Hubs of knowledge: using the functional link structure in Biozon to mine for biologically significant entities

BMC Bioinformatics

Table 4 Performance evaluation: focused subgraph vs. global vs. extended global.

Query Term	Query Type	Graph	Number of Results	Average No. Neighbors	Average Consistent Neighbors	Average Ratio	Q(R)	UROC (R)
ubiquitin	protein	Baseline	50	1.92	1.0	0.64	329	8278
		Focused	50	11.72	2.9	0.20	771	18516
		Global	50	12.18	3.56	0.27	977	26021
		Global-ext	50	12.18	3.56	0.27	977	25967
stromelysin	protein	Baseline	46	1.30	0.80	0.49	387	9038
		Focused	46	1.30	0.80	0.49	387	13180
		Global	43	1.19	0.79	0.48	357	11593
		Global-ext	46	1.30	0.80	0.49	387	13045
cancer	protein	Baseline	50	2.01	1.03	0.73	323	8202
		Focused	50	6.72	1.74	0.28	516	14901
		Global	50	6.68	1.8	0.30	535	16082
		Global-ext	50	6.72	1.8	0.30	534	16010
autoimmune	protein	Baseline	50	1.92	0.84	0.60	455	11512
		Focused	50	2.33	1.02	0.73	530	14330
		Global	30	1.1	0.67	0.58	223	4582
		Global-ext	50	2.24	0.98	0.70	530	14054

All results were generated with the PageRank prominence model, using the Principal Eigenspace scoring function. In the cases of autoimmune and stromelysin, the global method returns less than 50 documents. The extended global mode reports more than 50 documents, which explains the larger overall quality.

ISSN: 1471-2105