Indian Journal of Chemistry

Vol. 42A, June 2003, pp. 1207- 1218

Review

Chemical graph theory-Facts and fiction

Milan Randic
National Institute of Chemistry, Hajdri hova 19, Ljubljana. Slovenia
Received 31 January 2003

Graph Theory (GT) and its applications in chemistry, the so-called Chemical Graph Theory (CGT), appear to be two of
the most misunderstood areas of theoretical chemistry. We outlin e briefly possible causes for mi sunderstanding and suggest
remedies, incl uding a test o n the knowledge of GT and CGT.

Introduction " primitive." The Conjugated Circuits method

Graph Theory (GT) is a not so young branch of enumerates circuits within individual Kekule valence
discrete mathematics. It is generally accepted that it structures of polycyclic conjugated hydrocarbons
started with Leonhard Euler's paper I on the seven circuits in which there is a regular alternation of CC
l9
bridges of Konigsberg published in 1736. It has single and CC double bonds . The outcomes of such
received due attention after the first book on Graph enumeration are analytical expressions for molecular
20
Theori, which appeared two hundred years later, was resonance energy (RE). Schaad and Hess have
published in 1936. Since then GT became one of the shown that the method of Conjugated Circuits is
fastest expanding branched of mathematics, the closely related to Herndon 's Resonance Theory21, a
importance of which has been particularly recognized variant of VB calculations based so lely on the set of
in its role with development of the algorithms for Kekule valence structures of a molecule, that has been
22
computer applications of GT3. Graph theory has been in fact considered some time ago by Simpson , but
4 was mostly (undeservingly) overlooked. Let us also
accepted and appreciated in physics as well as in
biologi, but its acceptance in chemistry has been point out that although the Resonance Theory and the
marred with numerous unwarranted obstructions, Conjugated Circuit Model if based on the same
despite that it made contributions in chemical parameterization become mathematically equivalent,
documentation 6, structural chemistri, physical the two approaches are conceptually and
chemistrl, inorganic chemistr/, quantum computationally different. This has become more
23
chemistry 10, orgamc chemistry II , chemical apparent as defJlOn strated by Klein and coworkers
synthesis l2 , polymer chemistry 13, medicinal with application of the Conjugated Circuit model ,
4 l5 particularly , to computation of stabilities of
chemistri , genomics and DNA studies , and of 24
recent date proteomics 16. fullerenes , for which no other computations offered
insights into their stabilities.
This paper on the facts and fiction surrounding
Chemical Graph Theory should be viewed not only
Chemical Graph Theory has been motivated by
as equal to other branches of theoretical chemistry but
comments received for one recent graph theoretical
also as complementary and necessary for better
paperl 7 in which new molecular descriptors have been
understanding of "the nature of the chemical
proposed. In some areas of chemistry, notably
structure". It is true that at one time it was not
physical chemistry, theoretical chemistry and uncommon to see GT misidentified with HMO, the
medicinal chemistry, there appears to be continuing Ruckel Molecular Orbitals model of early Quantum
hesitation to accept graph theoretical concepts and Chemistry. The likely reason for this is because for
methodology as a legitimate theoretical tool. Thi s is graphs of conjugated hydrocarbons the adjacency
not the place to list numerous cases of matrix corresponds to the Huckel matrix of HMO
mi sunderstanding of CTG but let us mention one such method. In thi s context one could understand that
case which reflects the position clearly. In an article l8 Chemical Graph theory was of lesser interest to many
on quantum chemical computation of the stability of chemists - but even at that time GT was applied to
[n]phenalenes graph theoretical method of numerous diverse problems of chemistry, some listed
"Conjugated Circuits" has been referred as in Table 1, which had nothing to do with HMO.
 1208 INDIAN J CHEM , SEC A, JUNE 2003

Table I- Areas of appl ication of Topologica l Indi ces

Physico-chemical properties of molec ules, including those havin g heteroatoms

Bi ological activity of dru gs, including toxi city
Searc h fo r pharmacophore
Sea rch o f large databases
Mo lecular Similarity
Mo lec ular di versity
Enumeration of isomers
Degenerate rearrangements
Drug desig n
Sc reenin g of combinatorial libraries
Characterizati o n of fold ed proteins
Characteri zation of DNA primary sequences
Characteri zation of t-RNA
Characteri zati o n of molecular shape
Characteri zati o n of mo lecular chirality
Doc kin g fo r molecular recog niti o n
Numeri cal characteriza tion of proteo mi s maps

Topo logical Indices have found use in chemical much different. Graphs, however, have the additional
and biological applications and that Chemical graph advantage in that they allow some flexibility in
theory has made important conceptual and associati ng with individual edges and individual
quantitative contributions to chemistry. In the later vertices various weights, whi ch can be different in
part of this contribution we will list a number of different applications. Embedded graphs are defined
important recent contributions of topological indices as graphs of fixed geometry, which may but need not
and Chemical Graph T heory to chemistry and let coincide with the geometry of a chemical structure.
readers to delineate facts from fiction . Part of the One of the major uses of graphs is to serve as
problem may be in that some concepts of GT are so so urces of various structural invariants, which is
close to the chemical language of structural chemistry tantamount to say in g various mathematical properties
with whi ch many chemists are familiar. Thus many of structure. Thus, compounds that typicall y exh ibi t .
chemists may get an impress ion that GT is (if not various physico-chemical properties and biological
simple, and even simpli stic) not very sophisticated, acti vities can now tn addition exhib it various
and hence not capable of offering proper insights on mathematical properties. There is an important
chemical structure. That such a position is fa lse and distinction between the collection of physico-
that GT and CGT are rich in content and include chemical properties and biological activities of a
numerou s profound propositio ns can be easily found molecule and the co ll ection of its mathematical
if one is interested in this subject. properties: The number of phys ico-chemical
properties and biological activities of a molecule is
Topological indices as molecular descriptors finite, while in contrast mathematical properties
We will only briefly outline the substance of appear unlimited in their number. However, for a
topological indices and molecular descriptors in order , mathematical property to be of interest in chemistry it
to facilitate readers unfamiliar with details of has to show its use. This may be in structure-property
Chemical Graph Theory to fo rm their own view on regressIOns, structure-acti vity relation shi ps,
the nature of topologi cal indices and to be able to establishing mo lecular similarity and diversity ,
form an opinion on their potential use. For more screentng combinatorial libraries, clustering of
detail. readers should consul t several of avai lable chemical compounds, design o f novel drugs,
25
review articles on topological indices , including also characterization of DNA structures, characterization
a brief introduction avai lable in the Encyclopedia of of proteomics maps, etc. (see Table I) .
Computational Chemistr/ G• Topological indices are The most common use of mathematical invariants,
structural invariants based on modeling of chemical which are also known as graph theoretical indices or
structures by molecular graphs. Hence, covalent topological indices, is as molecular descriptors in
bo nds are represented by edges and atoms as vertices. QSPR and QSAR (quantitative structure-property
Superficially molecular graphs and molecular relationships and quantitative structure-actIvIty
structural formulas that chemists often use are not relationships, respectively). There are at least two
 RANDle: CHEMICAL GRAPH THEORY 1209

aspects of QSPR and QSAR with slightly different pathways, in agreement with Wong's coevolution
emphasis: (1) One is interested in as good as possible theory of the genetic code, were obtained."
predictions of properties without being concerned Katritzky, Lobanov & Karelson 30.3 1 have developed
with the interpretation of the descriptors used, if they computer software, which is distributed freely to
do the job; (2) One is interested in as good as possible people in academic institutions. CODESSA computes
characterization of properties and one is much some 400 molecular descriptors of which about a
concerned with the interpretation of the descriptors third are topological indices. This software has been
used. In the first case typically one selects a subset of used in numerous applications in QSPR and QSAR
structures for testing the model, in the second case ever since it was introduced .
one is using all available data and trying to Agrafiotis 32 describes a novel diversity metric for
"understand" the model. Since properties are usually use in the design of combinatorial chemistry and
expressed numerically, clearly if one considers high-throughput screening experiments. We cite a
regressions one needs descriptors that will · also short extract of this paper: "The data set used in this
numerically characterize a structure - hence a need study is based on the reductive amination reaction ...
for topological indices and other structural invariants. and is utilized for the construction of structurally
diverse druglike molecules with useful
Recent accomplishments in use of topological pharmacological properties . . . For demonstration
indices as molecular descriptors purposes, 300 primary and secondary amines and 300
We collected a dozen illustrations from the aldehydes were selected at random . . . and were used
literature in which topological indices have been used to generate a virtual library of 90,000 products . . .
for diverse problems of chemistrl7 . 39 , to examine the Each compound in the 90,000-membered library was
papers and then judge if topological indices are characterized by an established set of 117 topological
useful. descriptors, which were subsequently normalized and
Flower27 in his paper on DISSIM, a computer decorrelated using principal component analysis,
program which addresses the problem of selecting resulting in an orthogonal set of 23 latent variables
diverse subsets from larger collections of chemical which accounted for 99% of the total variance of data
compounds considered relationships of 159
topological indices (listed in his Table 1) of which 39 Liu and collaborators 33 reported on a molecular
were used in his CACS (Computer-Aided Compound electronegativity distance vector based on 13 atomi c
Selection) protocol. types, called MEDV -13, as a descriptor for predicting
Lahana and coworkers 28 by selecting four from two the biological activities of molecules based on QSAR.
dozen topological indices for screening a Their conclusion was that the study on steroids shows
combinatorial library of some 280,000 virtual that the performance of the MEDV-13 method has
decapeptides were able, by selecting "windows" of comparability with the previous methods containing
allowed values for individual topological indices and 3D QSAR, and the peptide study gives a high quality
nine other molecular descriptors, to reduce this of the QSAR model based on the MEDV -13 method.
enormous library to 26 compounds on which they However, the MEDV -13 method only employs
performed more advanced calculations. Finally they information about an element atom type, valence
selected five compounds and synthesized them electron state, and chemical bond type from the 20
finding that one compound had an molecular topology and requires no information
immunosuppressive activity that was almost hundred related to 3D structures or physicochemical properties
times more active than the lead compound! or molecular alignment. So, the MEDV -13 descriptor
Andrade and coworkers 29 used novel topological is fast, easy to use, reproducible and predictable one
indices to characterize t-RNA. They proposed for the QSAR studies.
"weighted structural descriptors-closely related to Galvez and collaborators 34 tried to combine
Randic connectivity and Balaban distance indices-as parameters of semiempirical (quantum chemical)
distinctive characteristics of each structure. Molecules calculations with topological indices in order to arri ve
were characterized by a set of weighted structural at any discriminant function for antibacterial activity.
descriptors and classified by a clustering method and They found that two quantum chemical descriptors
discriminant function analysis. Two main groups of (QDs) played important role for discriminant function
tRNAs that correspond to the biosynthetic amino acid but at the same time they have shown that these two
1210 INDIAN J CHEM, SEC A, JUNE 2003

QDs can be well expressed by topological indices connectivity indices (I XV, 3xv. 4 XV), molecular weight,
(TIs). The utility of expressing QDs "as a function of and temperature as an input parameters. . . .The
the TIs is to the extent that their predictive capability performance of the QSPR for temperature-dependent
is for larger sets of compounds for which the vapor pressure, which was developed from a simple
calculation of TIs is much faster and easier than the set of molecular descriptors, displayed accuracy of
calculation of QOs .. .". They concluded: "Using the better than or well within the range of other avrulable
multilinear regression and the LDA (linear estimation methods.
discriminant analysis), a pattern of topological 37
Estrada and Molina considered the classification
simjlarity of antibacterial activity has been obtained. of antibacterial activity of 2-furylethylene derivatives.
This pattern has been applied successfully for the According to them: Topographic (3D) molecular
search of drugs that together with other connectivity indices based on molecular graphs
pharmacological activities can also show antibacterial weighted with quantum chemical parameters are used
activity. An added advantage is that this screening can in QSPR and QSAR studies. These descriptors were
be carried out using large databases and with low compared to 2D connectivity indices (vertex and edge
time-consuming." ones) and to quantum chemical descriptors in
Ga0 35 reported that the genetic algorithm was very modeling partItIon coefficient (log P) and
efficient method for variable selection or antibacterial activity of 2-furylethylene derivatives. In
optimization. Besides available topological indices he describing log P the 3D connectivity indices produced
also used BCUT metric (an extension of Burden's a significant improvement (more than 29%) in the
parameters, which are based on a combination of the predictive capacity of the model compared to those
atomic number for each atom and a description of the derived with topological and quantum cherrucal
nominal bond type for adjacent and non-adjacent descriptors. The best linear discrimjnant model for
atoms and incorporates both connectivity information classifying antibacterial activity of these compounds
and atomic properties). He made binary QSAR was also obtained with the use of 3D connectivity
analysis of Carbonic Anhydrase II Inhibitors. "The indices. The global percent of good classification
best binary QSAR model was obtained with a obtained with 3D and 20 connectivity indices as well
combination of 23 molecular descriptors including as quantum chemical descriptors were 94.1,91.2 and
. . 111
f our connectIvIty . d exes ( 2X, 0 X,
v 1
X,v an d 2 X,
V) 88.2 respectively.
three shape indexes CZK, 1Ka, 3 Ka ), log P(o/w), and 15 We included this paper as it clearly shows that
BCUTs .. . The cross-validated accuracy is 91 % on graph theoretical approaches are not lirruted to 20
active compounds, 92% on inactive compounds, and objects (graphs) but can be extended to 3D objects
91 % for all compounds. Thus the predictive power of (embedded graphs) . To quote from the introduction of
the binary QSAR model is quite high." He also made the paper of Estrada & Molina: An important step
binary QSAR analysis of Estrogen Receptor Ligands forward in the development of graph-based molecular
and obtained sirrular results: 'The best binary QSAR descriptors has been the definition of topographic
model was obtained with a combination of 24 descriptors. These kinds of molecular descriptors are
molecular descriptors including four connectivity based on molecular graphs with appropriate weights
indexes (0 X, 1 X, 2 X, and 1 XV), two shape indexes (IKa, to account for 3D molecular features. The pioneering
2Ka), flexibility index <1>, log P(o/w), and 16 BCUTs works in this direction were done by Milan Randic at
... The cross-validated accuracy is 73% on active the end of 1980s. In these works Randic proposed the
compounds, 90% on inactive compounds, and 88% use of topographic distance matrices, first based on
for all compounds. " graph embedded on a hexagonal lattice and then on a
Yaffe and Cohen 36 reported on estimation of vapor 3D diamond lattice.
pressure using topological indices as descriptors. Here Burden 38 considered a large Benzodiazepine Data
is an extract from their Abstract: A neural network Set (245 compounds), Muscarine Data Set (162
based quantitative structure-property relationship compounds) and Toxicity Data Set (277substituted
(QSPR) was developed for the vapor pressure- benzenes). Here is an extract from his paper: Gaussian
temperature behavior of hydrocarbons based on a data processes method (GPM) constitutes a method of
set for 274 compounds. The optimal QSPR model was solving regression problems. The usual coefficients or
developed based on 7-29-1 back-propagation neural weights associated with other regression methods are
network architecture using valence molecular absent and an exact Bayesian analysis is
 RANDle: CHEMICAL GRAPH THEORY 1211

accomplished using matrix manipulations. . . A classification biological actIvItIes from models of

combination of five set of easily computed indices structural similarity; comparison of a neural net-based
were employed in this work: the well studied Randic QSAR algorithm with holograms and multiple linear
index (R); the valence modification to the Randic regression - based QSAR; graph theoretical analysis
index by Kier and Hall (K); and an atomistic (A) of tunneling electron transfer in large polycyclic
index developed by Burden which has now been aromatic hydrocarbon networks; on distance related
enhanced by recognition of aromatic atoms and indexes; an extension of the Wiener index and the
hydrogen atom donors and acceptors (B) . . . Two "overall Wiener index;" interpretation of well-known
further indices have been added the first counts the topological indices, and structural interpretation of
number of rings of various sizes (G) and the second several distance-related topological indices ;
counts some common functional groups (F) .. . The characterization of DNA primary sequences based on
four type of index, R, K, A, and B, are average distances between bases, and characterization
complementary, and we have shown in previous of DNA by triplet of nucleic acid bases; illustration of
studies that their combination yields better QSAR CODESSA-based theoretical QSPR model with
models than the individual indices alone. variable · molecular descriptors based on distance
Now if all this is not sufficient to convince readers related matrices; novel shape descriptors for
that topological indices are useful in chemistry and molecular graphs and use of graph shells as molecular
found applications in quite diverse problems then I descriptors; characterization of 2D chirality; and
would like to draw attention of readers to a paper of several papers on the use of the variable connectivity
mine written in collaboration with Novic and index and the hierarchical approach to QSAR.
Vracko l6 • In this paper the authors explore the
characterization of 2-D electrophoresis proteomics Real world chemistry-what is it?
maps by certain structural invariants derived from In the past some cntIcs of CGT were
matrices constructed by considering for all pair of "complaining" that researchers in GT are mostly
spots in a proteomics maps the shortest (Euclidean) preoccupied with hydrocarbons . Hydrocarbons are
distances and distances me red along zigzag lines also part of the "real-world chemistry". If not why
connecting protein spots of the neighboring not? In Table 2 are listed a few hydrocarbons and
abundance. This paper is a sequel to previous papers their properties just to remind readers that
in which we outlined the idea of characterizing 2-D hydrocarbon chemistry, including fullerenes that may
proteomics maps by graph-theoretical descriptors. To or may not be decorated with hydrogens, may be as
illustrate the approach we selected data of Anderson fascinating as any branch of chemistry. Fullerenes are
et al.39 on protein abundance in mouse liver under even a better illustration of simple chemical graphs in
series of dose of peroxisome proliferator L Y 1711883. view that usually hydrogens are not present and need
We found strong linear correlation between the not be suppressed!
experimentally applied doses and the leading What critics overlook in objecting to modeling
eigenvalue of Distance/Distance type matrix 40 CGT on hydrocarbons is to see that this is only the
constructed for the experimental proteomics maps. first stage in developing novel molecular descriptors
and models. If a new topological index is not useful in
Discussion characterizing selected properties of hydrocarbons,
Readers will have to judge whether the utility of they are probably even less suitable for characterizing
topological indices has been demonstrated "beyond of properties of heteroatomic compounds. Once
the reasonable doubt." Observe that half a dozen descriptors are found useful they, as a rule, are
"cases" have been taken from not only a single generalized to extend to molecules having
journal, but a single volume of that journal. Moreover, heteroatoms. The list of the dozen papers outlined in
in the same volume of J chem In! Comput Sci. that we the previous section has numerous such illustrations.
screened 28 papers41 are published from the Second
Indo-US Workshop on Mathematical Chemistry, held Concluding remarks
in Duluth, MN, in May 2000 which is "crowded" with We will end with a brief outline of the manuscript
papers on application of GT to Chemistry, including that was the reason behind this review. The manuscript
numerous publication on novel topological indices. considers several novel topological indices for octane
For example, there are papers on: criteria for isomers (but equally applicable to other molecules),
1212 INDIAN J CHEM, SEC A, JUNE 2003

Table 2-Diverse properties of different hydrocarbons

Formula Name Property

CJ2HIO heat transfer agent; fungistat for oranges

C 1sH I8 Guaiazul ene anti-inflammatory; anti-ulcerative
C I9H 32 Tridecylbenzene detergent; forms stable foam in the presence of fat
C I9 H40 Pristane lubricant, ant i-corrosion agent
C22 H I4 Pentacene large crystal semiconductor
C 21 H I6 3-Methylcholanthrene experimental use in cancer research
C2) H46 Muscalure sex pheromone
C 30 H SO Squalene oil, agreeable odor; bactericide
C30 H62 Squalane lubrica nt; transfo rmer oil; perfume fixative; skin lubricant
C.jQH sl Lycopene carotenoid occurring in ripe fruit
C4o Hs6 a-Carotene vitamin A precursor
C4o H68 Phytofluene polyene hydrocarbon widespread in the vegetable kingdom
Ci soHI86 Hexabenzocoronene derivative liquid crystal (component for photovoltaic films)

all derived from a new graphical matrix. Matrix Table 3-The regression coefficient for quadratic regressions of
steric contributions for iso mers of octane with selection of
elements of a graphical matrix, in contrast to the
topological indices. Indices indicated by asterisk are introduced in
traditional graph theoretical matrices (such as the ref. 13.
adjacency matrix, distance, matrix, detour matrix,
Wiener and Hosoya matrices, etc.) are not numerical , Descriptor F Ref.
but are defined through qualified subgraphs of a
W(W) 0.9846 0.331 237 .6 *
graph considered. Graphical matrices have been for
WW 0.9839 0.339 227.3
the first time introduced only a few years ago, and
WWP 0.9839 0.339 226.8 *
besides the seminal paper42, the paper we are talking
111 0.9813 0.364 195.3
about is the only other manuscript considered this
subject 17 • Graphical matrices are not widely known W 0.9806 0.371 187.7
even among chemical graph theory , being published eig W(W) 0.9777 0.398 162. 5 *
in less available and less read mathematical journals. J 0.9766 0.407 154.8
So this particular paper was the first opportunity for a Shell S2 0.9760 0.412 150.9
wider circle of readers interested in structural eig WW 0.9753 0.419 145.9 *
chemistry to hear about graphical matrices! p) 0.9770 0.417 145.5 *
Because matrix elements of graphical matrix are H 0.9639 0.504 98.3
non-numerical the first step in applications is to arrive IIJJ 0.9497 0.593 69.0
at a numerical representation for such matrices, which R*R 0.9365 0.664 53 .5 *
is based on a selection of a particular graph invariant RRW 2 0.9008 0.823 32.3
to serve as "transition" to a numerical matrix. Once a ID 0.8998 0.827 31.9
numerical matrix is constructed, one has choice to 0.8953 0.844 30.3
P2 /w 2
select any of already known procedures to extract 0.7824 l.l80 11.8
X
structural invariants from such a matrix . Thus if one 1.346 7.4
Z 0.7036
chooses the Wiener index of subgraphs appearing in
the matrix as a recipe to arrive at a numerical matrix,
and then again choosing the Wiener index of the so 18 different descriptors, including half a dozen
obtained numerical matrix one has novel topological (shown by asterisks) derived from the novel graphical
index, referred to briefly as Wiener of Wiener, and matrix. The message of Table 3 is to point out that
labeled as W(W). As shown in Table 3, index W(W) while many topological indices may yield simjlar
was found to be the best descriptor for regression of results, nevertheless some are better than others when
Scott's steric contributions of octane isomers. a particular property is considered. For instance, it is
Moreover in Table 3 we show statistical parameters known from the literature that X and Z, which are at
for Scott's steric contributions of octane isomers for the bottom of the table (hence among those shown the
 RANDle: CHEMICAL GRAPH THEORY 1213

"worst" descriptors for the property considered) are in 14 Randic M, Jerman-Blaz ic B, Rouvray D H, Seybold P G &
Grossman S C, Int J QlIantum Chem Quantum Bioi Symp, 14
fact the best molecular descriptors for the boiling
(1987) 245; Randic M in: Rein E (ed), Molecular bases of
points of alkanes! cancer, Part A: Macromolecular structure, carcinogens, and
As discussed at some length in the manuscript (and ollcogens (Liss, New York), 1985, p 309.
outlined in the abstract) novel indices give the best 15 Randic M & Balaban AT, J chem Inf Comput Sci (in press).
regression for selected physicochemical properties of 16 Randic M, Novic M & Vracko,M , J Proteome Res, I (2002)
217;Randic M, In! J Quantum Chern, 90 (2002) 848 ;Randic
octanes, hence are superior the already existing
M, Zupan J & Novic M, J chem Inf Comput Sci, 41 (2001)
indices for the properties considered. 1339.
Graph Theory has been found a useful tool and has its 17 The manuscript: Randic M & Basak N, Novel graphical
place in physics, biology, including biochemistry, matrix and novel distance based molecular descriptors, has
molecular and cellular biology, proteomics and been since resubmitted to SAR & QSAR environ Res (in
press).
genomics, ecology, geography, in economics, in
18 Kovacek D, Margetic D & Maksic Z B, J molec Siruct
psychology, linguistics, social sCiences and In (Theochem) 285 (1993) 195.
chemistry. 19 Randic M, Chem Phys Lett, 38 (1976) 68 ; Randic M, J Am
chem Soc, 99 (1977) 444; Randic M, Tetrahedron, 33 (1977)
Acknowledgment 1905.
The author would like to thank Professor A T 20 Schaad L J & Hess B A, Pure appl Chem, 54 (1982) 1097 ;
Balaban (Texas A & M at Galveston, TX) for Herndo n W C, MA TCH Commull math Comput Chern, 11
discussions concerning this manuscript and many of (1981) 3.
21 Herndon W C, J Am chem Soc, 95 (1973) 2404.
his valuable comments.
22 Simpson W T , J Am chem Soc, 75 (1953) 597 ; Simpson W T ,
References Theories of electrons in molecules (Prentice-Hall,
I EulerL, Solutio Problematic ad Geometriam Situs Eng lewood Cliffs), 1962.
Pertinantis, Academimae Petropolitanae, 8 (1736) 128 . 23 Klein D J, Topics curr Chern, 153 ( 1990) 57;Klein D J, J
2 Konig D, Einfiihrung in die Theorie der endlichen lind
chern Educ, 67 ( 1990) 633; Klein D J & Trinajstic N, Pure
unendliched Graphes (Chel sea, New York), 1950. appl Chem, 61 ( 1989) 2107; Klein D J, Pure appl Chern, 5
( 1983) 299;Klein D J, J chem Edllc, 69 (1992) 691 ;Zivkovic
3 Gibbons A, Algorithmic graph theory (Ca mbridge Univ
T , Randic M, Klein D J , Zhu H Y & Trinajstic N, J comput
Press, Cambridge), 1985.
Chem, 16 ( 1995 ) 517 ; Klein D J, Zhu H, Valenti R & Garcia-
4 Temperley H N V, Graph theory and applications (Ellis
Bach M A, Int J quan! Chem, 65 (1997) 421; Zhu H & Klein
Horwood, Chichester), 1981.
D J, MATCH Commun math Comput Chem , 31 (1994) 205.
5 Randic M & Krilov G, Chern Phys Lett. 272 ( 1997) 115 ; Zhu H, Balaban A T, Klein D J & Zivkovic T p, J chem
Balaban A T, Rev Roum Chern, 21 (1976) 1049. Phys, 101 (1994) 5281; Liu X, Schmalz T G & Klein D J,
6 Read R C, J chern InfCompllt Sci, 23 (1983) 135; Randic M, Chem Phys Lett 188 ( 1992) 550; Seitz W A, Hite G E,
Int J Quantum Chern Quantum Biol Symp, (1980) 187 ; Schmalz T G & Klein D J in: King R B & Rouvray D H
Randic M, J chem Inf Comput Sci, 35 ( 1995) 357 ; Randic M (eds), Graph theory alld topology in chemistry (Elsevier,
Croat chern Acta, 59 (1986) 327; Godson A L, J chern Inf Amsterdam ), 1987; Klein D J, Hite G E & Seitz W A, J
Comput Sci, 20 (1980) 167; Godson A L, Croat chem Acta, comput Chern, 7 (1986) 443.
56 (1983) 315; and Lozac' h N, J chem Inf Comput Sci, 25 24 Klein D J, Schmalz T G, Hite G E & Seitz W A, JAm chem
( 1985) 180. Soc, 98 (1986) 1301. Schmalz T G, Seitz W A, Klein D J &
7 Randic M, Acta chim Sloven, 45 (1998) 239; Randic M, Acta Hite G E, Chem Phys Lett, 130 (1986) 203.
Cryst, A34 (1978) 275. 25 Balaban A T & Ivanciuc 0 in: Devillers J & Balaban A T
8 Schmalz T G , Klein D J & Sandleback B L, J chem Inf (eds), Topological indices and related descriptors in QSAR
Comput Sci, 32 (1992) 54. and QSPR (Gordon & Breach, Amsterdam), 1999, pp 21-57 ;
9 King R B, Chern Rev, 101 (200 1) 1119; Randic M, Oakland Basak S C in: Devillers J & Balaban A T (eds), Topological
D 0 & Klein D J, J comput Chern, 7 (1986) 35. indices alld related descriptors in QSAR and QSPR (Gordon
10 Matsen F A, Int J Quantum Chem Symp, 8 (1974) 379; & Breach, Amsterdam), 1999, pp 563-593 ; Randic M, J
Paid us J, J chem Phys, 61 (1974) 5321; Paldus J, Lecture chem Inf Comput Sci, 37 (1997) 672; Randic M, J mol
Notes in Chemistry, No. 22 (Springer-Verlag, New York), Graphics Model, 20 (200 1) 19; Estrada E in: Devillers J &
1981 ; Shaviu I, Int J quant Chem, 11 (1977) 131. Balaban A T (eds), Topological indices and related
II Balaban A Tin: Bonchev D & Melenyan 0 (eds), Graph descriptors in QSAR and QSPR (Gordon & Breach,
theory approaches to chemical reactivity (Kluwer, Amsterdam), 1999, pp 403-45 3; Estrada E & Molina E in: M
Dordrecht). 1994. V Diudea (ed), QSARlQSPR studies by molecular descriptors
12 Corey E J, Rev chem Soc, 25 (1971) 455; Corey E, The logic (Nova, New York), 2000, pp 91-1 16; Estrada E & Uriarte E,
of chemical synthesis (Wiley, New York), 1989; Zefirov N S Curr med Chern , 2001 (in print).
& Gordeeva E V, Usp Khim, 56 (1987 ) 1753. 26 Randic M in : Schleyer P v R, Allinger N L, Clark T ,
13 Klein D J, J chem Phys, 75 (1981) 5186; Gordon M & Gasteiger J, Kollman P A, Schaefer III H F & Schreiner P R
Temple W B in: Balaban AT (ed), Chemical applications of (eds), The encyclopedia of computational chemistry (Wiley,
graph theory (Academic Press, London), 1976, pp 299-332. Chichester), (1998), pp 3018-3032.
1214 INDIAN J CHEM, SEC A, JUNE 2003

27 Flower D R, J 11101 Graphics Mode, 16 (1998) 239. Books on Chemical Graph Theory:
28 Grassy G, Calas B, Yasri A, Lahana R, Woo J, Iyer S, Chemical ApplicatiollS 0/ Graph Theory (Balaban, A. T. Ed.)
Kaczorek M, Floc' h R & Buelov R, Nature Biotechnol, 16 Academic Press, London, (1976)
(1998)748. Trinajstic, N. Chemical Graph Theory, (2nd ed.) CRC Press, Boca
29 Andrade J, Th eor Bioi, 197 ( 1999) 193. Raton, FI 1992.
30 Katritzky A R. Karelson M & Lobanov Y, CODESSA Computational Chem ical Graph TheOlY (Rouvray, D. H. Ed.),
(Comprehensive Descriptors for Structural and Statistical Nova Sci. Pub!.: Commack, N. Y. (1990).
Analysis); University of Florida: Gainesville, FL, 1994. Merrifield, R. E.: Simmons, H. E. Topological Methods in
31 For a summary of earlier results obtained with CODESSA Chemistry, John Wiley & Sons, New York, NY, 1989
see: Katritzky A R, Karelson M & Lobanov Y From Chemical Topology to Three-Dimensional Geometry,
"Understanding How Chemical Structure Detennines Balaban, A. T. Ed. (Balaban, A. T. Ed.) Plenum, New
Physical Properties" in which there are summary of 46 York, (1997).
publications involving CODESSA (a preprint - courtesy of Topological Indices and Related Descriptors in QSAR and QSPR
Professor A. R. Katritzky, Dept. of Chemistry, Univ. of (Devilers, 1. ; Balaban, A. T. (Eds.); Gordon & Breach Pub!.
Florida, Gainesville, FL 326 11 ). Amsterdam (1999)
32 Agrafiotis D K, J chem In/ Comput Sci, 41 (2001) 159. QSPR I QSAR Studies by Molecular Descriptors, Diudea, M. Y.
33 Liu S S, J chem In/ Comput Sci, 41 (200 I) 321. Ed. Nova Sci .: Pub!. Huntington, NY, 2002.
34 Galvez J, J chem In/ Comput Sci, 41 (2001) 387.
35 Gao H, J chem In/Comput Sci, 41 (2001) 402. A Selection of review articles on various topics of CGT:
36 Yaffe D & Cohen Y, J chem InfComput Sci, 4 1 (2001) 463. Balaban, A. T. ; Ivanciuc, O. in : Topological Indices and Related
37 Estrada E & Molina E, J chem 111/ Comput Sci, 41 (200 I) Descriptors in QSAR alld QSPR (Devi lers, J.; Balaban, A.
791. T. (Eds.); Gordon & Breach Pub!. Amsterdam (1999), pp
38 Burden F R, J chem In/ Comput Sci, 41 (2001) 830. 2 1-57.
39 Anderson N L, Esquer-Blasco R, Richardson F, Foxworthy P Basak, S. C. in : Topological Indices and Related Descriptors in
& Eacho P, Toxicol Appl Pharmacol, 137 (1996) 75. QSAR and QSPR (Devilers, J.; Balaban, A. T., Eds.);
40 Randic M, Kleiner A F & DeAlba L M, J chem Ill/ Comput Gordon & Breach Pub!.: Amsterdam ( 1999), pp 563-593.
Sci, 34 (1994) 277. Trinajstic, N.; Klein, D. J.; Randic, M. On Some Solved and
41 Second Indo-US Workshop on Mathematical Chemistry, Unsolved Problems of Chemical Graph Theory. Int. J.
May 3~-June 3, 2000, Duluth, MN ; J chem In/ CompL/l Sci QuantuTll Chem: Quantum Chem. Symp. 1986,20,699-742.
41 (2001) pp 479-701 (total of28 papers). Randic, M.; Trinajstic, N. Notes on Some Less Known Early
42 Randic M, Plavsic D & Razinger M, MATCH Commun math Contributions to C hemical Graph Theory, Croat. Chern.
Comput Chem, 35 (1997) 243. Acta 1994,67, 1-35.
Balaban, A. T. Chemical Graphs: Looking Back and Glimpsing
Ahead. J. Chern. Illf. Compo Sci. 1995, 35, 339-350.
Appendix 1 Randic, M. On characterization of Chemical Structure. J. Chern.
Inf. Compo Sci. 1997,37,672-687.
Introductory Graph Theory books: Randic, M. The Connectivity Index 25 Years After. J. Mol.
Ore, O. Graphs and Their Uses, Random House: New York: Graphics & Modelling 2001, 20,19.
( 1963). Randic, M. Chemical Structure - What is "She." J. Chern. Educ.
Wilson, J. R. Introduction to Graph Theory, Oliver and Boyd: 1992, 69,713-718.
Edinburgh (1972). Balasubramanian, K. Applications of Combinatorics and Graph
Bondy, J. A.; Murty, U. S. R. Graph Theory with Applications, Theory to Spectroscopy and Quantum Chemistry. Chern.
MacMillan Press, Ltd: London (1976). Rev. 1985,85,599-618.
Chartrand, G. Graphs as Mathematical Models, Prindle, Weber & Pogliani , L. From Molecular Connectivity Indice to
Schmidt, Inc.: Boston 1977. Semiempirical Connectivity Terms: Recent Trends in
West, D. B. Introduction to Graph Theory. Prentice-Hall, Upper Graph Theoretical Descriptors. Chern. Rev. 2000, 100,
Saddle River, NJ , 1996. 3827-3858.
Konig, D. Ein/iihrung in die Theorie der Endlich en !/lui Balaban, A. T. Solved and Unsolved Problems in Chemical Graph
Unendliched Graphes, Chelsea: New York (1950). Theory, Annals Discrete Math. 1993,55, 109-126.
Balaban, A. T. Chemical Graphs. Part 49: Open Problems in the
Textbooks and advanced textbooks on GT: Area of Condensed Polycyclic Benzenoids: Topological
Ore, O. Theory 0/ Graphs, Am. Math. Soc. Providence: RI (1962) Stereoisomers of Coronoids and Congeners. Rev. Rourn.
Harary, F. Graph Theory, Addison-Wesley : Reading, MA (1969) Chim. 1988, 33,699-707.
Bucvkley, F.; Harary, F. Distance in Graphs, Addison-Wesley Balaban, A. T. Challenging Problems involving Benzenoid
Pub. Redwood City, CA, 1990. Polycyclics' and Related Systems. Pure Appl. Chern. 1982,
Bursaker, R. G.; Saaty, T. L. Finite Graphs and Networks, 54, 1075-1096.
McGraw-Hill, NY, 1965. Balaban, A. T. Is Aromaticity Outmoded? Pure Appl. Chern.
Harary, F.; Palmer, E. M. Graphical Enumerations. Academic 1980,52, 1409-1492.
Press, New York, NY 1973
Tutte, W. T. Graph Theory, Addison-Wesley: Reading, MA 1984 Appendix 2
For definitions of graph theoretical terminology see: We collected numerous questions that cover various topics and
Essam, J. W.; Fisher, M. E. Rev. Mod. Phys. 1970,42, 271. results from GT and CGT, which can serve as a test for those who
 RANDle: CHEMICAL GRAPH THEORY 1215

think that they know graph theory in order to find out how much 18. What is Dellaney triangulation of a map? How is it
they know. We do not supply the answers but they can be found constructed? Why may it be of interest in chemistry?
by contacting the literature of CG and CGT. To assist interested 19. What is the " monster" graph ? How big is the monster graph
readers to find correct answers we listed in the Appendix (how many vertices)?
references that will help to find answers to specific questions. 20. How was the Wiener index defined by Harry Wiener and
how is it related to the distance matrix?
List of questions to test one's knowledge of Graph Theory and 2 1. What is a structural interpretation of the Wiener index?
Chemical Graph Theory 22. What is the difference between )(jer & Hall' s valence
There is no significance to the order in which the following con nectivity index and the rel atively recently proposed
questions listed appear. Equally, some questions are more variable connectivity index?
important and some are less important e ither for GT or CGT, 23. Why is phenanthrene more aromatic than anthracene?
some questions are more diffi cult even if they may be 24. Is it true that n-alkane always have the highest BP (Boiling
intermingled with easier questions. In reviewin g the questi ons we point) among 'all isomers having th e same n? If yes - why
advise that one keep his/her score by adding + I for every correct yes? [f no - why no?
answer and subtracting a penalty of -2 for every incorrect answer, 25. What is the difference between a graph matri x (such as the
while hav ing zero fo r unanswered questions. Penalty has been adjacency matrix, the distance matri x) and graphical
deliberate ly given a greater weight as a wrong answer requires matrices of graphs?
first to eradicate the error and then to learn then correct answer. 26. It is well known that one can easi ly obtain the adjacency
Observe that several questi ons have two and three sub-questi ons, matri x fro m a di stance matri x. Can one co nstruct (not
which ought to be considered separate ly and co unted separately. If loo king at a picture of a graph) the distance matri x from the
you answer all questions correctly yo u ca n have as much as 100 adjacency matri x?
points and proudly claim to be an expert on GT and CGT. Eve n a 27. Do topo logical indices have to have phys ical meaning? If the
score of 50 would indicate a fair fa mili arity with GT and CGT, answer is yes, then why?
but anything below 50 poi nts to lack of familiarity with GT and 28. [s parti al ordering of interest in chemistry? Give an
CGT. illustrati on if the answer is positi ve.
29. What is the difference between molecular topological indices
Here is a list of questions: and mo lecul ar topographic indices?
I. What is Clarke's theorem about? Why is this theorem so 30. What is the major result of a work of Collatz and
important? Sinogowitz?
2. What is the di ffe rence between the Characteristi c polyno mi al 3 1. What is "graph Reco nstructi on" problem? Who and when
and Matching polynomial? proposed the problem? Has the problem been solved?
3. John Platt made two important co ntributions to chemical GT. 32. Wh at is the contribution of Joshu a Lederberg (Nobel
What are they? Laureate) to chemical notation?
4. Why is Petersen graph of interest in chemistry? What does it 33. What is the contribution of Vlado Prelog (Nobel Laureate) to
represent in chemistry? chemi cal notation?
5. When has been published the first book on GT? In what 34. What are Ugi 's BE matrices?
language? Who was the author? What was his nationality? 35. What are the uses of canonical labeling of verti ces of graphs:
6. In what year was published th e first book on Chemical GT? To solve Graph Isomorphi sm proble m? To solve Graph
Who was the editor? automorphism problem? Both?
7. What is a DID matrix? What structural interpretation has its 36. Who was first to publish a paper illustrating subgraphs
leading eigenvalue? contributing to the construction of the characteristic
8. What is the elegant algorithm of Gordon and Davidson for? polynomial ?
9. What is a theorem of John and Sachs about?
37. Who and when was the first to use term "graph" for
10. What is the difference between the "conjugated circuits" and
mathematical objects known today as graphs?
"circuits of conjugation"? Both terms were used in the
38. Who was first to consider enumeration of graphs?
literature.
39. Who and when published the first paper on en umeration of
II. Consider use of a single variable connectivity index (based
chemical isomers? In what journal? In what language?
on several variables) in structure-property-activity
regressions. Does this then represent a simple regression or a 40. What are (vertex and edge) transitive graphs? Why are they
multivariate regression analysis? of interest in chemistry? Who and when introduced first such
12. What is the major advantage of orthogonalized molecular graph in chemistry?
descriptors in comparison with standard (non-orthogonal) 41. What is the size of the smallest edge (but not vertex)
descriptors? transitive graph? Who constructed this graph?
13. What is the difference between graphical and graph 42. What graphs are cages labeled as cages? Is the Coxeter graph
theoretical methodology? (vertex and edge transitive cubic graph on n = 28 vertices) a
14. Can a graph theoretical index characterize chirality? Gi ve an cage?
example if possible or state " not possi ble". 43. What is graph center? What is graph diameter? What is girth
15. What is the difference between a graph theoretical invariant of a graph?
and a topological index? 44. Who classified graph algorithms as NP (non-polynomial )
16. Who introduced Distance matrix in GT? and P (polynomi al)?
17. What are the Kronecker and the Hadamard product of 45 . Is enumeration of paths of different length in a graph solved
matrices? by a non-polynomial (NP) or polyno;nial algorithm?
 1216 INDIAN J CHEM, SEC A, JUNE 2003

46. Is enumeration of walks of different length in a graph solved Q 6: See the first edition of Konig, D. Einfiihrung in die
by a non-polynomial (NP) or polynomial algorithm? Theorie der Elldlichell und Unelldliched Graphes, which
47. What is the detour matrix? Who introduced the detour matrix was reprinted by Chelsea: New York in 1950.
in GT? Does it have another name? Q 7: See: Randic, M.; Kleiner, A. F.; DeAlba, L. M. J. Chem.
48. Can two non-isomorphic graphs have identical di stance In! Comput. Sci.1994, 34, 277.
matrix? Can two non-isomorphic graphs have identical Q 8: See: Gordon, M. ; Davison, W. H. T. 1. Chem. Phys. 1952,
detour matrix? 20,428.
49. How is Hosoya's topological index constructed? See also: Randic, M. Aromaticity of Po lycyclic Conjugated
50. What is the Frobenius theorem that defined bounds on the Hydrocarbons. Chem. Rev. (in print).
leading eigenvalue of a matrix? Q 9: See: John, P.; Sachs, H. Calculating the Numbers of
Perfect Matchings and of Spanning Trees, Pauling' s
51. What are "migrating" sextets?
Orders, the Characteristic Polynomial, and the
52. Has Klaus Ruedenberg published a paper that can be viewed
Eigenvectors of a Benzenoid system. Topics Curro Chem.
as CGT paper?
1990, /53, 146-179.
53. Has Hans Primas published a paper that can be viewed as Q 10: See: Randic, M. Aromaticity of Polycyclic Conjugated
CGT paper? Hydrocarbons. Chem. Rev. (in print).
54. Has C. A. Coulson published a paper that can be viewed as Q II: Some think "yes" (Zefirov, N. S.; Palyulin, V. A. QSAR
CGT paper? for boi ling points of "small" sul tides. Are the "H igh-
Has Linus Pauling (doubly Nobel Laureate) published a paper that Quality Structure-Property-Activity Regression s" the real
can be viewed as CGT paper? high quality QSAR model. J. Chem. In! Compul. Sci.
Has G. Wheland published a paper that can be viewed as CGT 2001 ,4 /, 1022-1U27) but some think the correct answer is
paper? " no" (see: Randic, M. Variable Molecular Descriptors and
Has Rudolf Marcus (Nobel Laureate) published a paper that can High-Quality Regression Analysis. Understanding and
be viewed as CGT paper? Misunderstandings. J. Chem. In! Comput. Sci.
55. What is the crossing number? What is the crossing number of (submitted).
4-D cube? Q 12: See: Randic, M. Resolution of ambiguities in structure-
56. Are fullerene graphs planar graphs? property studies by use of orthogonal descriptors. 1. Chem.
57 . What is Laplace matrix and what for it is used in GT and In! Comput. Sci. 1991, 31 , 3 11 -320; Randic, M.
CGT? Orthogonal molecular descriptors. New.l. Chem. 1991 , 15,
58. Are th e row sums of a sy mmetric matrix invaraints? 517-525 ; Randic, M. Fitting of non-linear regressions by
orthogonalized power series. J. Compul. Chem. 1993, 14,
Is the sum of all matrix elements of a sy mmetric matrix an
363-370.
invariant?
Q 13: For illustrations of a graphical approaches see:
59. Should a shape index depend on molecular size (number n of
Randic, M. Graphical Enumeration of Conformations of
vertices)? Do Kier's K2 and K3 depend on n? Chains. Int. J. Quantum Chem: Quantum Bioi. Symp.
60. Do we have enough topological indices? If yes - why yes? If 1980, 7, 187-197.
no - why no? Klavzar,S.; Zigert, P.; Gutman, I., Theochem (submitted).
Randic, M. Polycyclic Aromatic Hydrocarbons
We list here sources for checking the answers to the 60 (submitted).
questions li sted in the text rather than providing the answer in Q 14: See: Liang, c.; Mislow, K. J. Math. Chem. 1995, 18, 1-
order that an effort is made to find the answer. For instance, on 24; Randic, M. ; Razinger, M. J. Chem. In! Comput. Sci.
question 5 we could have li sted easily the name of the author, the 1996, 36, 429-442 ; Randic, M. 1. Chem. Ill! Comput. Sci.
title of the book (which would reveal the language), indicate the 2001 ,41,639-649.
nationality of the author, and even giving some biographical Golbraikh, A.; Bonchev, D.; Tropsha, A. J. Chem. In!
details (for instance, the author of the book has committed suicide Comput. Sci. 2001 , 41,147- 158.
during WW II). But that would be counterproductive to our desire Schultz, H. P.; Schultz, E. B.; Schultz, T. P. 1. Chem. In!
that readers do the investigative research. Hence, we give hints Comput. Sci. 1995, 35, 864-870.
where the answers could be found and at a later time we plan to Q 15: See any of many review articles on topological indices,
give the answers and accompanying information . which are more popular (but somewhat incorrect) name
for graph theoretical invariants also referred to as
Answers to question can be found in: molecular descriptors.
Q I: Journal a/Graph Theory Q 16: See: Harary, F. Graph Theory, Add ison-Wesley: Reading,
Q 2: See: Trinajstic, N. Computing the Characteristic MA (1969)
Polynomial of a Conjugated System Using the Sachs Q 17: See any textbook on Linear Algebra or MATLAB Manual.
Theorem. Croat. Chem. Acta 1977,49, 593-633. Q 18: Dellaney triangulation is dual of a map which is divided in
Q 3: See: Platt, J. R. in Handbuch der Physik (Fliigge, S. Ed.), regions such that all points with a region are closer to the
Springer-Verlag: Berlin (1961), pp. 205-209; and: Platt, J. center of the region than to any other neighboring center.
R. Prediction of isomeric differences in paraffin See any textbook on advanced geometry.
properties. J. Phys. Chem. 1952,56,328-336. Q 19: See: Randic, M.; Oakland, D. O.; Klein, D. 1. J. Compul.
Q 4: See: Dunitz, J. D. ; Prelog, V. Allgew. Chem. 1968, 80, Chem. 1986, 7, 35.
700; Balaban, A. T.; Farca~iu , D.; Banicii, R. Rev. Roum. Zivkovic, T. Croat. Chem. Acta
Chim. 1966, II, 1205. Q 20: See: Wiener, H. Prediction of isomeric differences In
Q 5: See: Graph Theory 1736 - 1936. paraffin properties. J. Am. Chem. Soc. 1947,69, 17-20.
 RANDle: CHEMICAL GRAPH THEORY 121 7

See: Hosoya, H. Topological index. A new ly proposed quantity Q 35: See: Read, R. c.; Corneil, D. G. The Graph Isomorphi sm
characterizing the to po log ical nature of structural isomers Di sease, l . Graph Theory 1977, 1,339.
of saturated hydrocarbons. Bull. Chem. Soc. l apan, 1971 , Q 36: See: Coulson , C. A. Proc. Cambridge Phil. Soc. 1950, 46,
44,2332-2339. 202.
Q 21 See: Ra ndi c, M.; Zupan. 1. On Structura l Interpretation of Q 37: See: Ro uvray, D. H., l . Mol. Stmct. 1989, 185, I ;Bell , E.
Topological Indi ces, In : Proc. Wiener Melllorial T. Men of Math elllatics; See also an article e ntitl ed
Conference. Athe ns 200 1 (R. B King a nd D. H. Ro uvray, Chemistry and Algebra by Sy lvester, 1. J . Nature, 1877-
Eds.); Rand ic, M.; Zupan, J . On interpretat io n of well - 1878, XVll. pp. 284, 309.
known topo logical indices. 1. Chen!. In! Comput. Sci. Q 38: See a chapter o n g raph e nume rati o n in: Biggs, R; Lloyd,
2001 ,41,550-560. E. K.; Wilson, R. J. Graph Theory 1736 - 1936,
Q 22:See: The Connectivity Index 25 Years after (J. Molecular Clarendon Press: Oxford, 1976.
Graphics alld Modelling, 2001 ,20, 19-35). Q 39: See: Balaban , A. T. Enumi!fatioll of Isomers. In Chemical
Q 23: See: Randi c, M. Aromaticity of Po lycyc lic Co njugated graph theo ry. Abacus Press, Gordon & Breach: New
Hydrocarbons, Chelll. Rev. (in press). Also: Randi c, M. l . Yo rk , 1990, p. 177-234.
Am. Chefl/. Soc. 1976, Randic M .; Trinajstic, N. On some Less Known Graph Theory
Q24: See Randic M; Wilkins. C. L. On g raph theoretical basis for Contributions, Croa t. Chell!. Acta 1994,67, 1-35.
ordering of structures. Chelll . Phys. U fl . 1979, 63, 332- Q 40: See: Trinajstic, N. Chemical graph theory, (2nd ed .) CRC
336: Ra ndi c M; Wilkins, C. L. Graph th eoretica l ordering Press: Boca Rato n, FI 1992.
of stru ctu res as a basis for systematic searches for
Q 41: See: Pi sanski, T.; Randi c, M . Bridges between Geometry
regularity in molec ul ar data. l. Phys. Chem. 1979, 83,
and Graph Theory; in : Geometry at Works (Papers in
1525- 1540.
Applied Geometry), Gorini C. A., Editor, MAA
Q 25: See: Randic, M .; Plavs ic, D.; Razinger, M . Double
(Mathem ati cal Association of America) Notes Number 53,
In vari a nts. MATCH-Comlllun. Math. COllljJLIl. Chem.
Washin gton, D. c.: 2000 o r:
1997,35,243-259.
Bondy, J. A.; Murty, U. S. R. Graph Theory with Applications,
Q 26: See: Kun z, M . An eq ui va le nce relation between distance
MacMillan Press, Ltd: Lo ndo n (1976).
and coordinate matrices. MATCH - Comlllull. Math.
Q 42 : See: Balaba n, A. T. 1. Combinato rial Theory, Ser. B
Comput. Chen!. 1995,32, 193-204.
1972, 12, I ; Balaba n, A. T. Rev. Roum. Math. Pures Appl.
Q 27: See: Randic, M.; Zupan , 1. On the structural interpretation
1973, 18, 1033; Also: W o ng, P. K Cages - a survey. l .
of topo logical indices, in : Topology ill Chemistry- Discrete
Graph Th eory. 1982, 6, 1-22.
Mathematics of Molecules (Ro uvray, D. H.; King, R. B.,
Eds.), Horwood Pub!.: Ch ichester, England, 2002, pp. Q 43: See: Harary. F. Graph Theory, Addison-Wesley: Reading,
249-29 1. MA ( 1969);Bo nc hev, D.; Balaban, A. T.; Randi c, M. Int.
Estrad a, E. The Physicochemical Interpre tatio n of Mo lec ul ar l . QuallIum Chem. 1981, 19,61.
Connectivity Ind ices, l . Phys. Chell!. (in press) Q 44: See: Mill e r, R. E.; Thatcher, J. W . Complexity of
Q 28: See: Kl ein, D. 1. Prolegomenoll on Partial Orderings in Computer Computations. Ple num Press, 1972, pp. 85-103.
Chemistry, MATCH - Commun. Math . Compllt. Chem. Q 45: See: Mill er, R. E.; Th atc her, J. W . Complexity of
2000,42, 7; Randi c, M. Vracko, M.; Novic, M. Basak, S. Computer Computations. Ple num Press, 1972, pp. 85- 103.
C. MATCH - Com/1/un. Math. Comput. Chern. 2000, 42. Q 46: Harary, F. Graph Theory, Addison-Wesley: Reading, MA
18 1. ( 1969).
Q 29: See : Rand ic, M.; Razinger, M . Molecular Topographical Q 47 : See: Trinajstic, N.; Niko li c, S.; Lucic, B.; Amic, D.;
Indices. 1. Chem. In! Comput. Sci. 1995, 35, 140- 147; Mihalic, Z. The Detoru Matri x in Chemistry. 1. C/leln. In!
Also: Estrada E. Characterization of 3D Molecular Comput. Sci. 1997, 37, 631-630;Nikolic, S.; Trin aj stic, N.;
Structure. Chem. Phys. Lefl. 2000, 3 19, 7 13-718. luric, A.; Mihalic, Z. The Deto ru Matrix and the Detour
Q 30: See: Cvetkovic, D . M. ; Doob, M. ; Sachs, H. Spectra of Index of Wei ghted Graphs. Croat. Chem. Acta 1996, 69,
Graphs, 3 rd ed. Huthi g GmbH - 1. A. Barth Verlag, 1577-1591 ;Lukovits, I. The Deto ur Index. Croat. Chem.
Heidelbe rg, 1995; Acta 1996, 69, 873-882; Lukovi ts, I.; Razin ger, M . On the
Collatz, L. ; Sinogowitz, U. Abh. Math. Sem . Ulliv. Calculation of the Detour Index, l . Chem. In! Comput.
Hamburg 1957, 21, 63. Sci. 1997, 37, 283-286.; Amic, D. ; Trinajstic, N. On the
Q 3 1: See: O'Neil, P. V. U la m's Conjecture and Graph Deto ur Matri x. Croat. Chem. Acta 1995,68,53-62.
Reco nstructi on. Amer. Math. Montly, 1970, 77, 35-43. Q 48: See: Ra ndic, M .; De Alba, L. M. ; Harri s, F. E. Graphs
Q 32: See: Lederberg, 1.; Sutherland, G. 1.; Buchanana, B. G.; with the same De tour Matrix . Croat. Chem. Acta 1998, 71 ,
Feigenbaum, E. E.; Robertson, A. V.; Duffield , A. M .; 53.
Dje rassi, C. Applicatio n of Artificial Intellige nce fo r Q 49: See : Hosoya, H. Topo logical index. A newly proposed
Chemical Inferences. I. Ther Numbe r of Possible Organic quantity characterizing the topological nature of structural
Compounds. Acyclic Structures Containing C, H, 0, and iso mers of saturated hydrocarbon s. Bull. Chem. Soc. lpn.
N. l . Am. Chern. Soc. 1969,91 ,2973-2976 .. 1971 , 44,2332-2239 .
Q 33: See: Nobel Laureates in Chemistry 1901 - 1992 (Lay lin Q 50: See a textbook o n Linear Algebra
K lames, Ed.) Amer. Chem. Soc. Washing ton , D. c., Q 51: Clar , E. The Aromatic Sextet J. Wiley & Sons: London
1993; Cahn, R. S. ; Ingo ld, C. K; Prelog, V. Angew. Chem. 1972
Int. Engl. Ed. 1966, 5, 385. Q 52: See: Ruedenbe rg, K l . Chem. Phys. 1954, 22, 1878;
Q 34: See: Dugundji , J.; Ugi, I. Topics Curro Chem.1973, 39, Ham, N. S.; Ruedenberg, K l . Chem. Phys. 1958, 29,
19. 1215.
1218 INDIAN J CHEM, SEC A, JUNE 2003

Q 53: See: Glinthard. H. H.; Primas, G. Helv. Chill/. Acta 1956, Hahn. G.: Sabidussi. G. (Eds.) Graph Symmetry:
39, 1645. Algebraic Methods alld Applications, (NATO ASI seri es,
Q 54: . See: Coulson, C. A. Proc. Call/bridge Phil. Soc. 1950,46, Series C, Mathematical and Physical sciences, vol. 497 ).
202: Pauling, L. 1. Chem. Phys. 1933, I, 280; Wheland, G. Dordrecht; Boston; London: Kluwer, 1997, pp. 225-270.
W. 1. Chem. Phys. 1933, 3, 356: Marcus, R. A. 1. Chem. Q58: True only for regular graphs in which case the row sums are
Phys. 1965, 43,2643. constant a,nd represe nt trivial invari ant (of little intereest).
Q 55: See: Harary, F. Graph Theory, Addi son-Wesley : Reading, Q 59: See: Randic, M. Novel Shape Descriptors for Molecular
MA ( 1969). Graphs. 1. Chem. Inf. Call/put. Sci. 2001, 41, 607 -613.
Q 56: All polyhedral graphs are planar Q 60: The answer is "yes" if you want to freeze the development
Q 57 : See, for instance, Mohar, B. Laplace eigenvalues of graphs of chemical graph theory as a scientific di scipline; if you
- a survey. Discrete Math. 1992, 109, 171-183; Mohar, B. think that scie nce is not static, the answer is an emphatic
Some applications of Laplace eigenvalues of graphs. "no."