Table 3.

2D-QSAR models for AChE and BACE-1 inhibitors.

AChE
pIC50 = −0.928 + (2.348 × BCUT_SLOGP_3) − (0.150 × reactive) − (0.004 × PEOE_VSA + 1) − (0.005 × PEOE_VSA−3) − (0.002 × SlogP_VSA2) − (0.004 × SMR_VSA2)
Internal Validation					External Validation
N	RMSE	R2	RMSELOO	Q2LOO	N	RMSE	R2	R2(PRED)	${\mathrm{r}}_{\mathrm{m}}^2$	$\overline{{\mathrm{r}}_{\mathrm{m}}^2}$	$\Delta {\mathrm{r}}_{\mathrm{m}}^2$	CCC
50	0.18	0.70	0.22	0.57	22	0.16	0.78	0.78	0.64	0.69	0.11	0.88
BACE-1
pIC50 = 1.268 + (0.870 × petitjean) + (6.370 × BCUT_PEOE_1) + (3.305 × a_ICM) − (0.478 × chiral_u) + (0.085 × rings) + (0.157 × a_Nn) + (0.006 × PEOE_VSA − 0) + (0.022 × PEOE_VSA − 6) − (0.260 × logS) + (0.009 × SlogP_VSA3) + (0.009 × SlogP_VSA5)
Internal Validation					External Validation
N	RMSE	R2	RMSELOO	Q2LOO	N	RMSE	R2	R2(PRED)	${\mathrm{r}}_{\mathrm{m}}^2$	$\overline{{\mathrm{r}}_{\mathrm{m}}^2}$	$\Delta {\mathrm{r}}_{\mathrm{m}}^2$	CCC
150	0.37	0.80	0.40	0.77	65	0.41	0.83	0.81	0.79	0.76	0.05	0.91

N: number of compounds; RMSE (root-mean-square error), R² (squared correlation coefficient), RMSE_LOO (cross-validated root-mean-square error), Q²_LOO (cross-validated squared correlation coefficient), CCC (concordance correlation coefficient), and ${\mathrm{r}}_{\mathrm{m}}^2,\overline{{\mathrm{r}}_{\mathrm{m}}^2},\Delta {\mathrm{r}}_{\mathrm{m}}^2$ (validation metrics suggested by Roy et al. [74]).