. 2010 Dec 2;79(1):31–57. doi: 10.3797/scipharm.1011-02

Tab. 3.

Comparison of the statistical quality of the various QSAR models developed in the present work.

Using topological, structural and thermodynamic descriptors

Mod.	Stat.	Eq.	Descriptors	LVs	n_train.	s	R²	R²_a	F*	PRESS

A1	GFA-linear	–	SC-0, S_aaCH, S_dssC, S_dO <SC-3_P-20>,	–	25	0.369	0.889	0.867	40.07	4.756
A2	GFA-spline	8	³χ_p, <1.79401-S_sCH₃>, S_aasC³χ_p, (³χ_p)²,	–	25	0.315	0.919	0.903	56.95	3.021
A2a	GFA-spline	9	<1.79401-S_sCH₃>, S_aasC S_aaCH,	–	25	0.369	0.889	0.867	40.17	4.557
A3	G/PLS-linear	–	S_aasC, S_dsN, S_sOH, MolRef <6.68154-¹χ >,	2	25	0.401	0.856	0.843	65.48	5.150
A4	G/PLS-spline	10	³χ_p, <1.98556-S_sCH3>, S_aasC	2	25	0.323	0.906	0.897	106.44	3.022

Mod.	Stat.	Eq.	Descriptors	Q²	r_m²_(LOO)	n_test	R²_pred	r_m²_(test)	r_m²_(overall)

A1	GFA-linear	–	SC-0, S_aaCH, S_dssC, S_dO <SC-3_P-20>,	0.806	0.676	8	0.859	0.839	0.685
A2	GFA-spline	8	³χ_p, <1.79401-S_sCH₃>, S_aasC ³χ_p, (³χ_p)²,	0.877	0.757	8	0.924	0.899	0.777
A2a	GFA-spline	9	<1.79401-S_sCH₃>, S_aasC S_aaCH,	0.814	0.677	8	0.917	0.818	0.711
A3	G/PLS-linear	–	S_aasC, S_dsN, S_sOH, MolRef <6.68154-¹χ >,	0.790	0.771	8	0.879	0.887	0.790
A4	G/PLS-spline	10	³χ_p, <1.98556-S_sCH3>, S_aasC	0.877	0.870	8	0.884	0.812	0.872

Using spatial, electronic, and thermodynamic descriptors

Mod.	Stat.		Descriptors	LVs	n_train.	s	R²	R²_a	F*	PRESS

B1	GFA-linear		MR, Jurs-TASA	–	25	0.443	0.824	0.808	51.55	5.271
B2	G/PLS-linear		MR, Jurs-SASA, Jurs-PPSA-3, Jurs-TASA	3	25	0.434	0.839	0.816	36.41	5.516
B3	G/PLS-spline		<55.0428-MR>, Jurs-PNSA-2, <121.354-Jurs-WNSA-1>, Jurs-WPSA-3, <472.813-Jurs-TASA>	3	25	0.397	0.865	0.846	44.88	4.901

Mod.	Stat.		Descriptors	Q²	r_m²_(LOO)	n_test	R²_pred	r_m²_(test)	r_m²_(overall)

B1	GFA-linear		MR, Jurs-TASA	0.785	0.645	8	0.754	0.683	0.639
B2	G/PLS-linear		MR, Jurs-SASA, Jurs-PPSA-3, Jurs-TASA <55.0428-MR>, Jurs-PNSA-	0.775	0.754	8	0.773	0.661	0.775
B3	G/PLS-2, spline		<121.354-Jurs-WNSA-1>, Jurs-WPSA-3, <472.813-Jurs-TASA>	0.800	0.787	8	0.678	0.525	0.761

Using combined descriptors

Mod.	Stat.		Descriptors	LVs	n_train.	s	R²	R²_a	F*	PRESS

C1	GFA-linear		³χ_c^v, Zagreb, S_aaCH, Jurs-RPSA	–	25	0.284	0.935	0.921	71.35	2.237
C2	GFA-spline		<4.19273-¹χ^v >, S_aasC, <1.83917-S_sCH₃>, RadOfGyration	–	25	0.304	0.925	0.910	61.31	2.756
C3	G/PLS-linear		⁰χ^v, S_aaCH, S_aasC, S_sOH, Jurs-TASA, HOMO <1.78363-S_sCH₃>,	2	25	0.362	0.883	0.872	82.95	5.205
C4	G/PLS-spline		<S_aasC-1.50199>, <5.22431-RadOfGyration>, <133.005-Jurs-WPSA-2>	1	25	0.294	0.919	0.915	260.7	3.064

Mod.	Stat.		Descriptors	Q²	r_m²_(LOO)	n_test	R²_pred	r_m²_(test)	r_m²_(overall)

C1	GFA-linear		³χ_c^v, Zagreb, S_aaCH, Jurs-RPSA	0.909	0.808	8	0.894	0.834	0.822
C2	GFA-spline		<4.19273-¹χ^v >, S_aasC, <1.83917-S_sCH₃>, RadOfGyration	0.888	0.768	8	0.892	0.826	0.791
C3	G/PLS-linear		⁰χ^v, S_aaCH, S_aasC, S_sOH, Jurs-TASA, HOMO <1.78363-S_sCH₃>,	0.788	0.758	8	0.880	0.800	0.785
C4	G/PLS-spline		<S_aasC-1.50199>, <5.22431-RadOfGyration>,<133.005-Jurs-WPSA-2>	0.875	0.848	8	0.800	0.737	0.829

Critical values of F distribution (two-tailed) at 98% significance level: F_{4, 20} = 4.431, F_{2, 22} = 5.719, F_{3, 21}= 4.874, F_{1, 23} = 7.881