471
14 Quantitative Structure-Property Relationship Analysis …
be represented as a system of different simplexes (tetratomic fragments of fixed
composition, structure, chirality and symmetry) (Fig. 14.1). Atoms in a simplex can
be differentiated on the base of different characteristics:
1. Atoms individuality expressed through the nature (e.g., nitrogen or carbon) or
more detailed type of atoms (C-sp
3
or C-sp
2
);
2. Partial atom charge that reflect their electrostatic properties;
3. Lipophilicity of atoms that reflects its hydrophobic/hydrophilic properties;
4. Atomic refraction that partially reflects the ability of atoms to dispersion
interactions;
5. A mark that characterizes the atom as a possible hydrogen donor or acceptor during H-bond formation (A: hydrogen acceptor in H-bond; D: hydrogen donor in
H-bond, I: indifferent atom).
For atom characteristics, that have real values (for example, charge, lipophilicity
and refraction) the division of continuous values into definite discrete groups is carried out at the preliminary stage. The number of groups (G) is a tuning parameter
and can be varied (usually G = 3–7).
The use of sundry variants of simplex vertexes (atoms) differentiation represents
an important part of SiRMS we consider that specification of atoms by their nature
alone (this actually reflects atom identity, for example, C, N or O), which is realized
in many QSAR methods limits the possibilities of active fragments selection. For
example, if the –NH– group has been selected as the determining activity fragment
and the ability of H-bond formation is a factor determining its properties, then we
will miss donors of H-bonds, such as OH-groups. The use of atom differentiation
by donor/acceptor of H-bond allows one to avoid the situation illustrated above.
One can make analogical examples for other atom properties (lipophilicity, partial
charge and refraction, for example). Different types of simplexes are generated depending on the level of detail of molecular structure.
1D models. 1D simplex is a combination of any four atoms contained in the
molecule (Fig. 14.1). The simplex descriptor (SD) at this level is a number of quadruples of atoms of the definite composition. 1D simplexes were not used in our
studies.
2D models. The connectivity of atoms in simplex, atom type and bond nature
(single, double, triple, or aromatic) have been considered. Thus, the SD at the 2D
level is a number of simplexes of fixed composition and topology (Fig. 14.1). Other
structural parameters corresponding to molecular fragments of different size, can
be used on 1 and 2D levels. The use of 1–4 atomic fragments is preferable, because
further extension of the fragment length could increase the probability of model
over-fitting and decrease its predictivity and DA.
3D models. Not only topology, but also the stereochemistry of the molecule is
taken into account at the 3D level. It is possible to differentiate all the simplexes as
right (R), left (L), symmetrical (S) and plane (P) achiral (Fig. 14.2). Stereochemical
configuration of simplexes is defined by a modified Kahn–Ingold–Prelog rule. An
SD at this level is a number of simplexes of fixed composition, topology, chirality
and symmetry (Fig. 14.1).
14 Quantitative Structure-Property Relationship Analysis …
be represented as a system of different simplexes (tetratomic fragments of fixed
composition, structure, chirality and symmetry) (Fig. 14.1). Atoms in a simplex can
be differentiated on the base of different characteristics:
1. Atoms individuality expressed through the nature (e.g., nitrogen or carbon) or
more detailed type of atoms (C-sp
3
or C-sp
2
);
2. Partial atom charge that reflect their electrostatic properties;
3. Lipophilicity of atoms that reflects its hydrophobic/hydrophilic properties;
4. Atomic refraction that partially reflects the ability of atoms to dispersion
interactions;
5. A mark that characterizes the atom as a possible hydrogen donor or acceptor during H-bond formation (A: hydrogen acceptor in H-bond; D: hydrogen donor in
H-bond, I: indifferent atom).
For atom characteristics, that have real values (for example, charge, lipophilicity
and refraction) the division of continuous values into definite discrete groups is carried out at the preliminary stage. The number of groups (G) is a tuning parameter
and can be varied (usually G = 3–7).
The use of sundry variants of simplex vertexes (atoms) differentiation represents
an important part of SiRMS we consider that specification of atoms by their nature
alone (this actually reflects atom identity, for example, C, N or O), which is realized
in many QSAR methods limits the possibilities of active fragments selection. For
example, if the –NH– group has been selected as the determining activity fragment
and the ability of H-bond formation is a factor determining its properties, then we
will miss donors of H-bonds, such as OH-groups. The use of atom differentiation
by donor/acceptor of H-bond allows one to avoid the situation illustrated above.
One can make analogical examples for other atom properties (lipophilicity, partial
charge and refraction, for example). Different types of simplexes are generated depending on the level of detail of molecular structure.
1D models. 1D simplex is a combination of any four atoms contained in the
molecule (Fig. 14.1). The simplex descriptor (SD) at this level is a number of quadruples of atoms of the definite composition. 1D simplexes were not used in our
studies.
2D models. The connectivity of atoms in simplex, atom type and bond nature
(single, double, triple, or aromatic) have been considered. Thus, the SD at the 2D
level is a number of simplexes of fixed composition and topology (Fig. 14.1). Other
structural parameters corresponding to molecular fragments of different size, can
be used on 1 and 2D levels. The use of 1–4 atomic fragments is preferable, because
further extension of the fragment length could increase the probability of model
over-fitting and decrease its predictivity and DA.
3D models. Not only topology, but also the stereochemistry of the molecule is
taken into account at the 3D level. It is possible to differentiate all the simplexes as
right (R), left (L), symmetrical (S) and plane (P) achiral (Fig. 14.2). Stereochemical
configuration of simplexes is defined by a modified Kahn–Ingold–Prelog rule. An
SD at this level is a number of simplexes of fixed composition, topology, chirality
and symmetry (Fig. 14.1).