|
Internet Electronic Journal of Molecular Design - IEJMD, ISSN 1538-6414, CODEN IEJMAT
| ABSTRACT - Internet Electron. J. Mol. Des. February 2006, Volume 5, Number 2, 116-134 |
QSTR Study on Aquatic Toxicity Against Poecilia reticulata and
Tetrahymena pyriformis Using Topological Indices
Adina Costescu and Mircea V. Diudea
Internet Electron. J. Mol. Des. 2006, 5, 116-134
|
Abstract:
Considering the importance of quantitative structure-toxicity
relationship (QSTR) studies in the field of aquatic toxicology from
the viewpoint of ecological safety assessment, fish and algae
toxicity of various aromatic compounds has been modeled by the
multiple regression technique. Topological indices are used to
relate the chemical structures to their toxic activity against aquatic
organisms. In our experiment we do not look for the best model for
a property on a given set of molecules, but for the best set of
descriptors modeling a given functional group. Correlation analysis
and Multiple Linear Regression (MLR) have been carried out to
derive the best QSAR models, giving important information on
functional groups by topological descriptors. The models
belonging to benzene derivative toxicity against Poecilia reticulata
were validated with an external validation set, and the indices in
the best models were used to predict the toxicity for other two sets
of aromatic compounds against Tetrahymena pyriformis. The
robustness and prediction power of each model was evaluated by a
leave-half-out (LHO) procedure. The classification performance of
topological descriptor models generated with multiple regressions
show that the statistical results depend strongly on the functional
groups description. The basic set, toxicity data of 92 diverse
aromatic compounds against Poecilia reticulata, provided very
good results witch were validated more then 93%. For the both
prediction sets against Tetrahymena pyriformis the best models
are: R = 0.98; R2 = 0.96;
Q2 = 0.94; F = 188.36; s = 0.13 (the set
with 37 nitrobenzene derivatives) and R = 0.92; R2 = 0.84;
Q2 = 0.82; F = 198.64;
s = 0.2 (for 167 aromatic compounds). Our
statistical study demonstrated that topological indices based on
Cluj matrices show a good predictive ability of the aquatic toxicity
against Poecilia reticulata and Tetrahymena pyriformis. These
indices account for molecular bulk, halogen, nitro and amino
substitutions in the benzene ring. The four descriptors that describe
the equation of nitrobenzene derivative are very useful to predict
the activity of other compounds containing the nitro functional
group. It appears that topological descriptors have significant
potential in QSTR studies, which warrants extensive evaluation.
This procedure can be used to approach the aquatic toxicity and to
select the appropriate model for new chemical compounds.
|