Table 2. The statistical data of pharmacophore-based 3D QSAR using Phase^a,b,c.

Variant	SD		R2	F	P	Stability	RMSE	Q2	Pearson R
4-site, training 1
ADRR.87	0.566		0.911	396	9.720 x 10−61	0.838	0.897	0.657	0.824
AADR.79	0.571		0.901	400	4.848 x 10−66	0.836	0.781	0.733	0.864
AADR.139	0.564		0.906	421	2.154 x 10−67	0.786	0.768	0.740	0.872
AADR.51	0.544		0.909	444	4.568 x 10−69	0.819	0.827	0.698	0.856
5-site, training 2
AAADR.17	0.363		0.967	734	1.779 x 10−55	0.624	1.020	0.531	0.820
AAADR.20	0.309		0.976	1,021	1.055 x 10−60	0.469	0.888	0.645	0.868
AAADR.18	0.366		0.967	721	3.426 x 10−55	0.645	1.053	0.500	0.825
5-site, training 3d
AAADR.1	0.498		0.909	652	5.215 x 10−102	0.809	—	—	—
AAADR.6	0.588		0.872	449	9.198 x 10−88	0.843	—	—	—
AAADR.4	0.591		0.871	444	2.270 x 10−87	0.881	—	—	—
4-site, training 3d
AADR.5	0.567	0.906		418	1.482 x 10−66	0.732	—	—	—
AAAR.2	0.521	0.921		508	1.410 x 10−71	0.786	—	—	—
ADRR.23	0.520	0.927		465	2.724 x 10−51	0.745	—	—	—

^a Only the top 10% - 20% hypotheses scored was moved forward for evaluation.

^b Statistics obtained when PLS = 3.

^c SD, standard deviation of regression; r², correlation coefficient; F, variance ratio; stability: Stability of the model predictions to changes in the training set composition, max = 1; P, significance level of variance ratio; RMSE: root-mean-square error of the test set; q², correlation coefficient for the predicted activities; Pearson R, value for the correlation between predicted and observed activities for the test set; PLS, partial least square regression method.

^d For training set 3, all 268 compounds were used to perceive hypotheses. There was no test set. Hence the statistics for test set were empty