Consensus features of CP-MLR and GA in modeling HIV-1 RT inhibitory activity of 4-benzyl/benzoylpyridin-2-one analogues

Figures & data

Figure 1.  (A) 7-Chloro-1-(2,6-difluorophenyl)-1H,3H-thiazolo[3,4-a] benzimidazole (7-Cl-TBZ), (B) 2,3-diaryl-thiazolidin-4-ones, and (C) 4-benzyl/benzoyl-pyridin-2-ones benzylpyridinones.

Table 1.  Observed and predicted HIV-1 RT inhibitory activity 4-benzyl/benzoylpyridin-2-ones (Figure 1C).

S. No.*	R1	R2	R3	R4	R5	−logIC50
						Obs	Pred^†
							Eq. (9)	Eq. (11)	PLS	ANN
1^#	Me	Et	Me	Me	3,5-diMe	8.10	7.66	8.12	8.31	8.07
2	Me	Et	Me	Me	3,5-diMe	8.40	7.88	8.14	8.48	8.04
3	Me	Me	Me	Me	3,5-diMe	8.30	7.86	8.07	8.30	8.06
4	Et	Me	Me	Me	3,5-diMe	8.00	7.90	7.52	8.00	7.80
5^#	Me	i-Bu	Me	Me	3,5-diMe	7.30	6.85	7.52	7.56	8.09
6^#	i-Pr	Me	Me	Me	3,5-diMe	6.10	7.62	6.99	7.05	6.22
7^§	Me	n-Pr	Me	Me	3,5-diMe	7.30	7.27	7.17	7.10	7.70
8^§	H	H	Me	Me	3,5-diMe	5.10	5.55	6.20	6.53	5.22
9	Me	H	Me	Me	3,5-diMe	6.40	6.21	7.99	6.40	6.10
10	-(CH2)4-^‡		Me	Me	3,5-diMe	8.00	7.93	7.38	8.00	8.01
11	Me	Et	Me	Me	3-Me	8.60	8.26	8.60	8.60	8.07
12^§	Me	Me	Me	Me	3-Me	7.80	8.38	8.52	8.64	8.04
13	Et	Me	Me	Me	3-Me	7.00	8.48	8.05	7.00	7.23
14^§	-(CH2)4-^‡		Me	Me	3-Me	7.20	8.37	7.92	7.68	7.83
15	-(CH2)3-^‡		Me	Me	3-Me	7.80	8.07	8.01	7.80	7.67
16	Me	(CH2)2OMe	Me	Me	3-Me	8.40	8.67	8.43	8.40	8.46
17^§	Me	(CH2)3OMe	Me	Me	3-Me	8.10	8.31	8.78	8.81	8.35
18	Me	(CH2)3OMe	Me	Me	3,5-diMe	8.70	7.90	8.38	8.70	8.61
19^#	MeOH	Et	Me	Me	3,5-diMe	9.00	9.13	8.52	9.16	9.07
20	Me	Et	H	CHO	3,5-diMe	7.10	6.21	5.67	7.10	7.10
21^§	Me	Et	H	Me	3,5-diMe	8.00	8.10	8.13	8.32	8.05
22	Me	Et	Me	Et	3,5-diMe	8.10	7.70	8.04	8.10	8.01
23	Me	Et	Me	n-Pr	3,5-diMe	7.80	7.01	7.32	7.80	7.95
24	Me	Et	Me	CH(Me)CH2OMe	3,5-diMe	8.22	8.18	8.05	8.22	8.25
25^#	Me	Et	Me	(CH2)3SMe	3,5-diMe	7.60	7.54	7.46	7.42	8.01
26^§	Me	Et	Me	CH2CH2 OMe	3,5-diMe	8.70	8.02	8.11	8.23	8.44
27	Me	Et	Me	(CH2)5OH	3,5-diMe	8.40	8.30	7.40	8.40	8.23
28	Me	Et	H	COMe	3,5-diMe	6.40	6.38	6.25	6.40	6.38
29	Me	Et	H	COEt	3,5-diMe	5.40	5.83	5.66	5.40	5.09
30	Me	Et	H	COC3H7	3,5-diMe	4.00	5.50	5.18	4.00	4.20
31^#	Me	Et	H	Et	3,5-diMe	7.80	7.71	7.69	7.66	7.94
32	Me	Et	H	n-Pr	3,5-diMe	7.70	7.19	7.08	7.70	7.86
33^§	Me	Et	H	n-Bu	3,5-diMe	6.90	6.82	6.31	6.42	6.30
34	Me	Et	Et	Et	3,5-diMe	7.80	7.57	7.88	7.80	7.84
35	Me	Et	n-Bu	n-Bu	3,5-diMe	4.30	5.98	5.73	4.30	4.34
36	Me	Et	H	Benzyl	3,5-diMe	6.60	6.27	6.06	6.60	6.51
37	Me	Et	Benzyl	Benzyl	3,5-diMe	4.00	5.15	5.34	4.00	4.04
38^#,‡	Me	Et			3,5-diMe	6.80	8.19	8.38	8.20	7.67
39^¶	Me	Et			3,5-diMe	6.20	4.44	4.85	6.20	6.19
40^§,║	Me	Et			3,5-diMe	7.90	7.31	6.95	6.92	7.80
41	Me	Et	Me	(CH2)2OH	3-Me	8.30	8.71	9.22	8.30	8.36
42	Me	Et	Me	(CH2)3OH	3-Me	8.52	8.70	8.61	8.52	8.80
43^#	Me	Et	Me	(CH2)5OH	3-Me	8.00	8.79	7.83	7.96	8.43
44	Me	Et	Me	(CH2)2OMe	3-Me	9.00	8.43	8.47	9.00	8.86
45	Me	Et	Me	(CH2)2OEt	3-Me	7.89	8.40	8.76	7.89	8.20
46	Me	Et	Me	CH2CN	3-Me	8.40	8.44	8.94	8.40	8.10
47	Me	Et	Me	(CH2)2CN	3-Me	7.80	8.34	8.73	7.80	8.08
48^#	Me	Et	Me	(CH2)3CN	3-Me	8.30	8.24	9.08	8.51	8.11
49^#	Me	Et	H	NH-CS-NHEt	3-Me	4.60	5.54	5.92	5.31	5.19
50	Me	Et	H	NH-CS-NHPh	3-Me	5.50	5.86	5.18	5.50	5.43
51	Me	Et	H	NH-CS-NHCOPh	3-Me	5.20	3.39	5.23	5.20	5.40
52^§	Me	Et	H	NH-CS-NH2	3-Me	6.50	5.87	6.52	5.69	6.69
53	Me	Et	Me	(CH2)2OCH3	3-Me	9.00	8.34	8.49	9.00	8.96
54	Me	Et	Me	Et	3-CHCHCN	9.00	8.91	8.70	9.00	8.09
55	Me	Et	Me	(CH2)2OCH3	3-CHCHCN	9.00	9.23	8.64	9.00	8.83

*Compounds 2, 11–19 and 53 are 4-benzoylpyridin-2-ones and rest of them are 4-benzylpyridin-2-ones.

^†Equation 9 is from CP-MLR; Equation 11 is from GA; PLS predicted activities are from PLS model of 11 common descriptors of CPMLR and GA (Table 4); ANN predicted activities are from BP-ANN model developed using 5 input features (Table 5).

^#,§Test set compounds for CP-MLR, GA and PLS models; in case of BP-ANN model, compounds identified with “#” and “§”, respectively correspond to validation and test sets.

^‡Morpholin-4-yl in place of -NR₃R₄ (Figure 1C).

^¶Piperidin-1yl in place of -NR₃R₄ (Figure 1C).

^∥Pyrrol-1-yl in place of -NR₃R₄ (Figure 1C).

Table 2.  Information content of descriptors identified from CP-MLR and GA approaches.

SNo	Descriptor	Class*	FS^†		Information content^#
1	nDB	Const	C	G	Number of double bonds
2	X2A	Topo	C		Average connectivity index chi-2
3	PW3		C		Ratio of path/walk 3 - Randic shape index
4	PW4		C		Ratio of path/walk 4 - Randic shape index
5	BIC5			G	Bond information content of neighborhood symmetry of 5 order
6	VRA2		C	G	Average Randic-type eigenvector-based index from adjacency matrix
7	T(N..O)		C		Sum of topological distances between N & O
8	T(O..O)		C	G	Sum of topological distances between O & O
9	BEHp6	BCUT		G	6^th Highest eigenvalue of Burden matrix weighted by atomic polarizabilities
10	BELp4		C	G	4^th Lowest eigenvalue of Burden matrix weighted by atomic polarizabilities
11	GGI4	Galvez	C	G	Topological charge index of order 4
12	GGI6		C	G	Topological charge index of order 6
13	JGI2		C		Mean topological charge index of order 2
14	JGI4		C	G	Mean topological charge index of order 4
15	JGI6		C		Mean topological charge index of order 6
16	MATS8e	2D-Auto	C	G	Moran autocorrelation of lag 8 weighted by atomic Sanderson electronegativities
17	GATS4e			G	Geary autocorrelation of lag 4 weighted by atomic Sanderson electronegativities
18	GATS8e		C		Geary autocorrelation of lag 8 weighted by atomic Sanderson electronegativities
19	GATS4p		C	G	Geary autocorrelation of lag 4 weighted by atomic polarizabilities
20	H-046	ACF	C	G	H attached to C0(sp3) with no X attached to next C
21	LogP	PROP	C	G	Octanol-water partition coefficient

*Descriptor class: Const, constitutional; Topo, topological; BCUT, BCUT; Galvez, Galvez topological charge indices; 2D-AUTO: 2D autocorrelations; ACF, atom-centered fragments; PROP; proprieties.

^†Feature selection approach involved in descriptor identification, C for CP-MLR and G for GA.

^#See Ref. (17).

Table 3.  Five parameter models for HIV-1 RT inhibitory activity of 4-benzyl/benzoylpyridin-2-ones (Table 1) from CP-MLR and GA along with statistics.

S. No.	Model*	FS^†	Normal stat.^#			Cross-valid. Stat				Test r^2t	Y-Rand r^2Yrand (max)
			R^2	s	F	Q^2	Q^2G5	SPRESS	SDEP
1	1, 6, 10, 14, 20	C;G	0.848	0.617	32.31	0.752	0.786	0.788	0.717	0.512	0.109 (0.351)
2	1, 6, 14, 16, 21	C	0.846	0.620	31.98	0.731	0.738	0.820	0.747	0.532	0.124 (0.331)
3	1, 6, 14, 18, 21	C	0.841	0.563	30.63	0.717	0.726	0.841	0.765	0.518	0.149 (0.387)
4	1, 3, 6, 14, 21	C	0.839	0.633	30.40	0.730	0.735	0.822	0.748	0.576	0.132 (0.383)
5	1, 6, 14, 15, 21	C	0.834	0.644	29.23	0.712	0.707	0.849	0.773	0.548	0.133 (0.383)
6	1, 8, 11, 12, 19	C;G	0.834	0.646	28.92	0.730	0.741	0.822	0.749	0.533	0.078 (0.364)
7	1, 4, 7, 13, 21	C	0.832	0.650	28.57	0.743	0.696	0.801	0.729	0.512	0.082 (0.289)
8^§	1, 2, 6, 14, 21	C	0.832	0.650	28.54	0.700	0.732	0.867	0.789	0.605	0.113 (0.325)
9	1, 5, 14, 16, 17	G	0.830	0.653	28.25	0.753	0.725	0.786	0.715	0.513	0.135 (0.393)
10^‡	1, 11, 16, 19, 21	G	0.839	0.635	30.17	0.735	0.751	0.815	0.742	0.564	0.137 (0.375)
11	1, 6, 10, 14, 20	C;G	0.848	0.617	32.31	0.752	0.774	0.788	0.717	0.512	0.143 (0.399)
12	1, 8, 11, 12, 19	C;G	0.834	0.646	28.92	0.730	0.756	0.822	0.748	0.533	0.132 (0.323)

*The number corresponds to the descriptor serial number given in (Table 2).

^†Feature selection approach involved in model formation, C for CP-MLR and G for GA.

^#In all the models, the number of observations is 35; all statistical abbreviations represent their standard meaning.

^§Equation 9 in discussion.

^‡Equation 11 in discussion.

Table 4.  MLR-like PLS models from the combined as well as common descriptors of CP-MLR and GA approaches (Table 2) for the HIV-1 RT inhibitory activity (−logIC₅₀) of 4-benzyl/benzoylpyridin-2-ones (Table 1).

S. No.	Descriptor	MLR-like coeff (f.c)*
		(CP-MLR) ∪ (GA)^†	(CP-MLR) ∩ (GA)^#
1	nDB	−1.164 (−0.180)	−1.612 (−0.257)
2	X2A	−6.139 (−0.007)
3	PW3	0.187 (0.0005)
4	PW4	9.942 (0.019)
5	BIC5	−0.022 (0.0002)
6	VRA2	0.008 (0.009)	0.025 (0.030)
7	T(N..O)	0.022 (0.051)
8	T(O..O)	0.050 (0.076)	0.071 (0.109)
9	BEHp6	2.687 (0.053)
10	BELp4	−2.607 (−0.063)	−3.712 (−0.093)
11	GGI4	0.639 (0.044)	0.990 (0.070)
12	GGI6	−0.980 (−0.037)	−1.541 (−0.060)
13	JGI2	2.209 (0.007)
14	JGI4	17.480 (0.059)	24.782 (0.086)
15	JGI6	−8.924 (−0.011)
16	MATS8e	2.252 (0.050)	4.338 (0.099)
17	GATS4e	−2.611 (−0.087)
18	GATS8e	−1.155 (−0.088)
19	GATS4p	2.783 (0.056)	4.267 (0.088)
20	H-046	−0.023 (−0.024)	−0.013 (−0.015)
21	LogP	−0.215 (−0.078)	−0.255 (−0.095)
	Constant	3.402	7.986
	Statistics
	N	35	35
	r^2	0.890	0.888
	S	0.515	0.520
	F	60.91	59.46
	Q^2	0.848	0.834
	Q^2G5	0.862	0.843
	SPRESS	0.605	0.633
	SEDP	0.564	0.586
	r^2t	0.569	0.607
	r^2Yrand(max)	0.116 (0.383)	0.080 (0.345)

*Coefficients of MLR-like PLS equation in terms of descriptors for their original values; f.c is fraction contribution of regression coefficient, computed from the normalized regression coefficients obtained from the autoscaled (zero mean and unit SD) data.

^†Combined descriptors of CP-MLR and GA.

^#Descriptors common to CP-MLR and GA.

Figure 2.  Plots of fraction contribution of MLR-like PLS coefficients (normalized) of the combined and common descriptors of CP-MLR and GA for the HIV-1 RT inhibitory activity of 4-benzyl/benzoyl-pyridin-2-ones; the numbers on the bars refer to the descriptors’ numbers (Table 2).

Table 5.  ANN Architecture and goodness of fit of HIV-1 RT inhibitory activity of 4-benzyl/benzoylpyridin-2-ones (Table 1) in training, validation and test sets with five most significant features from PLS in BP-ANN model*.

ANN architecture and parameters
Layer	Nodes	Training parameters
Input	5 + 1(bias)	Learning rate (μ)		0.611
Hidden	4	Momentum (α)		0.661
Output	1	Transfer function		Sigmoid
		Optimization algorithm		Levenberg–Marquardt
		Iterations (λ)		19
ANN statistics of HIV-1 RT model
Sample	Sample size	r^2	RMSEP		RSEP (%)	MAE (%)
Train	35	0.931	0.377		4.999	7.907
Valid	10	0.925	0.465		6.234	19.083
Test	10	0.890	0.342		4.609	16.875

*ANN input features are nDB, T(O..O), BELp4, MATS8e and LogP.

Figure 3.  The plots of observed versus predicted activities of 4-benzyl/benzoyl-pyridin-2-ones’ training (open diamonds), validation (open citrcles) and test (open triangles) sets from BP-ANN. The solid line indicates the best fit. The dashed line passing through the origin, making an angle of 45° with the axis, bisects the plot area.