In silico prediction of the aniline derivatives toxicities to Tetrahymena pyriformis using chemometrics tools

Abstract

A hybrid variable selection procedure was proposed to derive the in-silico models enabling the calculated chemical descriptors to be correlated to the toxicity of aniline derivatives on Tetrahymena pyriformis. The modified ant colony optimization (MACO) combined with the shuffling cross-validation (SCV) technique and the adaptive neuro-fuzzy inference system (ANFIS) were utilized as variable selection and mapping tool to obtain quantitative structure–toxicity relationship (QSTR) models. It should be noted that there are few QSTR models in the literature dealing with aniline derivatives. Satisfactory results ( = 0.943 and RMSE_LOO = 0.153) in comparison with the previous works and other techniques indicate that the proposed models have reasonable predictive abilities. The QSTR model suggests that the toxicity of the studied compounds mainly depends on the size of the molecule along the longest axis, logarithm of octanol/water partition coefficient, and polarity of the molecule. The obtained in-silico model may be used for the prediction of toxicity and risk assessment of chemicals to achieve better ecotoxicological management and prevent adverse health consequences.

Keywords:

• quantitative structure–toxicity relationships
• aniline derivatives
• Tetrahymena pyriformis
• modified ant colony optimization
• adaptive neuro-fuzzy inference system