Non-Linear QSAR Treatment of Genotoxicity

Abstract

The nonlinear QSAR approach using the Chebyshev polynomial expansion and neural networks has been applied for the prediction of genotoxicity of compounds. The mutagenic toxicity of heteroaromatic and aromatic amines, measured by the Ames test, was correlated with the molecular descriptors calculated from the molecular structures using quantum-chemical methods. The quantitative models obtained were compared with the results of the linear QSAR treatment. The descriptors appearing in the models reveal the importance of mutagenic interactions of heteroaromatic amines via hydrogen bonding, of effects induced by the solvent, and of the size of compound. The dependence of molecular descriptors on environmental effects and on molecular conformations was analysed.

Computational methods in drug discovery

“…It is only possible to use linear regression methods if the activity descriptor relation is linear. However the relationship between biological activity and the molecular descriptors are not always linear [262]. Machine learning approaches such as neural networks and support vector machine methods are used to generate QSAR models to address this issue of non-linear fitting [263–265].…”

The Present Utility and Future Potential for Medicinal Chemistry of QSAR / QSPR with Whole Molecule Descriptors

“…However, we were able to show that other simpler structure-based descriptors can be an efficient replacement for logP. A combination of step forward selection of descriptors and back-propagation NN improved the quality of the model with slightly different descriptor content of the model, indicating the possible non-linear relationship between structural determinants and genotoxicity of the compounds [131].…”

Neural Networks Convergence Using Physicochemical Data

“…In recent years, the literature concerning ANN as applied to QSPR/QSAR has grown drastically, suggesting that the importance of applications of ANN in molecular modeling is a major driving force. Among numerous studies using ANN in QSPR/QSAR, major contributions are due to the groups of Jurs, Zupan and Gasteiger, and Zefirov and ourselves − among others. These authors have shown that the ANN models frequently possess better predictive characteristics compared to models using standard multilinear regressions.…”