TABLE 8

Summary of the DDI predictions of tested approaches for induction-based DDI of 69 predicted AUC ratios

Three experiments with one batch of HepaRG cell line, 23 trials per experiment. One trial is one observed AUC ratio compared with the corresponding predicted value. See Materials and Methods for calculation of PPE, NPE, RMSE, and GMFE.

Inducer Plasma C_max Concentration	Approaches	n Trials Included in the 2-Fold Error	Percentage of Trials Included in the 2-Fold Error	FN (n)^{^a}	FP (n)^{^a}	TN (n)^{^a}	TP (n)^{^a}	NPE (%)	PPE (%)	RMSE^{^b}	GMFE^{^b}
Total	1	53	77	0	10	2	57	0	15	0.23	1.6
	d = 1
	2	54	78	0	10	2	57	0	15	0.23	1.6
	dRif
	3
	GMFE^{^c}	53	77	0	10	2	57	0	15	0.24	1.6
	RMSE^{^c}	55	80	2	8	4	55	33	13	0.24	1.7
	RMSLE^{^c}	52	75	0	9	3	57	0	14	0.24	1.6
	MAE^{^c}	53	77	0	10	2	57	0	15	0.24	1.6
Unbound	1
	d = 1	44	64	21	3	9	36	70	7.7	0.33	1.9
	2
	dRif	46	67	21	3	9	36	70	7.7	0.33	1.9
	3
	GMFE^{^c}	50	73	18	3	9	39	67	7.1	0.29	1.7
	RMSE^{^c}	44	64	14	3	9	43	61	6.5	0.31	2.2
	RMSLE^{^c}	49	71	18	3	9	39	67	7.1	0.30	1.7
	MAE^{^c}	26	38	16	3	9	41	64	6.8	0.32	2.7

d, induction scalar; 2-fold envelope, number of trials included in the 2-fold error, as also presented in Figure 4; GMFE, geometric mean fold error; inducer concentration, total or unbound plasma concentration in clinic (C_max) and in vitro (incubated inducer concentration); NPE, negative predictive error; PPE, positive predictive error; RMSE, root mean square error, RMSLE, root mean squared logarithmic error; MAE, mean absolute error.
↵^a Total number of clinical trials that were predicted as false negative (FN), false positive (FP), true negative (TN), or true positive (TP) with respect to induction based on the 0.8-fold cutoff criterion.
↵^b RMSE and GMFE used as quality criteria.
↵^c GMFE, RMSE, RMSLE, and MAE retained distances used for d calculation through a minimization process as described in Material and Methods.