In order to optimise in silico methods, such as (Q)SAR, grouping and read-across, for the purpose of the prediction of the hazard of cosmetic ingredients after long-term exposure, a characterisation of the chemical space of the cosmetics inventory was performed, both in terms of the physico-chemical properties and chemical substructures covered in the inventory.
Subsequently, in silico approaches will be refined to incorporate kinetic and metabolic studies to permit quantitative interpretation of results in terms of consumer risk.
Chemical space is a representation of the structural features and/or molecular properties covered by a defined set of chemicals.
The COSMOS inventory and TTC dataset compiled were assessed for their representation of the chemical space of cosmetic ingredients, by analysing structure and subgraph features, as well as physico-chemical property descriptors.
Although not containing information on steroids, the COSMOS TTC dataset populates all other CTFA (Cosmetic, Toiletries and Fragrance Association) classes that the original Munro TTC dataset lacked, in particular long aliphatic chains, glycol ethers, ketones, and non-ionic alcohol ethoxylate surfactants. Moreover, the COSMOS TTC dataset represents all use types found in the COSMOS Cosmetics Inventory, such as especially skin-care (conditioning/ moisturisers), emulsifiers, perfuming (fragrances), hair dyes, colorants, and UV absorbers/ filters, antimicrobials, vitamins, and plasticisers.
Furthermore, the COSMOS TTC dataset showed a good representation of the Cosmetics Inventory regarding physico-chemical properties and was therefore considered to be suitable for investigating the applicability of the TTC approach to cosmetics.
Modelling chronic, low dose chemical exposure involves a complex series of biological effects. Synergistic workflows for the prediction of repeated dose toxicity to humans for cosmetics are required to mimic the complexity of chronic toxicity. Therefore, the aim of the COSMOS project is to provide an alternative assessment strategy by developing computational workflows including different models and approaches.
Innovative toxicity prediction strategies based on chemical categories, read-across and (Q)SARs should be related to key events in Adverse Outcome Pathways (AOPs).
Categories of similar compounds can be formed using reactive fragments, associated with known mechanisms of toxicity. The data available for the compounds in the category may be used for read-across to predict missing toxicological data.
Mechanism-based profilers have been developed and coded as SMARTS patterns to define chemotypes, allowing the grouping of similar compounds and searching of sets of data, e.g. data collected in the COSMOS database.
Replacing testing for chronic, organ level, toxicity elicited by chemicals requires a paradigm shift in thinking, belief and understanding of the traditional role of the 3Rs. To make a success of modern technologies and the spirit of 21st Century toxicology, frameworks are being proposed to bring together information on biochemical pathways and their perturbation which may result in a biologically adverse effect, commencing with the molecular initiating event (MIE), to downstream effects of organ level/organism or population level effects, as shown in Figure 1. A mode of action framework, which may be formalised into an adverse outcome pathway (AOP), is such an approach (Ankley et al, 2010).
This knowledge, of associating the MIE(s) to specific adverse outcomes, can consequently be used in the production of endpoint specific categories, supported by data from in vitro / in vivo studies. Due to the identifiable chemistry associated to the MIE it provides a means of classifying the ‘domain’ of the AOP and hence defining chemical space of the AOP.
The advantage of the AOP approach is that it provides a transparent link from chemistry to toxicological effect.
COSMOS supports the development and promotion of Adverse Outcome Pathways (AOPs):
The role of COSMOS in AOP development is in particular to organise the chemistry involved in the process.
Within the SEURAT-1Cluster there is an opportunity to collect AOPs for organ level toxicity, with an emphasis on liver toxicity. linking also into other projects to capture the information (e.g. Effectopedia, OHT, …). The important consideration will be that this is seen as a means of compiling information from disparate sources and commercial sectors into a single coherent platform.
Within COSMOS chemoinformatic methods are being applied to gain information about “groups” of molecules that may cause these effects. These groupings are then supported by mechanistic understanding. The mechanistic understanding is being put within the AOPconcept.
Structural alerts are being developed for key events allowing mechanistic interpretation, e.g. covalent protein / DNAbinding.
The first AOPs to be developed will be focused on hepatotoxicity.
PBPK modelling / biokinetics can be used to determine the internal exposure (dose at target organ level) necessary for eliciting the effect.
COSMOS will thus help to identify highly targeted in vitro / in chemico assays that could be developed and used to provide evidence to support the pathways.
The COSMOS/SEURAT-1 AOPs developed will be provided as freely available tools within KNIME workflows.
Figure1. Summary of the steps within an Adverse Outcome Pathway illustrated with examples.
Ankley GT,Bennett RS, Erickson RJ, Hoff DJ, Hornung MW, Johnson RD, Mount DR, Nichols JW,Russom CL, Schmieder PK, Serrano JA, Tietge JE, Villeneuve DL (2010) AdverseOutcome Pathways: A Conceptual Framework to Support Ecotoxicology Research and Risk Assessment. Environmental Toxicology and Chemistry 29: 730-741