Gene prioritization and clustering by multi-view text mining

BMC Bioinformatics

Table 3 Prioritization performance obtained by the single controlled vocabularies and the multi-view approach

Single CV	Error of AUC	Integration of 9 complete CVs	Error of AUC
LDDB	0.0792	Order statistics	0.0990
eVOC	0.0852	Average score	0.0782
MPO	0.0974	Maximum score	0.0957
GO	0.1027	1-SVM μ_min= 0	0.0620
MeSH	0.1043	1-SVM μ_min= 0.5/N	0.0583
SNOMED	0.1129	1-SVM μ_min= 1/N	0.0509
OMIM	0.1214
Uniprot	0.1345
KO	0.1999
Integration of 9 LSI	Error of AUC	Integration of 35 subset CVs	Error of AUC
Order statistics	0.0645	Order statistics	0.0870
Average score	0.0382	Average score	0.0674
Maximum score	0.0437	Maximum score	0.0883
1-SVM μ_min= 0	0.0540	1-SVM μ_min= 0	0.1036
1-SVM μ_min= 0.5/N	0.0454	1-SVM μ_min= 0.5/N	0.0851
1-SVM μ_min= 1/N	0.0335	1-SVM μ_min= 1/N	0.0625

The experiments are repeated 20 times on random candidate gene sets and the standard deviations are all smaller than 0.01.

ISSN: 1471-2105