Implementation of multiple-instance learning in drug activity prediction

BMC Bioinformatics

Table 5 Predictive performance for different dissimilarity measures

Data set	Dissimilarity measure	Training set		Test set
		Accuracy	MCC	Accuracy	MCC
I	Soergel	0.972	0.944	0.854	0.714
	Dice	0.979	0.959	0.825	0.653
	Manhattan^a	0.941	0.881	0.861	0.725
	Rogers-Tanimoto	0.961	0.923	0.861	0.725
II	Soergel	0.965	0.933	0.860	0.725
	Dice	0.965	0.933	0.868	0.745
	Manhattan^a	0.978	0.956	0.904	0.807
	Rogers-Tanimoto	0.973	0.946	0.897	0.793
III	Soergel	0.989	0.977	0.846	0.706
	Dice	0.989	0.977	0.855	0.717
	Manhattan	0.979	0.954	0.838	0.686
	Rogers-Tanimoto^a	0.947	0.885	0.846	0.711
IV	Soergel	0.904	0.823	0.667	0.301
	Dice	0.904	0.823	0.635	0.307
	Manhattan	0.957	0.918	0.714	0.433
	Rogers-Tanimoto^a	0.898	0.811	0.794	0.584

ISSN: 1471-2105