Structure Generation for Nucleoside Analogous
As done for the barbiturates, structure generation from various starting points, i.e.
compound no., atom no., was carried out for the NA series of compounds too. In
doing that, activity-related vertices have been picked up from the strong ranges in
the ordering of D
−4 index values for the vertices (atoms) of the H-suppressed graphs
of these compounds. It has been found that a few carbon skeletons resembling the
structure of other active compounds than the ones from where the activity-related
vertices and the corresponding distance distribution values are taken have been
generated.
For the purpose of illustration, the structure of the compound no. 6 and the
generated structure which corresponds to compound no. 8 are shown in Fig. 9. It
can be seen that in this case too, the algorithm is able to generate a structure with
significantly different scaffold than the starting compound and has a higher MPS
value (MPS = 65) too compared to that (MPS = 8) of the starting structure indicating that this generated structure has the potential of being highly active and
therefore may be picked/prioritized for further studies. In fact, compound no. 8 is a
potent anti-HIV drug—Didanosine. Therefore, the method may be regarded as a
useful tool for generating, prioritizing and discovering potent anti-HIV compounds.
Moreover, the generated compound belongs to the test set indicating that the
structure of a compound that has not been used for training the system can also be
Fig. 9 Compound no. 6, its molecular graph with root vertex and one of the structures generated
from compound no. 6 that resembles the topology of compound no. 8
Combinatorial Drug Discovery from Activity-Related Substructure …
97
As done for the barbiturates, structure generation from various starting points, i.e.
compound no., atom no., was carried out for the NA series of compounds too. In
doing that, activity-related vertices have been picked up from the strong ranges in
the ordering of D
−4 index values for the vertices (atoms) of the H-suppressed graphs
of these compounds. It has been found that a few carbon skeletons resembling the
structure of other active compounds than the ones from where the activity-related
vertices and the corresponding distance distribution values are taken have been
generated.
For the purpose of illustration, the structure of the compound no. 6 and the
generated structure which corresponds to compound no. 8 are shown in Fig. 9. It
can be seen that in this case too, the algorithm is able to generate a structure with
significantly different scaffold than the starting compound and has a higher MPS
value (MPS = 65) too compared to that (MPS = 8) of the starting structure indicating that this generated structure has the potential of being highly active and
therefore may be picked/prioritized for further studies. In fact, compound no. 8 is a
potent anti-HIV drug—Didanosine. Therefore, the method may be regarded as a
useful tool for generating, prioritizing and discovering potent anti-HIV compounds.
Moreover, the generated compound belongs to the test set indicating that the
structure of a compound that has not been used for training the system can also be
Fig. 9 Compound no. 6, its molecular graph with root vertex and one of the structures generated
from compound no. 6 that resembles the topology of compound no. 8
Combinatorial Drug Discovery from Activity-Related Substructure …
97