Approaches for read-across in chemical risk assessment

Summary

Read-across exploits information on structurally related (similar) analogues to derive hypotheses about the activity of the new chemical and hence predict its toxicity without experimental testing.  Large existing databases on traditional toxicological endpoints and mechanisms of action are available that can be searched by data mining and cheminformatics tools (a selection is presented in the report).  In addition, high-information-content techniques such as ̒omics (toxicogenomics and metabolomics in particular) can be utilised to generate and test these hypotheses, notably about the mechanism of action.  Examples are given in the report for phthalates, oestrogens and skin sensitisers.

There is scope for improvement of the heuristics of analogue identification and hypothesis generation.  Furthermore, real examples of using high-information-content data are needed to support read-across, e.g. to provide a biology based rationale for chemical grouping.

Background

Traditional toxicity testing methods are resource-intensive.  With the advent of legislation such as REACH that mandates the development of comprehensive toxicity information for large numbers of chemicals, it is clear that current methods cannot keep pace with the demand for information.  Read-across approaches provide a practical alternative for the development of information on chemicals that are structurally related to already tested compounds.  Read-across methods are supported by the large database of toxicology studies that have been conducted over the past 50+ years.  These data have been compiled into formats that can be searched by chemical structure / substructure.  Heuristics that identify appropriate structure search strategies, and which can be used to classify the suitability of analogues, have been developed for some aspects of toxicity, most notably toxicity that is dependent upon reactive chemistry.  Some rules have also been developed for receptor affinity (e.g. the oestrogens receptor), although the development of heuristics for weak interactions such as receptor binding and enzyme inhibition is an area for further research.

In many cases, the outcome of a structure search is sufficiently robust that it can be viewed as definitive.  However, in other cases the result can be viewed as a hypothesis that requires further testing.  Toxicogenomics, metabolomics, and other high-information-content technologies provide an opportunity to test such hypotheses rapidly and completely with a minimum of animals.  They also provide the potential to rapidly populate toxicity databases with information that will both augment the traditional toxicology information and facilitate the discovery of chemical classes based on biological activity in addition to chemical structure.

The objective is to improve the process of read-across by the development of new heuristics that support the identification of appropriate analogues, the development of high-information-content approaches to test the results of read-across and to augment databases with enough information that chemical classes can be identified by similar biological activity, not just chemical structure.