Assessing Toxicity of Fullerene Nanostructures Using Human Proteins by Combined Computational Chemistry and Cheminformatics Approach

In recent years nanomaterials have found a widespread application in material science, pharmaceutical, biomedical and medical fields. Various nanomaterials are applied, as well as in development, as high effective drugs or as drug delivery agents for targeted treatment, including cancer, viruses and etc. Here we show how the combination of computational chemistry, data mining and cheminformatics approaches can help in assessment of toxicity and virtual screening of fullerene nanostructures for combined drug-like and toxicity properties. We demonstrate how the combination of inverse molecular docking, quantum chemistry and machine learning methods can be applied to investigate these nanostructures. A high-throughput virtual screening was applied to both large set of fullerene nanostructures and known biological targets (human proteins), to assess binding scores and to investigate the potential interactions with biological targets. We show several case studies, including analysis of aquatic toxicity of fullerene derivatives, since it is crucial to understand their fate in the environment, particularly their effects on aquatic organisms. Regression models were built and artificial neural network algorithms were applied for property prediction and screening FDs for structural features responsible for binding activity and then highlighted the most active fulleren derivatives that may have the greatest impact on aquatic organisms.<br/>The outcome of this study has the potential to reveal which structural features of fullerene nanostructures have the main impact on toxicity, to be able to decide what fullerene derivatives should be prioritized first. This will help to decrease the need for animal testing and make decisions on applications in advance of manufacturing.