The unique properties of nanomaterials (NMs), relative to their bulk form, have seen them used in a rapidly increasing number of commercial applications. However, with these novel properties of NMs come potential health and environmental hazards [1,2]. Thus, as part of a responsible innovation approach, NMs potential risks must be assessed in parallel to the exploitation of their benefits. Due to their enormous variability, NM risk assessment needs advanced in silico methodologies capable of extraction of advanced descriptors of the materials reflecting their ability to interact with biomolecules and live tissues, and machine learning (ML) from experimental and computational datasets.

In nanotechnology applications involving biological tissues, the responses and impacts are believed to be induced and steered by interactions at the bionano interface [3,4]. We expect that the interactions of NMs with water, amino acids, segments of lipids, sugars, and nucleic acids are especially important. The interactions with these representative fragments may constitute the material's fingerprint that is predictive of consecutive biological responses. Indeed, the NM–protein corona, the layer of adsorbed molecules formed upon immersion of the NP into a biological fluid, was found to be predictive of the NM biological activity such as association with live cells [5]. Several recent studies have demonstrated that properties of NMs responsible for their state in solution and interactions (e.g. dipole moments, dispersion energy, zeta potentials, surface energies, band gaps, ionisation potentials, hydrophilicity or Hamaker constants) correlate with the biological activity of these materials and can be related to oxidative stress, specific gene expression and development of specific adverse outcome pathways [6,7]. The descriptors responsible for reactivity, solubility, structure and stability (dissolution, heat of formation and absolute hardness) seem to be key players in the induced response. Metal ion release rates from metal oxide NPs and conduction band gap energies correlate with cytotoxicity [8,9]. The overlap of band gap energy with the cellular redox potential in such systems determines the ability of NMs to induce oxygen radicals, oxidative stress, and inflammation. The optimisation of NMs for the best performance and the lowest risk may require experiments with live cells or live animals, which often represents the most expensive and time-consuming part of the material development [1,2,6]. Replacement of such experiments by in silico methods could not only help to save resources but would also enable a much deeper approach to material development that is based on the knowledge of the relationships between the material characteristics and their activities.

A significant effort has been made recently to evaluate the relevant NM properties and build predictive models. The materials models involved quantum chemical methods to evaluate the NM surface reactivity and parameterise atomistic force fields [6,10-12], evaluate the interactions between biomolecular fragments with inorganics surfaces [13-15], and quantify the surface hydrophilicity [16] using atomistic simulations. Coarse-grained models were also proposed to predict the protein-NM interactions and the kinetics of the formation of NM protein corona [14,17]. The calculated NM properties were intended as inputs to build quantitative structure-activity relationships (QSAR) for the prediction of NM activities such as NM uptake, membrane binding, etc. [18,19].  

Another promising research direction is to couple the NM properties to systems biology, which aims to understand the molecular mechanisms of toxicity (MoA) induced by NMs. This approach provides knowledge of the complex interaction between the structure and activity of the genome and adverse biological effects caused by exogenous agents such as NMs. ML can pinpoint biomarkers of toxic endpoints, while molecular network inference to characterize NM MoA will help to define AOPs [7,20]. A key objective of  the workshop will be to develop a practical strategy to integrate  physics and biology-based models with ML and statistical modelling to determine the structure-function relationships between NM properties, biological interactions and physiological outcomes such as toxicity.