A computational view on nanomaterial intrinsic and extrinsic features for nanosafety and sustainability

In recent years, an increasing number of diverse Engineered Nano-Materials (ENMs), such as nanoparticles and nanotubes, have been included in many technological applications and consumer products. The desirable and unique properties of ENMs are accompanied by potential hazards whose impacts are difficult to predict either qualitatively or in a quantitative and predictive manner. Alongside established methods for experimental and computational characterisation, physics-based modelling tools like molecular dynamics are increasingly considered in Safe and Sustainability-by-design (SSbD) strategies that put user health and environmental impact at the centre of the design and development of new products. Hence, the further development of such tools can support safe and sustainable innovation and its regulation. This paper stems from a community effort and presents the outcome of a four-year-long discussion on the benefits, capabilities and limitations of adopting physics-based modelling for computing suitable features of nanomaterials that can be used for toxicity assessment of nanomaterials in combination with data-based models and experimental assessment of toxicity endpoints. We review modern multiscale physics-based models that generate advanced system-dependent (intrinsic) or timeand environment-dependent (extrinsic) descriptors/features of ENMs (primarily, but not limited to nanoparticles, NPs), with the former being related to the bare NPs and the latter to their dynamic fingerprinting upon entering biological media. The focus is on (i) effectively representing all nanoparticle attributes for multicomponent nanomaterials, (ii) generation and inclusion of intrinsic nanoform properties, (iii) inclusion of selected extrinsic properties, (iv) the necessity of considering distributions of structural advanced features rather than only averages. This review enables us to identify and highlight a number of key challenges associated with ENMs’ data generation, curation, representation and use within machine learning or other advanced data-driven models to ultimately enhance toxicity assessment. Finally, the set up of dedicated databases as well as the development of grouping and read-across strategies based on the mode of action of ENMs using omics methods are identified as emerging methodologies for safety assessment and reduction of animal testing.