC Systems Biology

Table 1 Summary of feature selection methods analysed.

Approach	Method name	Feature selection	Classification Feature	Feature value	Existing method?	Ref.
Single-gene	Mod-t	Rank genes by p-value	Single gene	Expression value	Yes	Smyth et al. [30]
Gene-set	Median expression	Rank gene-sets by p-value	Gene-set	Median expression value of all genes in the gene-set	No
Network	NetRank	Rank genes using NetRank algorithm	Single gene	Expression value	Yes	Winter et al.[11]
	Taylor	Rank sub-networks using Taylor's correlation-based measure	Sub-network edge	Expression difference between hub and interactor gene connected by the edge	Yes	Taylor et al. [8]
	BSS/WSS	Rank sub-networks using the BSS/WSS correlation-based measure	Single (hub) gene	Expression value of the gene	No

Description of the methods assessed in terms of the way in which external information is incorporated, the feature selection procedure, the features used for classification, and a reference if the method is obtained from a previous publication.

ISSN: 1752-0509