Computational Approaches to Fragment and Substructure Discovery and Evaluation - Fragment-Based Drug Discovery - page 211

and number of unique activities. In case of structurally similar consensus substructures, only
the highest in rank was kept. The steroid skeleton was found as a fine example of a multi-
activity structure due to the many physiological processes in which steroid hormones are
involved. Similarly highly ranked were tricyclic structures as in imipramine and doxepine.
They bind to many G-protein-coupled receptors and transport proteins.

8.3.4 Predictive Models

In an attempt to organize available data in mutagenicity databases, Kho et al . described
an automated approach to extract and organize ring systems occurring in a mutagenicity
dataset. [ 37 ] They suggested that this method can be applied to any other set of molecules
classified by some property, e.g. biological activity. A common assay for mutagenicity
prediction is the Ames test, in which Ames-positive compounds are suspected to have
mutagenic characteristics, whereas Ames-negatives are not. The database [ 38 ] was searched
for the occurrence of ring types and their frequency in the Ames-positive and -negative
categories. Emphasis was not so much on the development of predictive algorithms, but
more on organizing the available data for use by chemists. Simple scaffolds were iden-
tified using a program that finds scaffolds by comparing all molecules in a set. [ 39 ] The
results were presented as a hierarchy according to complexity. In this approach, simple
rings are placed at the highest level and more complex ring systems that contain the par-
ent rings as descendants. An example hierarchy is presented in Figure 8.8 Note that the

Confidence interval
of proportions (%)

Complexity

→

50-64

Equal odds

30-50

Equal odds

6-33

Ames +

100

Ames-

Figure 8.8 Scaffold (cyclohexene) hierarchy derived from mutagenicity data. [38] The propor-
tion of Ames negative to Ames positive counts is qualitatively indicated below each scaffold.
The confidence interval of the proportions is shown on the right of the scaffolds. Data taken
from Kho et al . [37]