Integrative approaches to the prediction of protein functions based on the feature selection

BMC Bioinformatics

Table 2 Contributions of genomic data sources

Data source		Exhaustive search			KL1LR
		# of GO terms (# in union)		AUC	# of GO terms (# in union)		AUC	NCG
Protein-protein interactions	OPHID	192		0.82	201		0.89	83
Protein domain	Interpro	522	(697)	0.87	408	(518)	0.89	266
	Pfam	600		0.86	311		0.89	210
Phenotype	MGI	213		0.87	346		0.90	129
Phylogenetic profile	BioMart	33	(95)	0.83	59	(166)	0.88	4
	Inparanoid	70		0.84	124		0.88	22
Disease	OMIM	41		0.85	32		0.88	3
Gene expression	Zhang et al.	28		0.81	147		0.90	10
	Su et al.	21	(55)	0.82	158	(309)	0.89	8
	Sage et al.	16		0.83	113		0.90	7

The numbers of GO terms satisfying the cut-off of prediction accuracies by the AUC and P20R values are presented for each data source along with the average AUC values of the GO terms. For the protein domain, phylogenetic profile, and gene expression data, the number of terms in the union set is shown in parentheses. The numbers of common terms between the two approaches are shown in the last column (NCG).

ISSN: 1471-2105