candidates from a relatively smaller set of compounds, quite exhaustive at the same
time within given limits, activity linked and rationally guided too may help drug
discovery more effectively.
Among the current drug discovery methods, data modelling and quantitative–
qualitative prediction of activity [2–4], use of molecular docking methods and
scoring functions for virtual high-throughput screening (vHTS) [5] and 3D quantitative structure–activity relationship (QSAR) studies [6] are some of the most used
ones. At the same time, combinatorial generation of chemical compounds is also
carried out since it increases the possibility of finding novel drug molecules from a
large number of chemically diverse compounds generated particularly for the need of
making scaffold hopping [7]. It also provides the opportunity to search for compounds having diverse structural characteristics which in turn may help decipher the
role of molecular components which may be responsible for the biological activities
of new drug molecules, particularly in situations where novel therapeutic candidates
are sought for to handle the challenges arising out of drug resistance problem [8].
So far generating molecular structures are concerned, molecular topology-based
approaches are in use for generating and designing molecular structures [9, 10] and
graph theory [11] and graph theoretical methods [12] have been suitably used for
doing that. However, in general these methods are used for generating structures
combinatorially [10] with no connection to their biological activities and a separate
method has to be used for the prediction of molecular properties and activities. It
appears, therefore, that a method that generates a large number of compounds
combinatorially and gets linked to their activities at the same time may be more
efficient in designing and discovering novel drug molecules. In particular, topological molecular descriptors [2] can be useful in this regard. Moreover, if this is
done using a single molecular (structural/substructural) descriptor, the process may
also be looked upon from inverse QSAR (iQSAR) point of view [13] since the basic
idea of doing iQSAR studies is to get molecular structures back from molecular
descriptor which has been used for activity prediction. In this context, it seems
reasonable to explore whether a method can be developed that is integrated in such
a way that it can be used for generating structures combinatorially that would have
molecules of diverse scaffold from a single molecular topological descriptor , can
be used for predicting molecular properties/activities and can be used for compound
prioritization and screening to help discover potential drug candidates.
So, the first question that may be asked in developing such an integrated method
is: Can we have a method such that structures can be generated combinatorially
from structural or substructural information that is already related to activity? In this
regard, there are two primary aspects in designing potential bioactive compounds
from activity-related substructural information—(1) identification of activity-related
vertices using a suitable method; (2) a method that can be used for structure
generation using topological information associated with such vertices. One of the
most useful activity-related substructure identification method was proposed by
Klopman [14] where molecular fragments of different length are identified from
active and inactive compounds, and the fragments are weighed on the basis of the
number of fragments obtained from active and inactive compounds using a suitable
Combinatorial Drug Discovery from Activity-Related Substructure …
73
time within given limits, activity linked and rationally guided too may help drug
discovery more effectively.
Among the current drug discovery methods, data modelling and quantitative–
qualitative prediction of activity [2–4], use of molecular docking methods and
scoring functions for virtual high-throughput screening (vHTS) [5] and 3D quantitative structure–activity relationship (QSAR) studies [6] are some of the most used
ones. At the same time, combinatorial generation of chemical compounds is also
carried out since it increases the possibility of finding novel drug molecules from a
large number of chemically diverse compounds generated particularly for the need of
making scaffold hopping [7]. It also provides the opportunity to search for compounds having diverse structural characteristics which in turn may help decipher the
role of molecular components which may be responsible for the biological activities
of new drug molecules, particularly in situations where novel therapeutic candidates
are sought for to handle the challenges arising out of drug resistance problem [8].
So far generating molecular structures are concerned, molecular topology-based
approaches are in use for generating and designing molecular structures [9, 10] and
graph theory [11] and graph theoretical methods [12] have been suitably used for
doing that. However, in general these methods are used for generating structures
combinatorially [10] with no connection to their biological activities and a separate
method has to be used for the prediction of molecular properties and activities. It
appears, therefore, that a method that generates a large number of compounds
combinatorially and gets linked to their activities at the same time may be more
efficient in designing and discovering novel drug molecules. In particular, topological molecular descriptors [2] can be useful in this regard. Moreover, if this is
done using a single molecular (structural/substructural) descriptor, the process may
also be looked upon from inverse QSAR (iQSAR) point of view [13] since the basic
idea of doing iQSAR studies is to get molecular structures back from molecular
descriptor which has been used for activity prediction. In this context, it seems
reasonable to explore whether a method can be developed that is integrated in such
a way that it can be used for generating structures combinatorially that would have
molecules of diverse scaffold from a single molecular topological descriptor , can
be used for predicting molecular properties/activities and can be used for compound
prioritization and screening to help discover potential drug candidates.
So, the first question that may be asked in developing such an integrated method
is: Can we have a method such that structures can be generated combinatorially
from structural or substructural information that is already related to activity? In this
regard, there are two primary aspects in designing potential bioactive compounds
from activity-related substructural information—(1) identification of activity-related
vertices using a suitable method; (2) a method that can be used for structure
generation using topological information associated with such vertices. One of the
most useful activity-related substructure identification method was proposed by
Klopman [14] where molecular fragments of different length are identified from
active and inactive compounds, and the fragments are weighed on the basis of the
number of fragments obtained from active and inactive compounds using a suitable
Combinatorial Drug Discovery from Activity-Related Substructure …
73