- Publication Date:
- Publication Type:Presentation
- Scientific Area:
- Industry Type:
An adverse outcome pathway framework to support the assessment of DART liabilities of compounds
Traditional developmental and reproductive toxicity (DART) safety assessments are costly and time-consuming due to the use of large numbers of animals involved. Alternative approaches to such assessments are receiving much attention, in order to improve the efficiency and increase the human relevance of DART safety assessments. Adverse outcome pathways (AOPs) are able to store relevant knowledge to ground alternative mechanistic approaches and provide context of their results to risk assessors. AOPs can be extended into larger ontologies, where assays and predictive in silico models can be mapped to an AOP network. The resulting predictive framework can support integrated approaches to testing and assessment (IATA) for DART endpoints. Putative molecular initiating events (MIEs) for DART were identified within expert rule-based in silico prediction systems, publications and from consultations with Lhasa Limited members. AOPs stemming from the identified MIEs were synthesised through literature review and linked to endpoints of regulatory significance, including embryo-foetal toxicity and fertility toxicity. Assays which model events in the AOPs were identified and both the assay and their results were mapped to the AOP. Linear AOPs were aggregated into an AOP network and in silico models were appended to relevant key events. The resulting framework of models was tested using datasets comprised of in vivo toxicity studies. The construction of an AOP network facilitated the centralisation of evidence which support adverse outcome pathways leading to DART including mechanisms involving the disruption of thyroid hormone signalling. Alongside literature review, data harvesting generated highly curated in vivo toxicity datasets and in vitro assay datasets. The datasets, their assays and their subsequently trained models, were mapped to relevant events in the AOP network, producing the DART AOP framework. Validating the predictive framework against toxicity datasets, demonstrated that the suite of models had an increased sensitivity, compared to an expert model trained solely on toxicity data. Organising knowledge, data and models into an AOP framework, can allow experts to intuitively explore the relevant properties of compounds present in the network, which are similar to the chemical under assessment. Such an approach will also facilitate the generation of future IATA using alternative methods for DART endpoints.