. 2016 Nov 16;1(5):923–929. doi: 10.1021/acsomega.6b00247

Table 2. Performance Measures in Predicting Pgp Inhibitors by Using the Data Set of Broccatelli et al¹⁰.

method	parameters	accuracy	sensitivity	specificity	κ	coverage
Training^a
v-NN	HC (h = 0.2; d₀ = 1.0)^b	0.85	0.86	0.84	0.70	1.00
v-NN	HA (h = 0.2; d₀ = 0.6)	0.91	0.93	0.88	0.81	0.67
FLAP/VolSurf+		0.88	0.84	0.91	0.75	1.00
Internal Test
v-NN	HC (h = 0.2; d₀ = 1.0)	0.84	0.84	0.83	0.67	1.00
v-NN	HA (h = 0.2; d₀ = 0.6)	0.89	0.88	0.91	0.78	0.66
FLAP/VolSurf+		0.85	0.82	0.87	0.69	1.00
External Test
v-NN	HC (h = 0.2; d₀ = 1.0)	0.76	0.81	0.67	0.48	1.00
v-NN	HA (h = 0.2; d₀ = 0.6)	0.88	0.91	0.80	0.71	0.53
FLAP/VolSurf+		0.86	0.90	0.80	0.70	1.00
Randomized Training Data^c
v-NN	HC (h = 0.2; d₀ = 1.0)	0.55	0.41	0.67	0.08	1.00
v-NN	HA (h = 0.2; d₀ = 0.6)	0.53	0.41	0.67	0.08	0.67

Performance of 10-fold cross validation.

v-NN parameters (smoothing factor, h, and Tanimoto-distance threshold, d₀).

Training set compounds for Pgp were randomly assigned as substrates or nonsubstrates.

Table 2. Performance Measures in Predicting Pgp Inhibitors by Using the Data Set of Broccatelli et al10.