Lipophilicity of Aminopyridazinone Regioisomers
J. Agric. Food Chem., Vol. 51, No. 18, 2003 5265
1
2
373, 1234, 1065, 882 cm-1; 1H NMR (deuteriochloroform) δ 3.32 (q,
H, NCH , J ) 5.6 Hz), 3.76 (s, 3H, NCH ), 3.80-4.00 (m, 3H, OH
), 6.02 (d, 1H, 5-H), 6.53 (s, 1H, NH), 7.58 (d, 1H, 6-H, J
(pH 7.4) mutually saturated with each other at room temperature. The
two partitioning phases (w vs o) were used in ratios of 50:1 (compounds
2b,c, 3b,c, and 2e, 2g,h, 3g,h), 25:1 (compound 2a, 3a), 10:2
(compound 3f), 10:1 (compound 2f, 3d,e), 5:2 (compound 6), and 1:1
(compounds 2d, 4, 5, 7). The starting amounts of the compounds were
in the range of 2-20 mg dissolved in S o¨ rensen buffer (50-220 mL),
2
3
and OCH
2
13
)
5.2 Hz); C NMR (deuteriochloroform) δ 40.2 (NCH
0.2 (OCH ), 97.6 (C-5), 138.8 (C-6), 144.2 (C-4), 157.1 (C-3). Anal.
Calcd for C : C, 49.70; H, 6.55; N, 24.84. Found: C, 49.64;
H, 6.39; N, 24.84.
Compound 6: white crystals (R
34%); mp 133-134 °C; IR (ν) 3284, 2926, 1611, 1522, 1445, 1402,
3 2
), 45.3 (NCH ),
6
2
7
11 3 2
H N O
which was made from an aqueous 0.067 M Na
2
HPO
4
2
(‚2H O) and from
) 0.16 toluene/methanol 8:2; yield
an aqueous solution of 0.067 M KH PO . In the experiments the tubes
2
4
f
)
were shaken mechanically in a shaking water bath for 1 h at 25 °C and
then centrifuged for 10 min at 2000 rpm. The concentration of the
compounds in the buffer was determined spectrophotometrically in
4-12 parallel measurements at the following wavelengths: 276 nm
(compound 3d), 284 nm (compound 7), 291 nm (compound 3a), 294
nm (compound 3e), 296 nm (compounds 4-6), 298 nm (compounds
3h), 301 nm (compound 2a), 304 nm (compounds 2d, 3g), 306 nm
(compound 3f), 318 nm (compound 3b), 320 nm (compounds 3c), 337
nm (compound 2b), 338 nm (compounds 2c), 340 nm (compound 2e,f),
344 nm (compounds 2g), and 348 nm (compounds 2h).
317, 1270, 1062 cm- ; H NMR (deuteriochloroform) δ 3.39 (t, 2H,
NCH ), 3.51 (qua, 2H, OCH ), 3.57 (s, 3H, NCH ), 4.86 (t, 1H OH, J
5.4 Hz), 6.49 (s, 1H, NH, J ) 6.2 Hz), 7.87 (s, 1H, 6-H); C NMR
deuteriochloroform) δ 39.5 (NCH ), 44.5 (NCH ), 60.5 (OCH ), 104.2
C-4), 126.7 (C-6), 144.9 (C-5), 156.7 (C-3). Anal. Calcd for C
: C, 41.29; H, 4.95; N, 20.64. Found: C, 41.05; H, 4.92; N, 20.42.
Compound 7: white crystals (R ) 0.12 toluene/methanol 8:2; yield
27%); mp 143-144 °C; IR (ν) 3252, 2935, 1630, 1535, 1459, 1335,
1 1
1
2
2
3
13
)
(
(
3
2
2
7 10
H -
3 2
CIN O
f
)
-1 1
1
289, 1073 cm ; H NMR (deuteriochloroform) δ 3.05 (qa, 2H, NCH
2
),
3
.52 (qua, 2H, OCH ), 3.45 (s, 3H, NCH ), 4.81 (t, 1H OH, J ) 5.4
2
3
Computational Methods. PM3 and DFT calculations were carried
out by using the Spartan program package [Spartan SGI version 5.1.1,
Wavefunction Inc., 1998, on a Silicon Graphics (INDY R4400)
computer]. For calculation of Bird’s indices (20) the optimized
structures obtained at the PM3 level were used for full-geometry
optimization with the LSDA/pBP86/DN/DFT model (21).
Hz), 5.48 (d, 1H, 4-H), 6.94 (t, 1H, NH, J ) 5.2 Hz), 7.49 (d, 1H,
1
3
6
(
3
-H, J ) 6.2 Hz); C NMR (deuteriochloroform) δ 38.1 (NCH
NCH ), 58.7 (OCH ), 94.0 (C-4), 130.8 (C-6), 149.2 (C-5), 160.9 (C-
). Anal. Calcd for C : C, 49.70; H, 6.55; N, 24.84. Found:
3
), 44.4
2
2
7 11 3 2
H N O
C, 49.20; H, 6.52; N, 24.74.
Crystal Structures of 2c, 3c and 2e, 3e. Single crystals of 2c, 3c
and 2e, 3e were mounted on glass fibers and transferred to the
diffractometer (Rigaku RAXIS-IIc imaging plate detector, Rigaku RU-
KOWWIN calculations were performed in two ways by using the
KOWWIN program (version 1.54, Syracuse Research Corp., Syracuse,
NY, 1997). [For an application, see: Meylan, W. M.; Howard, P. H.
Atom/fragment contribution method for estimating octanol-water
partition coefficient. J. Pharm. Sci. 1995, 84, 83-92.] A priori
calculations were done by using the molecular structures or by operating
with the EVA option, and predictions are made by considering the
experimental log P value of one member of the series.
2
00HB X-ray generator, graphite monochromated Mo KR radiation)
for data collection. All data were collected at 293 K. Crystal data are
shown below with estimated standard deviation in the final digits in
parentheses. 2c: C14
3 2
H16ClN O , M ) 293.75, orthorhombic, space
group Pbcn, a ) 1.6195(6) nm, b ) 0.7074(3) nm, c ) 2.4167(8) nm,
3
-1
V ) 2.769(2) nm , Z ) 8, D
c
) 1.409 g cm , µ (Mo-KR) ) 0.281
mm , 1756 unique reflections. 3c: , M ) 293.75,
triclinic, space group P-1, a ) 0.9082(3) nm, b ) 0.9577(9) nm, c )
3DNET Calculations (3DNET 4W, Version Beta 1.1.14, Vichem
Chemie GmbH, Budapest, Hungary, 2001). [For an application, see:
K o¨ vesdi, I.; Dominguez-Rodriguez, M. F.; Orfi, L.; N a´ ray-Szab o´ , G.;
Varr o´ , A.; Papp, G. J.; M a´ tyus, P. Application of neural networks in
structure-activity relationships. Med. Res. ReV. 1999, 19, 249-269.]
In the present study, geometric optimization for all 20 molecules was
carried out by the MM2 program as implemented in HyperChem
(HyperChem Professional, version 7.0 for Windows, Hypercube Inc.,
2002) to obtain 3D structures. Then, for all molecules, 1437 structural
descriptors were calculated with the Dragon program (version 2.1, by
R. Todeschini, V. Consonni, and M. Pavan, 2002; available for
download at http://www.disat.unimib.it/chm/). Of the descriptors, 1119
nonconstant ones were used in the subsequent computations. Next,
-
1
14 3 2
C H16ClN O
0
0
1
.8994(9) nm, R ) 116.98(4)°, â ) 99.98(9)°, γ ) 82.47(9)°, V )
3
-1
-1
.685 (1) nm , Z ) 2, D
c
) 1.414 g cm , µ (Mo-KR) ) 0.284 mm
,
295 unique reflections. 2e: C
7
H
9
ClN , M ) 232.63, monoclinic,
4 3
O
space group Pc, a ) 0.4027(5) nm, b ) 0.879(1) nm, c ) 1.386 (2)
nm, â ) 95.8(1)°, V ) 0.488(1) nm , Z ) 2, D
3
-1
c
) 1.583 g cm , µ
-
1
(
Mo-KR) ) 0.385 mm , 1045 unique reflections. 3e: C ClN O ,
7
H
9
4 3
M ) 232.63, monoclinic, space group C2/c, a ) 1.139(7) nm, b )
0
D
3
.994(6) nm, c ) 1.717(3) nm, â ) 97.4(4)°, V ) 1.93(2) nm , Z ) 8,
-
1
-1
c
) 1.602 g cm , µ (Mo-KR) ) 0.390 mm , 1986 unique
reflections. For all data sets data processing was carried out using the
software supplied with the diffractometer. Structure solutions with direct
methods were carried out with the teXsan Crystal Structure Analysis
Package (Molecular Structure Co., Houston, TX, 1992). The refinements
were carried out using the SHELXL-93 program (Sheldrick, G. M.
University of Gottingen, Germany, 1993) with full matrix least-squares
2
contingency filtering with the XY linear correlation coefficient R with
the threshold of 0.4 yielded 423 descriptors for the subsequent variable
subset selection. This large number of well-correlating descriptors
indicated the possible existence of a good linear model. Therefore, a
partial least-squares (PLS) analysis and quantitative structure-activity
relationships (QSAR) model was built for the present study. In the
Euclidean space of the 423 descriptors 6 uniformly distributed molecules
were selected for external validation. With the remaining 14 molecules,
a repeated evaluation ensemble method was used, in which 64 random
selections of 7 molecules from 14 were generated. In each of these
selections, 7 molecules were used to calculate the PLS model with a
given number of components for a given subset of variables; the model
was evaluated by the performance on the remaining set of 7 molecules.
This process was repeated for all 64 random selections.
The ensemble average of the standard error of predictions over these
64 randomly selected evaluation sets was next chosen to be optimized
during the variable subset selection. For each subset of variables, the
number of PLS components was optimized to give the lowest ensemble-
averaged standard error of predictions.
We used a genetic algorithm for the administrator of the variable
subset selection, in which the negative of the ensemble-averaged
standard error of predictions was taken as the fitness value of a given
model. One hundred and thirty-nine generations of 16 models with 16
offspring resulted in the best model with 9 descriptors and 4 PLS
components. This model was the best fitted one of the last 24
generations when the algorithm stopped. The final external validation
2
method on F . All non-hydrogen atoms were refined anisotropically.
Hydrogen atoms were generated on the basis of geometric evidence,
and their positions were refined by the riding model. Hydrogen atoms
of the OH groups were generated to form the best hydrogen bonds.
For 2c two disordered OH hydrogen atoms were generated with
occupancies of 0.5. Final R indices for 2c are R ) 0.1001 and R
w
)
0
.2283 (all reflections) and R ) 0.0740 and R
w
) 0.2020 [I > 2σ(I)];
) 0.1699 (all reflections) and R )
those for 3c are R ) 0.0689 and R
w
0
0
0
.0566 and R
.0992 and R
.1508 [I > 2σ(I)]; those for 3e are R ) 0.1235 and R
) 0.1963 [I > 2σ(I)]. The maximal
residual peak and hole in the final difference electron density map are
w
) 0.1571 [I > 2σ(I)]. Final R indices for 2e are R )
) 0.2448 (all reflections) and R ) 0.0636 and R
) 0.2736 (all
w
w
)
w
reflections) and R ) 0.0674 and R
w
-
3
-3
2
-
57 and -382 e nm for 2c, 267 and -287 e nm for 3c, 182 and
-
3
-3
191 e nm for 2e, and 270 and -380 e nm for 3e.
Full lists of atomic coordinates, bond lengths, angles, and thermal
parameters have been deposited at the Cambridge Crystallographic Data
Centre under deposition codes (2c) 205847, (3c) 205848, (2e) 205849,
and (3e) 205850.
Experimental Determination of Partition Coefficient by Shake
Flask Method. The solvents used were octanol and S o¨ rensen buffer