activities against HIV [20, 21], and a data set of 3779 compounds (named GTB data
set) for which minimum inhibitory concentration (MIC) values have been measured
against H37Rv strain of Mycobacterium tuberculosis (Mtb) [22]. The GTB data set
may be obtained from the link [23] given in the reference section. The results
described here will therefore substantiate the findings obtained earlier [15].
Regarding activity prediction, results have been reported for NA and barbiturate
data sets. For barbiturates data set, we have considered the same training set and test
set as used in an earlier study [18]. However, for the NA data set, we have identified
a reasonably well-performing training set–test set split and have reported the results
for individual compounds present in that split. For prioritization of the generated
active compounds that help screen potential active compounds, Molecular Priority
Score (MPS) [15] has also been used and the results obtained for NA and barbiturate series of compounds have been given in the tables alongside their activity
prediction results. We have carried out combinatorial generation of structures using
topological distance-based substructural information associated with identified
activity-related vertices (atoms) in some compounds of the data set. We have been
able to reconstruct the structures of active NA and barbiturate compounds from the
substructural information associated with activity-related vertices of other active
NA and barbiturate compounds. Regarding substructure searching exercise, we
have reported identified potential active compounds from GTB data set [22, 23]
considering activity-related atoms (vertices) in the structures of Isoniazid and
Streptomycin, both of which are known antitubercular drugs in use.
It appears from the outcome of the results that the integrated method would find a
place as a useful drug discovery tool for designing and discovering novel bioactive
compounds. In particular, the method is believed to be of much help in situations
where novel drug candidates having very different structural characteristics/scaffolds
are sought for particularly to overcome the drug resistance problem.
2 Methods
In this section, we have described in detail different mathematical approaches/tools
which have been used to develop the present integrated drug discovery method and
the related computer programs. Examples with tables and figures have been used to
illustrate underlying concepts of the methods used. While we have leveraged few
existing mathematical aspects for the present purpose, we have introduced some
new algorithms as well.
2.1 Computation of Vertex Index
Let G be the carbon skeleton of n-butane and D G
ð Þ, the corresponding distance
matrix is shown in Fig. 1. Computation of D
À4 indices for the vertices of D(G) has
been illustrated below.
Combinatorial Drug Discovery from Activity-Related Substructure …
75
set) for which minimum inhibitory concentration (MIC) values have been measured
against H37Rv strain of Mycobacterium tuberculosis (Mtb) [22]. The GTB data set
may be obtained from the link [23] given in the reference section. The results
described here will therefore substantiate the findings obtained earlier [15].
Regarding activity prediction, results have been reported for NA and barbiturate
data sets. For barbiturates data set, we have considered the same training set and test
set as used in an earlier study [18]. However, for the NA data set, we have identified
a reasonably well-performing training set–test set split and have reported the results
for individual compounds present in that split. For prioritization of the generated
active compounds that help screen potential active compounds, Molecular Priority
Score (MPS) [15] has also been used and the results obtained for NA and barbiturate series of compounds have been given in the tables alongside their activity
prediction results. We have carried out combinatorial generation of structures using
topological distance-based substructural information associated with identified
activity-related vertices (atoms) in some compounds of the data set. We have been
able to reconstruct the structures of active NA and barbiturate compounds from the
substructural information associated with activity-related vertices of other active
NA and barbiturate compounds. Regarding substructure searching exercise, we
have reported identified potential active compounds from GTB data set [22, 23]
considering activity-related atoms (vertices) in the structures of Isoniazid and
Streptomycin, both of which are known antitubercular drugs in use.
It appears from the outcome of the results that the integrated method would find a
place as a useful drug discovery tool for designing and discovering novel bioactive
compounds. In particular, the method is believed to be of much help in situations
where novel drug candidates having very different structural characteristics/scaffolds
are sought for particularly to overcome the drug resistance problem.
2 Methods
In this section, we have described in detail different mathematical approaches/tools
which have been used to develop the present integrated drug discovery method and
the related computer programs. Examples with tables and figures have been used to
illustrate underlying concepts of the methods used. While we have leveraged few
existing mathematical aspects for the present purpose, we have introduced some
new algorithms as well.
2.1 Computation of Vertex Index
Let G be the carbon skeleton of n-butane and D G
ð Þ, the corresponding distance
matrix is shown in Fig. 1. Computation of D
À4 indices for the vertices of D(G) has
been illustrated below.
Combinatorial Drug Discovery from Activity-Related Substructure …
75