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

BMC Bioinformatics

Table 3 The effect of normalization on the ranking.

Query Term	Query Type	Prominence Model	Average No. Neighbors	Average Consistent Neighbors	Average Ratio	Q(R)	UROC (R)
		Hubs & Authorities	4.38	0.6	0.13	167	4879
		Hubs & Authorities(N)	7.64	2.58	0.52	705	23809
ubiquitin	protein	Eigenvector Centrality	4.38	0.6	0.13	167	4879
		Eigenvector Centrality(N)	1.04	0.9	0.88	307	7871
		PageRank	12.18	3.56	0.27	977	26021
		Hubs & Authorities	1.19	0.79	0.48	357	10086
		Hubs & Authorities(N)	1.19	0.79	0.48	357	10112
stromelysin	protein	Eigenvector Centrality	1.19	0.79	0.48	357	9912
		Eigenvector Centrality(N)	1.19	0.79	0.48	357	9057
		PageRank	1.19	0.79	0.48	357	11593
		Hubs & Authorities	5.22	1.08	0.25	317	6137
		Hubs & Authorities(N)	3.62	1.2	0.67	372	11947
cancer	protein	Eigenvector Centrality	5.2	0.9	0.21	245	5346
		Eigenvector Centrality(N)	1.24	0.98	0.86	321	8137
		PageRank	6.68	1.8	0.30	535	16082
		Hubs & Authorities	1.1	0.67	0.58	223	4520
		Hubs & Authorities(H)	1.1	0.67	0.58	223	4345
autoimmune	protein	Eigenvector Centrality	1.1	0.67	0.58	223	4520
		Eigenvector Centrality(H)	1.1	0.67	0.58	223	4830
		PageRank	1.1	0.67	0.58	223	4582

For each model we repeated the computations with the unnormalized and the normalized (N) matrices. See Table 2 for details.

ISSN: 1471-2105