compounds are first found out using some rules [18, 19] and the activity is predicted
on the basis of the number of vertex index values falling in these ranges as defined
in the rule-based system [18, 19]. For the present purpose, the values of vertex
index D
À4 are computed for the vertices of the training set compounds (molecular
graphs). Once the indices are computed, they are arranged in an ascending order
and ranges of values coming from both active and inactive compounds are found in
the ordering and are tagged as “Active” and “Inactive” ranges by applying certain
rules [18, 19] given below:
1. Three or more consecutive vertex index values coming exclusively from
active compounds and exclusively from inactive compounds are said to
form an “active range” and an “inactive range”, respectively. However, at
least three index values in a range have to be distinct if they come from the
same compound and at least two index values in a range have to be
distinct if they come from different compounds.
2. Some single vertex index value coming from both active and inactive
compounds is not considered to form an “active range” or “an inactive
range” by itself or along with other vertex index values unless two-thirds
of that single vertex index comes from active compounds or inactive
compounds, respectively.
It has been discussed earlier [24] in connection with identifying ranges that the
vertices which correspond to the vertex index values forming active ranges may be
regarded as topological features responsible for making the compounds active. In
other words, they may be regarded as a set of features forming “Topological
Biophore ” which are responsible for exhibiting a given biological activity of the
compound under consideration. From this point of view, it may be said that if the
index values of some (or, all) of the vertices of a compound fall in active ranges,
then those vertices may be regarded as forming certain topological biophore which
make the compound active. Presumably, some of the vertex index values of a
compound may fall in inactive ranges as well. Thus, in order to predict activity from
the occurrences of the vertex index values in active and inactive ranges, another set
of rules [18, 19], given below, are applied:
A compound is predicted “ACTIVE” if all or some of its vertices fall:
1. Only in active ranges or
2. In both active and inactive ranges, the number of index values falling in
active ranges is greater than those falling in inactive ranges.
Otherwise, the compound is predicted “inactive”.
Combinatorial Drug Discovery from Activity-Related Substructure …
77
on the basis of the number of vertex index values falling in these ranges as defined
in the rule-based system [18, 19]. For the present purpose, the values of vertex
index D
À4 are computed for the vertices of the training set compounds (molecular
graphs). Once the indices are computed, they are arranged in an ascending order
and ranges of values coming from both active and inactive compounds are found in
the ordering and are tagged as “Active” and “Inactive” ranges by applying certain
rules [18, 19] given below:
1. Three or more consecutive vertex index values coming exclusively from
active compounds and exclusively from inactive compounds are said to
form an “active range” and an “inactive range”, respectively. However, at
least three index values in a range have to be distinct if they come from the
same compound and at least two index values in a range have to be
distinct if they come from different compounds.
2. Some single vertex index value coming from both active and inactive
compounds is not considered to form an “active range” or “an inactive
range” by itself or along with other vertex index values unless two-thirds
of that single vertex index comes from active compounds or inactive
compounds, respectively.
It has been discussed earlier [24] in connection with identifying ranges that the
vertices which correspond to the vertex index values forming active ranges may be
regarded as topological features responsible for making the compounds active. In
other words, they may be regarded as a set of features forming “Topological
Biophore ” which are responsible for exhibiting a given biological activity of the
compound under consideration. From this point of view, it may be said that if the
index values of some (or, all) of the vertices of a compound fall in active ranges,
then those vertices may be regarded as forming certain topological biophore which
make the compound active. Presumably, some of the vertex index values of a
compound may fall in inactive ranges as well. Thus, in order to predict activity from
the occurrences of the vertex index values in active and inactive ranges, another set
of rules [18, 19], given below, are applied:
A compound is predicted “ACTIVE” if all or some of its vertices fall:
1. Only in active ranges or
2. In both active and inactive ranges, the number of index values falling in
active ranges is greater than those falling in inactive ranges.
Otherwise, the compound is predicted “inactive”.
Combinatorial Drug Discovery from Activity-Related Substructure …
77