Predicting the toxicological effects of nanomaterials with novel modeling approach
(Nanowerk Spotlight) Nanotoxicology can be defined as the science of engineered nanodevices and nanostructures that deals with their effects in living organisms, emerging from the toxicology of nanoparticles and gaining increasing importance with the growth of nanotechnological applications. When adapting the existing definition of 'toxicology' of the Society of Toxicology to nanomaterials one would describe nanotoxicology as the study of the adverse effects of engineered nanomaterials on living organisms and the ecosystems, including the prevention and amelioration of such adverse effects (source).
The question, of course, is whether it will be possible to rationally design environmentally benign engineered nanoparticles that are not expected to cause toxicity – rather than forge ahead with creating new nanomaterials just on the basis of their intended functionality and then test their toxicity after they have been produced or even included in commercial products and applications.
In a new research field that could be called 'experimental nanotoxicology', scientists have now, for the first time, demonstrated that biological effects of manufactured nanoparticles (MNPs) can be predicted using their chemical, physical, and geometrical properties. The results successfully demonstrate the high potential of cheminformatics approaches for improving the experimental design and prioritizing the biological testing of novel MNPs.
"Our main motivation was our general interest in extending statistical molecular modeling techniques that in principle can relate structural features to the function or biological effects of complex molecules towards increasingly more complex systems – such as proteins or manufactured nanoparticles" Alexander Tropsha, K. H. Lee Distinguished Professor and Chair, UNC School of Pharmacy, explains to Nanowerk.
Study design for quantitative nanostructure-activity relationship (QNAR) modeling using both calculated as well as experimentally measured properties of manufactured nanoparticles as descriptors. (Reprinted with permission from American Chemical Society)
Reporting their findings in a recent issue of ACS Nano ("Quantitative Nanostructure-Activity Relationship Modeling"), Tropsha and his collaborators have, for the first time, tested the feasibility of modelling the very complex problem of biological properties of nanoparticles. They termed their approach quantitative nanostructure-activity relationship (QNAR) modeling.
The team modeled their approach after a process used in drug design and chemical synthesis, where the chemical structure of a new compound is quantitatively correlated with a well defined process, such as biological activity or chemical reactivity. Such a Quantitative Structure Activity Relationship (QSAR) can then be utilized to help guide chemical synthesis and drug design.
"The chief scientific core of our work is that we have thought of applying statistical molecular modeling approaches to datasets of very structurally complex molecules – such as manufactured nanoparticles – in the context of relating their structural features to – also complex – biological effects e.g., cellular uptake" Tropsha explains. "What we have demonstrated, perhaps unexpectedly, is that despite the complexity of the underlying system the use of rigorous modeling procedures resulted in models that were both statistically significant and externally predictive."
He points out that modeling MNPs and their biological effects is challenging due to two major issues: "First, because of the high structural complexity and diversity of MNPs, it is difficult to develop quantitative parameters capable of characterizing the structural and chemical properties of MNPs. Second, systematic physicochemical, geometrical, structural, and biological studies of MNPs are nearly absent in the public domain, making the development of statistically significant computational models and their validation difficult as these procedures require relatively large amounts of data."
Following the established principles of conventional QSAR modeling workflows, the UNC team set out to develop predictive QNAR models. Similar to general QSAR modeling strategies, the overall objective of QNAR models is to relate a set of descriptors characterizing MNPs with their measured biological effects, for example, cell viability, or cellular uptake.
The intended result of these models would be the ability to apply them to newly designed or commercially available MNPs in order to quickly and efficiently assess their potential biological effects.
Hierarchical clustering analysis of 51 MNPs using their biological activity profiles. The clustered distance matrix reveals three distinct clusters of MNPs based on their biological activity profiles (on the distance matrix, blue colors = high similarity between nanoparticles, red/green/yellow colors = low/medium similarity between nanoparticles). (Reprinted with permission from American Chemical Society)
"In both case studies, QNAR calculations led to statistically validated and externally predictive models; these models quantitatively relate the chemical, physical, and geometrical properties of MNPs with their biological effects measured in vitro in different cell-based assays" says Tropsha. "We believe that this report, which to the best of our knowledge is the first example of QNAR analysis of relatively large data sets of MNPs, successfully demonstrates the high potential of cheminformatics approaches for improving the experimental design and prioritizing the biological testing of novel MNPs."
Instead of current practice, where nanomaterials are tested, and if needed modified, after their production, the models built by the UNC scientists can be used to prioritize and bias the synthesis of MNPs towards particles with the desired biological activity and safety profiles.
Tropsha says that the team's long-term goal is to extend their approach to a broader variety of nanomaterials – which is daunting given the large variety of nanomaterials possible. "This will require complete datasets, where different materials are routinely evaluated using the same repertoire of assays, and would be helped by an ethos where investigators make their data publicly available for analyses that include multiple datasets and different types of materials."