M3S-GRPred: a novel ensemble learning approach for the interpretable prediction of glucocorticoid receptor antagonists using a multi-step stacking strategy

Table 4 Performance comparison of M3S-GRPred and top-five ML base-classifiers developed using different balanced training subsets

Evaluation strategy	Method	ACC	BACC	SN	SP	MCC	AUC
Cross-validation	PLS-Pubchem-BTS1	0.774	0.776	0.835	0.716	0.554	0.833
	SVM-CDKExt-BTS2	0.778	0.777	0.791	0.763	0.555	0.837
	XGB-FP4C-BTS5	0.780	0.780	0.796	0.764	0.560	0.855
	SVM-Pubchem-BTS4	0.782	0.781	0.796	0.767	0.563	0.824
	RF-Pubchem-BTS1	0.781	0.782	0.786	0.777	0.563	0.849
	M3S-GRPred	0.898	0.900	0.903	0.897	0.708	0.964
Independent test	PLS-Pubchem-BTS1	0.696	0.770	0.884	0.657	0.413	0.836
	SVM-CDKExt-BTS2	0.704	0.758	0.841	0.675	0.396	0.832
	XGB-FP4C-BTS5	0.744	0.742	0.739	0.745	0.387	0.819
	SVM-Pubchem-BTS4	0.774	0.835	0.928	0.742	0.523	0.880
	RF-Pubchem-BTS1	0.774	0.823	0.899	0.748	0.508	0.883
	M3S-GRPred	0.867	0.891	0.928	0.854	0.658	0.953

ISSN: 1471-2105