Table 2. Model performance estimation by 10-fold cross validation.

Multi-instance transfer learning		Multi-instance Learning			Multi-instance Learning Novel				Single Instance Learning
		SP	SE	MCC	SP	SE		MCC	SP	SE	MCC
	Positive	0.7692	0.8065	0.6338	0.7246	0.8065		0.5936	0.7031	0.7258	0.5290
	Negative	0.7966	0.7581	0.6284	0.7818	0.6935		0.5780	0.7167	0.6935	0.5234
	Accuracy	78.23%			75%				70.97%
	MCC	0.6306			0.5833				0.5260
	ROC-AUC	0.8335			0.8176				0.8003
	PR-AUC	0.8678			0.8369				0.8325
	F1 Score	0.80			0.76				0.71
*Random forest [16]		SP					SE
	Positive	0.817					0.407
	F1 Score	0.52
*Multi-task learning [17]	F1 Score	0.758

Note: * denotes the existing models.