In silico exploration of graphene nanoflakes: From DFT simulations to machine learning-driven toxicity predictions

Abstract

The present theoretical work provides a ground-breaking and comprehensive study of graphene nanoflakes integrating Density Functional Theory (DFT) simulations, toxicity predictions and a machine learning approach. The properties of graphene nanoflakes as a function of size, shape, and symmetry are systematically analysed using DFT calculations. The interaction of these nanoflakes with human proteins and cell membranes, considered as Molecular Initiating Events for diverse Adverse Outcome Pathways, is explored to infer potential toxicity effects. Leveraging the generated data, machine learning models were developed to predict flake properties and biological interactions. A single score representing the biological interaction or impact of graphene nanoflakes on both proteins and plasma membranes is assigned to each evaluated nanoflake to infer its potential toxicity. Our multiscale approach bring valuable insights into the structure-property-toxicity relationships of graphene nanoflakes, paving the way for their safe and efficient design and application.