|
Internet Electronic Journal of Molecular Design - IEJMD, ISSN 1538-6414, CODEN IEJMAT
| ABSTRACT - Internet Electron. J. Mol. Des. June 2004, Volume 3, Number 6, 350-360 |
An Information-Theoretical Measure of Similarity and a Topological Shape
and Size Descriptor for Molecular Similarity Analysis
Chandan Raychaudhury and Indira Ghosh
Internet Electron. J. Mol. Des. 2004, 3, 350-360
|
Abstract:
Molecular similarity studies play an important role in today's
drug discovery research for finding new lead compounds with
the expectation that similar molecules would exhibit similar
biological activities. Such approaches are of special importance
when one tries to find lead compounds from the databases of
thousands to millions of compounds. Although a number of
similarity measures are available in the literature, it appears
that a similarity measure that takes care of the variety/diversity
in structural components, such as substructures, of chemical
compounds might find useful applications in this regard and
Shannon's measure of information content of a discrete system
may be useful in formulating such a similarity index. However,
for doing similarity analyses using molecular descriptors,
availability of suitable molecular descriptors becomes an
essential requirement too. Graph-theoretical descriptors have
been found to be very useful in this purpose since they are fast
to compute and take care of important structural aspects. Also,
there is a special interest for the descriptors reflecting shape
and size aspects of molecules since they are related to the
fitting/docking of small molecules in the macromolecular
receptor sites. We intend to address these aspects in the present
communication. In this paper, we have proposed a new
information-theoretical measure of similarity, INFSIM, based
on Shannon's measure of information content of a discrete
system. In our study, we have also used a topological shape
and size index, TSS, defined for small molecules. These
indices have been used to carry out molecular similarity
analysis for qualitative discrimination (active/inactive) of
eleven beta-lactams, taken from the literature, with respect to
the anti-bacterial activity of penicillin G. These studies have
been carried out using the software AZMOLTOP. It appears
from the present study that the molecules under consideration
may be effectively classified using INFSIM and TSS indices
since the similarity values seem to be reflected in the
experimentally determined activities of the compounds. A
comparative study with Tanimoto similarity indicates that the
proposed method has performed relatively better similarity
assessment for the studied compounds. The results indicate that
the similarity index INFSIM may find useful application in
classifying compounds of a database qualitatively according to
their activities on the basis of the structural features encoded by
TSS. Since TSS is believed to translate the topological shape
and size of chemical compounds effectively, it may find
applications in database screening for the identification of new
lead compounds in drug discovery.
|