Similarly, if we regress the distances (d) of the chlorine and fluorine substituted
acids (SAA) from acetic acid taken from the data used to generate the plot in
Fig. 3.7 we get the relationship:
pK a ðSAA) % 8:4075 expðÀ0:644 dÞ
ðr
2
¼ 0:996; n ¼ 7Þ
ð3:32Þ
This suggests that the 2-dimensional projection of the 6-dimensional eigenvalue
descriptor set for these molecules retains most of the information of the LDM (and
DM) representations of the molecules. We have also shown [95] that the six
eigenvalues for the molecules in Table 3.4 can be used to build robust models for
multiple physical properties (pK a , LogP, LD50 (oral:rat), Henry’s Law Constants,
melting point, boiling point, vapour pressure and the Atmospheric OH rate
constant).
In another preliminary evaluation of the use of the 6-dimensional eigenvalues
from the LDM as universal descriptors we looked at building a model of the
inhibitive properties of heterocyclic diazoles for acidic iron corrosion [96].
Table 3.5 shows the computed eigenvalues and corrosion inhibitor efficiencies (CIE
%) for a set of diazoles reported by Babic-Samardzija et al. [96].
From these eigenvalues a relatively simple model can be built:
CIE(%Þ ¼ 77:6097 þ 703:1552ðF4 Â F5Þ À 261:6683 Â ðF4Þ
2 À 464:9701ðF5Þ
2
ðr
2
¼ 0:987; q
2
¼ 0:898; n ¼ 8Þ
ð3:33Þ
with both a high correlation coefficient and a high cross-validation score. A plot of
the experimental versus predicted corrosion inhibition efficiencies is shown in
Fig. 3.8.
Table 3.5 Eigenvalues and corrosion inhibitor efficiencies (CIE%) for a set of diazoles (CIE%
taken from Babic-Samardzija et al. [96])
Compound
F1
F2
F3
F4
F5
F6
CIE%
3-amino-1H-isoindole
3.160
2.854
2.306
1.968
1.566
1.487
90.7
Indazole
2.700
2.432
1.974
1.857
1.540
1.385
83.6
Imidazole
2.608
2.230
1.832
1.081
0.785
0.216
83.4
4-bromoimidazole
2.461
2.140
1.598
1.168
0.847
0.517
83.1
4-methylimidazole
3.471
2.254
1.964
1.262
0.890
0.760
82.8
Pyrazole
2.516
2.321
1.782
1.127
0.799
0.204
79.3
4-nitropyrazole
2.406
2.104
1.981
1.107
1.061
0.990
59.4
4-sulfopyrazole
2.405
2.266
1.993
1.617
1.088
1.069
80.3
80
C.F. Matta et al.
acids (SAA) from acetic acid taken from the data used to generate the plot in
Fig. 3.7 we get the relationship:
pK a ðSAA) % 8:4075 expðÀ0:644 dÞ
ðr
2
¼ 0:996; n ¼ 7Þ
ð3:32Þ
This suggests that the 2-dimensional projection of the 6-dimensional eigenvalue
descriptor set for these molecules retains most of the information of the LDM (and
DM) representations of the molecules. We have also shown [95] that the six
eigenvalues for the molecules in Table 3.4 can be used to build robust models for
multiple physical properties (pK a , LogP, LD50 (oral:rat), Henry’s Law Constants,
melting point, boiling point, vapour pressure and the Atmospheric OH rate
constant).
In another preliminary evaluation of the use of the 6-dimensional eigenvalues
from the LDM as universal descriptors we looked at building a model of the
inhibitive properties of heterocyclic diazoles for acidic iron corrosion [96].
Table 3.5 shows the computed eigenvalues and corrosion inhibitor efficiencies (CIE
%) for a set of diazoles reported by Babic-Samardzija et al. [96].
From these eigenvalues a relatively simple model can be built:
CIE(%Þ ¼ 77:6097 þ 703:1552ðF4 Â F5Þ À 261:6683 Â ðF4Þ
2 À 464:9701ðF5Þ
2
ðr
2
¼ 0:987; q
2
¼ 0:898; n ¼ 8Þ
ð3:33Þ
with both a high correlation coefficient and a high cross-validation score. A plot of
the experimental versus predicted corrosion inhibition efficiencies is shown in
Fig. 3.8.
Table 3.5 Eigenvalues and corrosion inhibitor efficiencies (CIE%) for a set of diazoles (CIE%
taken from Babic-Samardzija et al. [96])
Compound
F1
F2
F3
F4
F5
F6
CIE%
3-amino-1H-isoindole
3.160
2.854
2.306
1.968
1.566
1.487
90.7
Indazole
2.700
2.432
1.974
1.857
1.540
1.385
83.6
Imidazole
2.608
2.230
1.832
1.081
0.785
0.216
83.4
4-bromoimidazole
2.461
2.140
1.598
1.168
0.847
0.517
83.1
4-methylimidazole
3.471
2.254
1.964
1.262
0.890
0.760
82.8
Pyrazole
2.516
2.321
1.782
1.127
0.799
0.204
79.3
4-nitropyrazole
2.406
2.104
1.981
1.107
1.061
0.990
59.4
4-sulfopyrazole
2.405
2.266
1.993
1.617
1.088
1.069
80.3
80
C.F. Matta et al.