2007Unpublished venueRequires access

Predicting Toxicity of Aromatic Pollutants Using QSAR Modeling

Bakhtiyor Rasulev, Hrvoje Kušić, Danuta Leszczyńska, Jerzy Leszczyński, Natalija Koprivanac

Open publisher page 0 citations

Abstract

A large amount of overall organic chemical compounds produced and used annually pertain to aromatic compounds, highly toxic to living organisms in aquatic systems and soil, but to humans too, and moreover, many of them are reported as carcinogenic and mutagenic. One of the most successful approaches for predicting their toxic effect could be found in the application of QSAR/QSPR (quantitative structure-activity/property relationship) modeling. This powerful technique quantitatively relates variations in biological activity, i.e. toxicity, to changes in molecular structure and properties. Hence, the goal of the study was to predict toxicity in vivo of aromatic compounds structured by single benzene ring and including presence and absence of different substitute groups such as hydroxyl-, nitro-, amino-, methyl-, methoxy-, etc, by using QSAR/QSPR tool. A Genetic Algorithm and multiple regression analysis were applied to select the descriptors and to generate the correlation models. Evaluation of models was performed by calculating and comparing their model performances (R2, s, F, Q2) after splitting set of organic compounds to training and test sets. As the most predictive model is shown the 3-variable model having also a good ratio of the number of descriptors and their predictive ability. The main contribution to the toxicity showed descriptors belonging to 2D autocorrelation and atom-centered fragments descriptors, respectively. The GA-MLRA approach showed good results in this study, which allows to built simple, interpretable and transparent model that can be used for future studies of predicting toxicity of organic compounds to mammals

About this research paper

What this paper is about

A large amount of overall organic chemical compounds produced and used annually pertain to aromatic compounds, highly toxic to living organisms in aquatic systems and soil, but to humans too, and moreover, many of them are reported as carcinogenic and mutagenic. One of the most successful approaches for predicting their toxic effect could be found in the application of QSAR/QSPR (quantitative structure-activity/property relationship) modeling. This powerful technique quantitatively relates variations in biological activity, i.e. toxicity, to changes in molecular structure and properties. Hence, the goal of the study was to predict toxicity in vivo of aromatic compounds structured by single benzene ring and including presence and absence of different substitute groups such as hydroxyl-, nitro-, amino-, methyl-, methoxy-, etc, by using QSAR/QSPR tool. A Genetic Algorithm and multiple regression analysis were applied to select the descriptors and to generate the correlation models. Evaluation of models was performed by calculating and comparing their model performances (R2, s, F, Q2) after splitting set of organic compounds to training and test sets. As the most predictive model is shown the 3-variable model having also a good ratio of the number of descriptors and their predictive ability. The main contribution to the toxicity showed descriptors belonging to 2D autocorrelation and atom-centered fragments descriptors, respectively. The GA-MLRA approach showed good results in this study, which allows to built simple, interpretable and transparent model that can be used for future studies of predicting toxicity of organic compounds to mammals

Why it matters

A significance statement is not available in the OpenAlex record.

Key contribution

A contribution statement is not available in the OpenAlex record.

Method / approach

Method details are not available in the OpenAlex metadata.

Main findings

Findings are not separately available in the OpenAlex metadata.

Limitations

Limitations are not available in the OpenAlex metadata.

Applications

Application details are not available in the OpenAlex metadata.

Available abstract

A large amount of overall organic chemical compounds produced and used annually pertain to aromatic compounds, highly toxic to living organisms in aquatic systems and soil, but to humans too, and moreover, many of them are reported as carcinogenic and mutagenic. One of the most successful approaches for predicting their toxic effect could be found in the application of QSAR/QSPR (quantitative structure-activity/property relationship) modeling. This powerful technique quantitatively relates variations in biological activity, i.e. toxicity, to changes in molecular structure and properties. Hence, the goal of the study was to predict toxicity in vivo of aromatic compounds structured by single benzene ring and including presence and absence of different substitute groups such as hydroxyl-, nitro-, amino-, methyl-, methoxy-, etc, by using QSAR/QSPR tool. A Genetic Algorithm and multiple regression analysis were applied to select the descriptors and to generate the correlation models. Evaluation of models was performed by calculating and comparing their model performances (R2, s, F, Q2) after splitting set of organic compounds to training and test sets. As the most predictive model is shown the 3-variable model having also a good ratio of the number of descriptors and their predictive ability. The main contribution to the toxicity showed descriptors belonging to 2D autocorrelation and atom-centered fragments descriptors, respectively. The GA-MLRA approach showed good results in this study, which allows to built simple, interpretable and transparent model that can be used for future studies of predicting toxicity of organic compounds to mammals

Key concepts: Quantitative structure–activity relationship, Molecular descriptor, Applicability domain, Aquatic toxicology, Biological system, Chemistry, Biochemical engineering, Computational chemistry

Related papers

Back to paper searchBrowse research topicsOriginal source
Predicting Toxicity of Aromatic Pollutants Using QSAR Modeling — Research Paper | ScholarLens