449
13 Continuous Molecular Fields Approach Applied to Structure-Activity Modeling
this method does not require counterexamples at all and therefore, in distinction
from traditional SVM, can be used in the cases (typical for drug discovery) when
counterexamples are not known or the dataset is highly imbalanced. As distinct
from the traditional similarity search, the suggested method can adapt to complex
structure-activity landscapes and, thus, makes it possible to avoid activity cliffs
[49]. In addition, as compared to similarity search methods based on the use of
fragmental descriptors, the suggested approach implies using the same model to the
sets of compounds belonging to different structural classes.
13.4 The Main Directions of Further Development
of the CMF Approach
In this article, we have considered a particular implementation of the CMF approach aimed at building 3D-QSAR models. This implementation is being actively
developed, and its future version will surely be better than the current one. This
section, however, concerns more global, strategic directions of further development
of the whole CMF approach.
13.4.1 Introduction of Additional Types of Molecular Fields,
Integration with Quantum Chemistry
The CMF approach is not confined to the simplest approximation scheme introduced by Eq. (13.5). Any number of Gaussian functions, (both isotropic, i.e. spherically symmetrical, and non-isotropic) as well as any other set of basic functions
(such as splines, wavelets, etc.) can be used for approximating continuous molecular fields. This provides the ability to work with complex types of molecular fields,
including those derived from the electron density function.
Close integration with quantum chemistry is a promising direction of further
development of the CMF approach. Indeed, wave and electron density functions
are the most natural, comprehensive and accurate way to describe molecules.
Moreover, all possible continuous molecular fields can be generated from the
electron density functions with the help of integral transforms [50]. Especially
promising is the use of conceptual DFT molecular fields [51] based on conceptual DFT [52]. Molecular fields (steric, electrostatic, local softness and LUMO)
derived from the electron density function have already been successfully used
for building 3D-QSAR models in the field of metal complex catalysis using the
traditional grid-based CoMFA approach [53, 54]. We can expect that due to the
ability to approximate all kinds of molecular fields with any degree of accuracy
the CMF approach will be very useful in the development of new catalysts and
supramolecular complexes, which require high accuracy in representing the electronic structure of molecules.
13 Continuous Molecular Fields Approach Applied to Structure-Activity Modeling
this method does not require counterexamples at all and therefore, in distinction
from traditional SVM, can be used in the cases (typical for drug discovery) when
counterexamples are not known or the dataset is highly imbalanced. As distinct
from the traditional similarity search, the suggested method can adapt to complex
structure-activity landscapes and, thus, makes it possible to avoid activity cliffs
[49]. In addition, as compared to similarity search methods based on the use of
fragmental descriptors, the suggested approach implies using the same model to the
sets of compounds belonging to different structural classes.
13.4 The Main Directions of Further Development
of the CMF Approach
In this article, we have considered a particular implementation of the CMF approach aimed at building 3D-QSAR models. This implementation is being actively
developed, and its future version will surely be better than the current one. This
section, however, concerns more global, strategic directions of further development
of the whole CMF approach.
13.4.1 Introduction of Additional Types of Molecular Fields,
Integration with Quantum Chemistry
The CMF approach is not confined to the simplest approximation scheme introduced by Eq. (13.5). Any number of Gaussian functions, (both isotropic, i.e. spherically symmetrical, and non-isotropic) as well as any other set of basic functions
(such as splines, wavelets, etc.) can be used for approximating continuous molecular fields. This provides the ability to work with complex types of molecular fields,
including those derived from the electron density function.
Close integration with quantum chemistry is a promising direction of further
development of the CMF approach. Indeed, wave and electron density functions
are the most natural, comprehensive and accurate way to describe molecules.
Moreover, all possible continuous molecular fields can be generated from the
electron density functions with the help of integral transforms [50]. Especially
promising is the use of conceptual DFT molecular fields [51] based on conceptual DFT [52]. Molecular fields (steric, electrostatic, local softness and LUMO)
derived from the electron density function have already been successfully used
for building 3D-QSAR models in the field of metal complex catalysis using the
traditional grid-based CoMFA approach [53, 54]. We can expect that due to the
ability to approximate all kinds of molecular fields with any degree of accuracy
the CMF approach will be very useful in the development of new catalysts and
supramolecular complexes, which require high accuracy in representing the electronic structure of molecules.