Integration of toxicity data from experiments and non-testing methods within a weight of evidence procedure

• Main Author: Golbamaki Bakhtyari, Azadi
• Format: Dissertation
• Language: English
• Published: Open University 2018
• DOI: 10.21954/ou.ro.0000d93f

Assessment of human health and environmental risk is based on multiple sources of information, requiring the integration of the lines of evidence in order to reach a conclusion. There is an increasing need for data to fill the gaps and new methods for the data integration. From a regulatory point of view, risk assessors take advantage of all the available data by means of weight of evidence (WOE) and expert judgement approaches to develop conclusions about the risk posed by chemicals and also nanoparticles. The integration of the physico-chemical properties and toxicological effects shed light on relationships between the molecular properties and biological effects, leading us to non-testing methods. (Quantitative) structure-activity relationship ((Q)SAR) and read-across are examples of non-testing methods. In this dissertation, (i) two new structure-based carcinogenicity models, (ii) ToxDelta, a new read-across model for mutagenicity endpoint and (iii) a genotoxicity model for the metal oxide nanoparticles are introduced. Within the latter section, best professional judgement method is employed for the selection of reliable data from scientific publications to develop a data base of nanomaterials with their genotoxicity effect. We developed a decision tree model for the classification of these nanomaterials. The (Q)SAR models used in qualitative WOE approaches mainly lack transparency resulting in risk estimates needing quantified uncertainties. Our two structure-based carcinogenicity models, provide transparent reasoning in their predictions. Additionally, ToxDelta provides better supported techniques in read-across terms based on the analysis of the differences of the molecules structures. We propose a basic qualitative WOE framework that couples the in silico models predictions with the inspections of the similar compounds. We demonstrate the application of this framework to two realistic case studies, and discuss how to deal with different and sometimes conflicting data obtained from various in silico models in qualitative WOE terms to facilitate structured and transparent development of answers to scientific questions.