A Defined Approach to Skin Sensitisation: Integrating Derek Nexus with In Chemico/In Vitro Assays Based on Exclusion Criteria
Presented by Dr Donna Macmillan at the 2018 Pan-American Conference for Alternative Methods.
Skin sensitisation is traditionally assessed using in vivo assays, however, there is a drive to reduce, refine and replace the use of animals partly due to legislation such as REACH and the TCSA, both of which aim to restrict their use and use alternatives where possible. To this end, several in chemico/in vitro assays been developed to measure specific key events in the skin sensitisation adverse outcome pathway (AOP).
It is generally accepted that these cannot be used in isolation to replace in vivo assays and instead multiple information sources (e.g. in chemico/in vitro assays, molecular descriptors, in silico tools) should be combined in a defined approach to maximise AOP coverage and improve the predictivity of skin sensitisation potential.
Derek Nexus, an expert knowledge-based in silico tool, covers the whole skin sensitisation AOP as alerts are based on mouse, guinea pig, and human data whilst also taking any necessary transformation required by a chemical (prehapten/prohapten) prior to sensitising into account. Consequently, information from Derek was used in combination with data from in chemico/in vitro assays in a defined approach.
The defined approach described uses a set of exclusion criteria based on known limitations and uncertainties to suggest to a user the most relevant assay(s) for the chemical of concern.
A decision tree was then constructed which assigns the chemical of concern as a sensitiser or non-sensitiser. Non-sensitisers are given a Basketter potency category of 5/6 and as GHS not classified. Sensitisers are then assessed for human risk, i.e. a Basketter category of between 1 and 4 and a GHS classification of 1A or 1B. This is achieved using a k-nearest neighbour methodology where LLNA data is augmented with human risk classifications.
A dataset with in vivo, in chemico and in vitro data from 210 publicly available chemicals was used to evaluate the defined approach. The correct Basketter potency category (Basketter category 1-4, 5/6) was predicted for 64% and the correct GHS classification (1A, 1B, no classification) predicted for 77% of the dataset. Future work will focus on accessing new in chemico/in vitro data to assess the defined approach using a novel test set.