Inferring biological networks with output kernel trees

BMC Bioinformatics

Table 1 AUC results.

Inputs	All	TF vs. LF	TF vs. TF	Kern. (All)
Protein-protein interactions
expr	0.851 ± 0.028	0.859 ± 0.027	0.819 ± 0.082	0.776
phy	0.693 ± 0.036	0.698 ± 0.035	0.617 ± 0.064	0.767
loc	0.725 ± 0.018	0.726 ± 0.017	0.710 ± 0.055	0.788
expr+phy+loc	0 887 ± 0 024	0 891 ± 0 023	0 845 ± 0 081	-
y2h	0.790 ± 0.023	0.795 ± 0.022	0.692 ± 0.068	0.612
expr+phy+loc+y2h	0.910 ± 0.019	0.914 ± 0.017	0.865 ± 0.057	0.939
Metabolic network
expr	0.714 ± 0.032	0.732 ± 0.035	0.619 ± 0.089	0.706
Phy	0.815 ± 0.033	0.819 ± 0.031	0.721 ± 0.086	0.747
loc	0.587 ± 0.022	0.587 ± 0.022	0.592 ± 0.042	0.577
expr+phy+loc	0.847 ± 0.025	0.853 ± 0.025	0.733 ± 0.057	0.804
y2h	0.639 ± 0.033	0.650 ± 0.034	0.490 ± 0.098	-
expr+phy+loc+y2h	0.844 ± 0.025	0.851 ± 0.026	0.721 ± 0.056	-

AUC results obtained with extra-trees and ten-fold cross-validation compared with full kernel-based methods. The best result in each row between tree-based and kernel-based methods (for all predictions) is underlined.

ISSN: 1471-2105