Joint Decision-Making Model Based on Consensus Modeling Technology for the Prediction of Drug-Induced Liver Injury

Table 2

Comparison results of five consensus models.

Descriptor or fingerprints	Performance (%)
	Training set				Fivefold cross validation set				Test set
	ACC	SE	SP	MCC	ACC	SE	SP	MCC	ACC	SE	SP	MCC	AUC
PaDEL	96.2	96.1	96.2	91.8	72.1	71.4	73.2	43.3	75.4	74.1	77.6	50.2	81.8
PubchemFP	97.2	98.0	95.8	94.0	73.2	78.3	64.4	43.6	78.4	82.8	70.9	53.8	81.4
KRFP	97.4	98.1	96.1	94.3	70.8	77.0	60.1	37.2	78.4	83.7	69.4	53.5	82.9
SubFP	93.0	95.3	89.2	85.0	71.2	75.9	63.2	38.8	75.2	78.0	70.4	47.8	81.8
EstateFP	86.4	91.6	77.4	70.4	68.1	76.7	53.6	30.7	69.9	78.9	54.6	34.3	75.2