AutoGPA: An Automated 3D-QSAR Method Based on Pharmacophore Alignment and Grid Potential Analysis

Table 1

Statistics of ten AutoGPA models based on the training set.

Model	Overlap	PH4	NOC	Grids	MSE	𝑟^2	𝑞^2
1	46.91	RRRd	6	28	0.092	0.923	0.760
2	46.45	RRHa	7	16	0.141	0.882	0.731
3	46.44	RRHd	4	28	0.214	0.821	0.724
4	46.83	RRda	4	22	0.174	0.854	0.695
5	46.91	RRRa	4	17	0.218	0.818	0.675
6	46.84	RRHd	3	21	0.252	0.789	0.661
7	46.27	RHda	4	25	0.201	0.832	0.660
8	46.49	RRda	3	19	0.233	0.805	0.650
9	46.92	RRRd	3	26	0.249	0.791	0.626
10	46.41	RHda	5	19	0.192	0.839	0.612
CoMFA*	—	—	5	—	0.354	0.907	0.737

Overlap: atomic overlapping score in pharmacophore-based alignment.
PH4: pharmacophore feature labels; R: aromatic or π-ring center, H: hydrophobic, d: projected donor, a: projected acceptor.
NOC: number of components.
Grids: number of grid points for QSAR model.
MSE: mean squared error.
𝑟²: correlation coefficient.
𝑞²: predictive coefficient in leave-one-out cross-validation.
*The CoMFA model obtained by Abdul-Hameed et al. [8].