Neural network classification of mutagens using structural fragment data

SAR and QSAR modeling of endocrine disruptors

A number of xenobiotics by mimicking natural hormones can disrupt crucial functions in wildlife and humans. These chemicals termed endocrine disruptors are able to exert adverse effects through a variety of mechanisms. Fortunately, there is a growing interest in the study of these structurally diverse chemicals mainly through research programs based on in vitro and in vivo experimentations but also by means of SAR and QSAR models. The goal of our study was to retrieve from the literature all the papers dealing with structure-activity models on endocrine disruptor xenobiotics. A critical analysis of these models was made focusing our attention on the quality of the biological data, the significance of the molecular descriptors and the validity of the statistical tools used for deriving the models. The predictive power and domain of application of these models were also discussed.

A new strategy for using supervised artificial neural networks in QSAR

A new type of environmental QSAR model is presented for the common situation in which the biological activity of molecules mainly depends on their 1-octanol/water partition coefficient (log P). In a first step, a classical regression equation with log P is derived. The residuals obtained with this simple linear equation are then modeled from a supervised artificial neural network including different molecular descriptors as input neurons. Finally, results produced by the linear and nonlinear models are both considered for calculating the activity values, which are compared with the initial actual activity values. A heterogeneous database of 569 organic compounds with 96-h LC50s measured to the fathead minnow (Pimephales promelas), randomly divided into a training set of 484 chemicals and a testing set of 85 chemicals, was used as illustrative example to show the potentialities of this new modeling strategy Finally, practical suggestions are given for designing this type of hybrid QSAR model.

Fragmental approach in QSPR

Methodological problems of using fragmental descriptors for construction of QSAR/QSPR equations are considered, and the main achievements in this field are summarized and discussed. If a structure-property data set is sufficiently large to allow building statistically significant models, then any topological index can be replaced with a set of fragmental descriptors. Several examples of using the fragmental approach for predicting retention indices and the normal boiling points of organic compounds are considered. Advantages of using fragmental descriptors, namely a "transparency" and interpretability of QSAR/QSPR models, are exemplified.

Use of graph theoretic parameters in risk assessment of chemicals

In many instances of risk assessment, one has to estimate the potential risk of chemicals using limited experimental data, or no empirical data at all. In such cases, the use of non-empirical parameters, which can be calculated directly from structure, is a viable option for the risk assessor. Graph invariants have been used in predicting properties of congeneric sets of chemicals and determining structural similarity/dissimilarity of molecules. In this paper we have used (a) topological parameters in predicting mutagenicity of a diverse set of 520 chemicals and (b) graph theoretic parameters in quantifying structural similarity for a selection of analogs. The results of these analyses are presented along with a critical discussion of the utility and limitations of these methods.