Table 1.

Statistical parameters of four different PLS models developed from GRIND using different 3D conformational inputs.

Conformational method	Fractional factorial design (FFD) cycle									Comment FFD2 (LV2)
	Complete variable			FFD1			FFD2
	${\mathit{\text{q}}}_{\mathrm{\text{LOO}}}^2$	r2	SDEP	${\mathit{\text{q}}}_{\mathrm{\text{LOO}}}^2$	r2	SDEP	${\mathit{\text{q}}}_{\mathrm{\text{LOO}}}^2$	r2	SDEP
Minimum energy conformation	0.38	0.51	1.09	0.45	0.56	1.09	0.45	0.56	1.09	Non-consistent with respect to auto and cross-correlogram
Stochastic search conformation	0.34	0.45	1.08	0.42	0.51	1.01	0.46	0.53	0.91	Non-consistent with respect to auto and cross-correlogram
Docking conformations	0.32	0.46	1.12	0.39	0.51	1.08	0.47	0.56	1.41	Non-consistent with respect to auto and cross-correlogram
Standard 3D conformations	0.54	0.62	0.94	0.61	0.67	0.86	0.63	0.69	0.84	Consistent with respect to TIP-TIP, DRY-TIP, and N1-N1 correlogram (Figure 6)

The bold number represents finally selected model.