Predictive models for anti-tubercular molecules using machine learning on high-throughput biological screening datasets

Table 2 Overall summary of the performance of the models

Dataset	Classifier*	TP%	FP%	TN%	FN%	Accuracy	ROC area	BCR ^#	MCC ^$
AID1626	CSCNB	44.4	19.3	80.7	55.6	80.32%	0.707	0.625	0.061
	CSCRF	65.7	17.5	82.5	34.3	82.36%	0.823	0.741	0.122
	CSCSMO	51.2	18.8	81.2	48.8	80.88%	0.750	0.662	0.080
	MetaCostJ48	59.3	19.6	80.4	40.7	80.24%	0.690	0.698	0.097
AID1949	CSCNB	46.9	19.4	80.6	53.1	80.04%	0.712	0.637	0.080
	CSCRF	69.2	18.5	81.5	30.8	81.25%	0.825	0.753	0.162
	CSCSMO	54.1	19.9	80.1	45.9	79.63%	0.744	0.671	0.107
	MetaCostJ48	61.6	18.5	81.5	38.4	81.17%	0.718	0.715	0.138
AID1332	NB	39.4	19.5	80.5	60.6	74.31%	0.686	0.599	0.171
	CSCRF	81.8	17.3	82.7	18.2	82.57%	0.876	0.822	0.521
	CSCSMO	72.7	16.2	83.8	27.3	82.11%	0.806	0.782	0.469
	MetaCostJ48	72.7	17.3	82.7	27.3	81.19%	0.762	0.777	0.454

ISSN: 1756-0500