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

Table 7

Comparison of the present model with the main published models (Nm > 1304).

Ref.	Method	Nm	Nde	Performance	Coverage	Mechanism analysis
[5]	SVM	2295	166	TV: 80.39% ACC, 88.15% SE, 65.73% SP	—	Yes
				EV: 82.78% ACC, 93.18% SE, 68.25% SP
[9]	RF	2513	206	68.7 ± 1.7% ACC, 81.4 ± 1.6% SE, 50.8 ± 4.6% SP	—	No
[31]	GA-SVM	2171	674	ACC: 75%, TV: 68% SE, 95% SP	—	Yes
[40]	MC4PC, MDL-QSAR, BioEpisteme, PDM^a	1608	Not given	CV: 39.2 ± 2.6% SE, 87.1 ± 2.6% SP	93.5 ± 3.8%	No
				EV: 88.9% SE
[39]	SVM, NB, kNN, CT, RF	1317	307	TV: 65.74% ACC, 85.16% SE, 34.38% SP	—	Yes
				EV: 75% ACC, 93.22% SE, 37.93% SP
Present	Joint decision-making based on SVM	2608	150	CSM(T = 50%): CV5 (70.8%–73.2% ACC); JDM(T = 70%): TV (80.0% ACC, 83.9% SE, 73.3% SE); EV (79.8% ACC, 96.5% SE, 66.8% SP)	JDM(T = 70%)	Yes
					TV: 99.24%
					EV: 98.20%

SVM: support vector machine; GA-SVM: genetic algorithm-support vector machine; RF: random forest; ^adetailed information of these methods can be found in [40]; NB: naive Bayes; kNN: k-nearest neighbor; CT: classification tree; N_m is the number of compounds in the modeling dataset; N_de is the number of descriptors/fingerprints used in (Q)SAR models; CV: cross-validation; CV5: fivefold cross-validation; EV: external validation; TV: validation on the test set; CSM: consensus model; JDM: joint decision-making model.