Predicting drug side effects by multi-label learning and ensemble learning

BMC Bioinformatics

Table 2 The 5-CV performances of individual feature-based MLKNN models on Liu’s dataset

Features	AUC	AUPR	Hamming loss	Ranking loss	One error	Coverage	Average precision
Enzyme	0.8861 ± 0.0006	0.3989 ± 0.0011	0.0483 ± 0.0001	0.0839 ± 0.0002	0.1695 ± 0.0053	837.7197 ± 1.6124	0.4551 ± 0.0005
Pathway	0.8884 ± 0.0006	0.4105 ± 0.0010	0.0477 ± 0.0001	0.0802 ± 0.0001	0.1865 ± 0.0076	827.1183 ± 2.9986	0.4721 ± 0.0007
Target	0.8947 ± 0.0009	0.4424 ± 0.0017	0.0464 ± 0.0001	0.0745 ± 0.0003	0.1695 ± 0.0061	812.6752 ± 2.9022	0.4919 ± 0.0010
Transporter	0.8863 ± 0.0006	0.4010 ± 0.0013	0.0482 ± 0.0001	0.0826 ± 0.0002	0.1661 ± 0.0041	836.2058 ± 2.8593	0.4644 ± 0.0007
Indication	0.8948 ± 0.0004	0.4566 ± 0.0020	0.0456 ± 0.0001	0.0762 ± 0.0003	0.1363 ± 0.0034	818.3745 ± 3.6611	0.4950 ± 0.0012
Substructure	0.8912 ± 0.0005	0.4255 ± 0.0015	0.0472 ± 0.0001	0.0754 ± 0.0004	0.1760 ± 0.0040	808.9192 ± 2.4440	0.4888 ± 0.0014

ISSN: 1471-2105