BOOK REVIEWS
6, 2002 1507
BOOK REVIEWS
Topological Indices and Related Descriptors in QSAR
and QSPR. Edited by James Devillers & Alexandru T.
Balaban. Gordon and Breach Science Publishers: Singapore.
1999. 811 pp. 90-5699-239-2. $198.00
There are several different broad approaches to making correlations
between chemical structure and some desired property or bioactivity
(either of which is here spoken of simply as the “activity”). The
relevance of such approaches to quantitative-structure-activity-relation-
ships (QSAR) has now achieved widespread use, perhaps most
especially in evaluating bioactivities (e.g., for drug development). One
general approach to QSAR seeks to correlate a desired activity with
another reference property (e.g., the octanol-water partition coefficient)
which is more easily measured; a second general approach seeks to
correlate a desired activity to quantum-chemically computed descriptors;
and a third general approach seeks to correlate a desired activity with
various “topological indices” (which in the mathematical graph-theory
literature are usually referred to as “graph invariants”). Especially the
first two approaches seem often to have been imagined to associate to
a surmised mechanism giving rise to the desired activity, though for
bioactivities mechanistic details are in practice often wanting. The
present book is dedicated to the third sometimes somewhat controversial
but now increasingly successful topological-index approachsindeed
the book is focused not only on QSAR but also on quantitative structure-
property relations (QSPR) for other properties.
The book consists of 17 chapters by several different leading
practitioners in the field. There is a rational plan, with evident
coordination between chapters, sometimes engendered by the participa-
tion of one of the editors. These chapters might be somewhat
approximately divided into four categories: the first consisting of two
or three introductory chapters; the second category consisting of six or
so chapters presenting a diversity of topological indices; the third
consisting of five or so chapters using particular chemical philosophies
for the selection of the topological indices to be used; and the fourth
consisting of three chapters attending to a few further computational
problems and related strategies.
The first introductory category includes chapters by Devillers, by
Balaban and O. Ivanciuc, and by O. Ivanciuc and Balaban (though the
last of these chapters could be included in the next category). Devillers
gives some general history of QSAR relating to the other broad
approaches not detailed in the present book and makes some related
philosophical remarks. Balaban and Ivanciuc present a history focusing
on the introductions of various topological indices. The third of these
chapters presents a general review of chemical graph-theoretic rudi-
ments, with nice illustrative examples. Multilinear regression analysis
is a presumed prerequisite throughout the present book (much as it is
also similarly utilized for other broad QSAR approaches outside the
scope of the present book).
The second set of chapters presenting a great variety of graph
invariants includes chapters by O. Ivanciuc, T. Ivanciuc, and Balaban,
by O. and T. Ivanciuc, by S. Nikolić, N. Trinajstić, and Z. Mihalić, by
L. H. Hall and L. B. Kier, and by D. Bonchev (though the last chapter
could be placed in the next category instead). Here the graph invariants
may often be viewed as obtained in some manner from various graph-
theoretic matrices, such as the adjacency, Laplacian, shortest-path
J. Chem. Inf. Comput. Sci., Vol. 42, No.
distance, Szeged, path, or detour matrices as well as matrices derived
therefrom by inverting the elements or “complementing” them or raising
them to a power. Particularly in the first chapter of the present set of
chapters, attention is directed to the incorporation of heteroatom and
bond-weighting aspects of molecular graphs into the design of the
topological indices, though such considerations appear in passing in
other chapters of this set too, and in the subsequent categories of
chapters such considerations are generally explicitly developed or
already presumed.
The third set of more tightly application-focused chapters are by E.
Estrada, by Kier and Hall, by Hall and Kier, by S. C. Basak, by
Devillers, and by J. E. Dubois, J. P. Doucet, A. Panaye, and B. T. Fan
(though here the first two of these chapters could also be placed in the
previous category). Often the topological indices in a chapter here
become much restricted in comparison with the possibilities enunciated
in the previous set of chapters, but the oft-major point then is that the
topological indices are rationally selected within the chemical philo-
sophical approach of the authors. For instance, Kier and Hall focus on
their set of “kappa” indices for encoding global shape and flexibility
information and their “electrotopological” indices for encoding mean
local structural information (including electronegativity characteristics
for the various atoms). And Dubois et al. focus on their ordering of
substructural features to describe activities which are viewable to be
determined by a local region as perturbed by the surrounding environ-
ment within the molecule. Basak also describes a method of average
neighborhood analysis, using information-theory-related graph invari-
ants.
The fourth and final category of chapters concerned with miscel-
laneous additional problems and related computational strategies are
by Basak, B. D. Gute, and G. D. Grunwald, by O. Ivanciuc, and by O.
Ivanciuc and Devillers. The chapter by Basak incorporates a variety of
extra-graph-theoretic (quantum-chemical or property) information in
fitting for an activity. (There are a couple of other earlier chapters which
also mention in passing the idea that various pieces of geometric
information can be incorporated in indices otherwise resembling
standard purely graph-theoretic topological indices.) The chapter by
Ivanciuc concerns the use of neural networks to aid in identifying
(especially nonlinear) correlations between activity and topological
indices. The final chapter is an overview mentioning several available
software packages.
Overall this ∼800-page book provides a reasonably comprehensive
and fair presentation of the current field of use of topological indices
for QSAR and QSPR. Throughout the book (in virtually every chapter)
a variety of illustrative examples are developed, and resultant fits are
noted. The book is certainly of value for anyone interested in QSAR
and QSPR, regardless of whether the researcher is a practitioner of the
topological-index approach or of one of the other oft-used approaches
to predict activities.
D. J. Klein
Texas A&M UniVersity/GalVeston
CI010441H
10.1021/ci010441h