In silico cytotoxicity estimation of ionic liquids based on their two- and three-dimensional structural descriptors

Abstract  

The cytotoxicity of a series of ionic liquids containing ammonium, pyrrolidinium, imidazolium, pyridinium, and piperidinium cations against leukemia rat cell line IPC-81 was estimated from their structural parameters using quantitative structure–activity relationship methodology. Linear and nonlinear models were developed using genetic algorithm multiple linear regression and multilayer perceptron neural network approaches. Robustness and reliability of the constructed models were evaluated by internal, external, and Y-randomization procedures. Furthermore, the chemical applicability domain was determined via a leverage approach for each model. The results of this study revealed that the contribution of structural characteristics of the anionic parts of the studied ILs were fewer than of the cationic parts.