Development and application of QSAR models for mechanisms related to endocrine disruption.

Humans are daily exposed to a wide variety of man-made chemicals through food, consumer products, water, air inhalation etc. For the main part of these chemicals no or only very limited information is available on their potential to cause endocrine disruption. Traditionally such information has been derived from animal studies, which are time-consuming, expensive and subject to ethical issues. For these reasons alternative methods such as cell culture studies and non-testing approaches such as quantitative structure-activity relationships (QSARs) are of high value as they can provide information on the mode of action of chemicals in a faster and cheaper way. The main purpose in this PhD project was to develop QSAR models for mechanisms related to endocrine disruption and apply the models to predict 10,000s of chemicals to which humans are potentially exposed. The first part of the thesis is a background section, comprising 1) an introduction to the endocrine system with a focus on thyroid hormones (THs) and their essential function in neurodevelopment as well as a description of how chemicals may interference with endocrine mechanisms and cause adverse effects, 2) an introduction to the applied methods to develop QSARs, and 3) an introduction to regulatory toxicology including the acceptance of predictions from QSARs under the European chemicals regulation, REACH. Following the background section, the four projects of the thesis are described. The first three projects focus on the development of QSARs for mechanisms that can affect TH levels: Thyroperoxidase (TPO) inhibition, Pregnane X receptor (PXR) activation, and Aryl hydrocarbon receptor (AhR) activation. TPO is an enzyme essential in the synthesis of THs, and both PXR and AhR are important regulators of enzymes involved in the turnover of THs and other hormones. The fourth project was part of a large international QSAR collaboration, CERAPP, in which a QSAR model for estrogen receptor (ER) agonism was developed, and used to predict 32,197 CERAPP chemicals. All models in the four projects were validated to assess how good they are at making correct predictions, and they all showed good predictive performance. The QSAR models were used to predict 72,524 REACH substances, and they were able to predict between 38,114 to 53,433 of these substances. To conclude, the QSAR models developed in this PhD project can provide important information on the 10,000s of chemicals in our surroundings. The predictions can for example be used for prioritizing chemicals for further evaluation, aid in chemical assessments, grouping approaches, and drug development as well as in the generation of new hypotheses on mode of actions in adverse health outcomes