In our modern society we are exposed to many natural and synthetic chemicals. The assessment of chemicals with regard to human safety is difficult but nevertheless of high importance. Beside clinical studies, which are restricted to potential pharmaceuticals only, most toxicity data relevant for regulatory decision-making are based on in vivo data. Due to the ban on animal testing of cosmetic ingredients in the European Union, alternative approaches, such as in vitro and in silico tests, have become more prevalent. In this thesis existing non-testing approaches (i.e. studies without additional experiments) have been extended, e.g. QSAR models, and new non-testing approaches, e.g. in vitro data supported structural alert systems, have been created. The main aspect of the thesis depends on the determination of data quality, improving modelling performance and supporting Adverse Outcome Pathways (AOPs) with definitions of structural alerts and physico-chemical properties. Furthermore, there was a clear focus on the transparency of models, i.e. approaches using algorithmic feature selection, machine learning etc. have been avoided. Furthermore structural alert systems have been written in an understandable and transparent manner. Beside the methodological aspects of this work, cosmetically relevant examples of models have been chosen, e.g. skin penetration and hepatic steatosis. Interpretations of models, as well as the possibility of adjustments and extensions, have been discussed thoroughly. As models usually do not depict reality flawlessly, consensus approaches of various non-testing approaches and in vitro tests should be used to support decision-making in the regulatory context. For example within read-across, it is feasible to use supporting information from QSAR models, docking, in vitro tests etc. By applying a variety of models, results should lead to conclusions being more usable/acceptable within toxicology. Within this thesis (and associated publications) novel methodologies on how to assess and employ statistical data quality and how to screen for potential liver toxicants have been described. Furthermore computational tools, such as models for skin permeability and dermal absorption, have been created.

Submitted by Luciana Ferreira (lucgeral@gmail.com) on 2017-06-12T15:19:44Z No. of bitstreams: 2 Dissertação - Vinicius de Medeiros Alves - 2014.pdf: 3082084 bytes, checksum: da4838d5fe24841429f43de84204d98a (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) / Approved for entry into archive by Luciana Ferreira (lucgeral@gmail.com) on 2017-06-12T15:21:40Z (GMT) No. of bitstreams: 2 Dissertação - Vinicius de Medeiros Alves - 2014.pdf: 3082084 bytes, checksum: da4838d5fe24841429f43de84204d98a (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) / Made available in DSpace on 2017-06-12T15:21:40Z (GMT). No. of bitstreams: 2 Dissertação - Vinicius de Medeiros Alves - 2014.pdf: 3082084 bytes, checksum: da4838d5fe24841429f43de84204d98a (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) Previous issue date: 2017-05-12 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - CAPES / Introduction: Skin sensitization is a major environmental and human health hazard. Although many chemicals have been evaluated in humans, there have been no efforts to model these data to date. Skin sensitization is commonly evaluated using structural alerts. However, there has been a growing concern that alerts disproportionally flag too many chemicals as toxic, which questions their reliability as toxicity markers. The main goal of this thesis was to develop and apply new cheminformatics methods to predict skin sensitization of chemical compounds that lack experimental data. Methodology: It has been compiled, curated, analyzed, and compared the available human data and the murine (performed in mice) animal model data, named LLNA (local lymph node assay). Using these data, it was developed reliable computational models and applied them for virtual screening of chemical libraries to identify putative skin sensitizers. It was developed a freely accessible web-based application for the identification of potential skin sensitizers. In addition, it was demonstrated that contrary to the common perception of QSAR models as “black boxes” they can be used to identify statistically significant chemical substructures (QSAR-based alerts) that influence toxicity. Results and discussion: The overall concordance between murine LLNA and human skin sensitization responses for a set of 135 unique chemicals was low (R = 28-43%), although several chemical classes had high concordance. We have succeeded to develop predictive QSAR models of all available human data with the external correct classification rate of 71%. A consensus model integrating concordant QSAR predictions and LLNA results afforded a higher correct classification rate of 82% but at the expense of the reduced external dataset coverage (52 %). We used the developed QSAR models for virtual screening of CosIng database and identified 1,061 putative skin sensitizers; for seventeen of these compounds, we found published evidence of their skin sensitization effects. The developed Pred-Skin web app (http://www.labmol.com.br/predskin/) is based on binary QSAR models of human (109 compounds) and LLNA (515 compounds) data with good external correct classification rate (70-81% and 72-84%, respectively). It is also included a multiclass potency model based on LLNA data (accuracy ranging between 73-76%). Conclusions: Models reported herein provide more accurate alternative to LLNA testing for human skin sensitization assessment across diverse chemical data. In addition, they can also be used to guide the structural optimization of toxic compounds to reduce their skin sensitization potential. The Pred-Skin web app is a fast, reliable, and user-friendly tool for early assessment of chemically-induced skin sensitization. A new approach that synergistically integrates structural alerts and rigorously validated QSAR models for a more transparent and accurate safety assessment of new chemicals was also proposed. / Introdução: A sensibilização cutânea é um importante parâmetro de avaliação de toxicidade humana e ambiental. Embora muitos compostos tenham sido avaliados em seres humanos, não foi reportado até o momento modelos de QSAR (do inglês, quantitative structure-activity relationships) gerados com esses dados. Comumente, a sensibilização cutânea é avaliada computacionalmente usando-se alertas estruturais. No entanto, tem havido uma preocupação crescente de que alertas sinalizam a maioria dos compostos como tóxicos, o que questiona sua confiabilidade como marcadores de toxicidade. O objetivo geral do presente trabalho foi desenvolver e aplicar novos métodos de quimioinformática para predizer a sensibilização cutânea de compostos químicos que carecem de dados experimentais. Metodologia: Foram compilados, preparados, analisados e comparados os dados de sensibilização cutânea de pele humana e do modelo animal murino (realizado em camundongos), denominado LLNA (local lymph node assay). Modelos de QSAR foram desenvolvidos utilizando esses dados e aplicados para a triagem de quimiotecas virtuais para identificar potenciais sensibilizadores. Foi desenvolvido um aplicativo gratuito para a identificação de potenciais sensibilizadores cutâneos. Além disso, foi demonstrado que modelos de QSAR podem ser usados para identificar subestruturas químicas estatisticamente significativas (alertas estruturais baseados em QSAR) que influenciam a toxicidade. Resultados e discussão: A concordância global (R) entre respostas de sensibilização cutânea humana e murina para um conjunto de 135 substâncias químicas únicas foi baixa (R = 28-43%), embora várias classes químicas apresentassem alta concordância. Foi possível desenvolver modelos de QSAR preditivos com taxa de classificação correta externa de 71%. Um modelo de consenso que integrava predições concordantes de QSAR e dados de LLNA proporcionaram uma acurácia 82%. Utilizou-se os modelos de QSAR desenvolvidos para a triagem virtual da base de dados CosIng e foram identificados 1061 potenciais sensibilizadores cutâneos. Para dezessete desses compostos, encontrou-se evidências publicadas de seus efeitos de sensibilização cutânea em seres humanos. O aplicativo desenvolvido, Pred-Skin (http://www.labmol.com.br/predskin/), baseia-se em modelos de QSAR classificatórios de dados humanos (109 compostos) e murinos (515 compostos) com boa taxa de classificação correta externa (70-81% e 72-84%, respectivamente). Esse aplicativo também possui um modelo de multiclassificatório desenvolvido com dados de LLNA (precisão que varia entre 73-76%). Conclusões: Os modelos de QSAR desenvolvidos forneceram uma alternativa mais precisa do que o modelo animal para avaliação da sensibilização cutânea humana. Além disso, a interpretação dos modelos de QSAR permitem orientar a otimização estrutural de compostos tóxicos para reduzir o potencial de toxicidade. O aplicativo Pred-Skin é uma ferramenta rápida, confiável e de fácil utilização para a avaliação da sensibilização cutânea de compostos químicos. Foi também proposta uma nova abordagem que integra sinergicamente alertas estruturais e modelos de QSAR rigorosamente validados para uma avaliação de toxicidade mais transparente e precisa de novos produtos químicos.