Chemical exposures affect the environment and may lead to adverse outcomes in its organisms. Omics-based approaches, like standardised microarray experiments, have expanded the toolbox to monitor the distribution of chemicals and assess the risk to organisms in the environment. The resulting complex data have extended the scope of toxicological knowledge bases and published literature. A plethora of computational approaches have been applied in environmental toxicology considering systems biology and data integration. Still, the complexity of environmental and biological systems given in data challenges investigations of exposure-related effects. This thesis aimed at computationally linking chemical exposure to biological effects on the molecular level considering sources of complex environmental data. The first study employed data of an omics-based exposure study considering mixture effects in a freshwater environment. We compared three data-driven analyses in their suitability to disentangle mixture effects of chemical exposures to biological effects and their reliability in attributing potentially adverse outcomes to chemical drivers with toxicological databases on gene and pathway levels. Differential gene expression analysis and a network inference approach resulted in toxicologically meaningful outcomes and uncovered individual chemical effects — stand-alone and in combination. We developed an integrative computational strategy to harvest exposure-related gene associations from environmental samples considering mixtures of lowly concentrated compounds. The applied approaches allowed assessing the hazard of chemicals more systematically with correlation-based compound groups. This dissertation presents another achievement toward a data-driven hypothesis generation for molecular exposure effects. The approach combined text-mining and deep learning. The study was entirely data-driven and involved state-of-the-art computational methods of artificial intelligence. We employed literature-based relational data and curated toxicological knowledge to predict chemical-biomolecule interactions. A word embedding neural network with a subsequent feed-forward network was implemented. Data augmentation and recurrent neural networks were beneficial for training with curated toxicological knowledge. The trained models reached accuracies of up to 94% for unseen test data of the employed knowledge base. However, we could not reliably confirm known chemical-gene interactions across selected data sources. Still, the predictive models might derive unknown information from toxicological knowledge sources, like literature, databases or omics-based exposure studies. Thus, the deep learning models might allow predicting hypotheses of exposure-related molecular effects. Both achievements of this dissertation might support the prioritisation of chemicals for testing and an intelligent selection of chemicals for monitoring in future exposure studies.:Table of Contents ... I Abstract ... V Acknowledgements ... VII Prelude ... IX 1 Introduction 1.1 An overview of environmental toxicology ... 2 1.1.1 Environmental toxicology ... 2 1.1.2 Chemicals in the environment ... 4 1.1.3 Systems biological perspectives in environmental toxicology ... 7 Computational toxicology ... 11 1.2.1 Omics-based approaches ... 12 1.2.2 Linking chemical exposure to transcriptional effects ... 14 1.2.3 Up-scaling from the gene level to higher biological organisation levels ... 19 1.2.4 Biomedical literature-based discovery ... 24 1.2.5 Deep learning with knowledge representation ... 27 1.3 Research question and approaches ... 29 2 Methods and Data ... 33 2.1 Linking environmental relevant mixture exposures to transcriptional effects ... 34 2.1.1 Exposure and microarray data ... 34 2.1.2 Preprocessing ... 35 2.1.3 Differential gene expression ... 37 2.1.4 Association rule mining ... 38 2.1.5 Weighted gene correlation network analysis ... 39 2.1.6 Method comparison ... 41 Predicting exposure-related effects on a molecular level ... 44 2.2.1 Input ... 44 2.2.2 Input preparation ... 47 2.2.3 Deep learning models ... 49 2.2.4 Toxicogenomic application ... 54 3 Method comparison to link complex stream water exposures to effects on the transcriptional level ... 57 3.1 Background and motivation ... 58 3.1.1 Workflow ... 61 3.2 Results ... 62 3.2.1 Data preprocessing ... 62 3.2.2 Differential gene expression analysis ... 67 3.2.3 Association rule mining ... 71 3.2.4 Network inference ... 78 3.2.5 Method comparison ... 84 3.2.6 Application case of method integration ... 87 3.3 Discussion ... 91 3.4 Conclusion ... 99 4 Deep learning prediction of chemical-biomolecule interactions ... 101 4.1 Motivation ... 102 4.1.1Workflow ...105 4.2 Results ... 107 4.2.1 Input preparation ... 107 4.2.2 Model selection ... 110 4.2.3 Model comparison ... 118 4.2.4 Toxicogenomic application ... 121 4.2.5 Horizontal augmentation without tail-padding ...123 4.2.6 Four-class problem formulation ... 124 4.2.7 Training with CTD data ... 125 4.3 Discussion ... 129 4.3.1 Transferring biomedical knowledge towards toxicology ... 129 4.3.2 Deep learning with biomedical knowledge representation ...133 4.3.3 Data integration ...136 4.4 Conclusion ... 141 5 Conclusion and Future perspectives ... 143 5.1 Conclusion ... 143 5.1.1 Investigating complex mixtures in the environment ... 144 5.1.2 Complex knowledge from literature and curated databases predict chemical- biomolecule interactions ... 145 5.1.3 Linking chemical exposure to biological effects by integrating CTD ... 146 5.2 Future perspectives ... 147 S1 Supplement Chapter 1 ... 153 S1.1 Example of an estrogen bioassay ... 154 S1.2 Types of mode of action ... 154 S1.3 The dogma of molecular biology ... 157 S1.4 Transcriptomics ... 159 S2 Supplement Chapter 3 ... 161 S3 Supplement Chapter 4 ... 175 S3.1 Hyperparameter tuning results ... 176 S3.2 Functional enrichment with predicted chemical-gene interactions and CTD reference pathway genesets ... 179 S3.3 Reduction of learning rate in a model with large word embedding vectors ... 183 S3.4 Horizontal augmentation without tail-padding ... 183 S3.5 Four-relationship classification ... 185 S3.6 Interpreting loss observations for SemMedDB trained models ... 187 List of Abbreviations ... i List of Figures ... vi List of Tables ... x Bibliography ... xii Curriculum scientiae ... xxxix Selbständigkeitserklärung ... xliii

Indiana University-Purdue University Indianapolis (IUPUI) / This dissertation presents a body of research that attempts to tackle the ‘overfitting’ problem for gene signature and biomarker development in two different aspects (mechanistically and computationally). In achievement of a deeper understanding of cancer molecular mechanisms, this study presents new approaches to derive gene signatures for various biological phenotypes, including breast cancer, in the context of well-defined and mechanistically associated biological pathways. We identified the pattern of gene expression in the cell cycle pathway can indeed serve as a powerful biomarker for breast cancer prognosis. We further built a predictive model for prognosis based on the cell cycle gene signature, and found our model to be more accurate than the Amsterdam 70-gene signature when tested with multiple gene expression datasets generated from several patient populations. Aside from demonstrating the effectiveness of dimensionality reduction, phenotypic dissection, and prognostic or diagnostic prediction, this approach also provides an alternative to the current methodology of identifying gene expression markers that links to biological mechanism. This dissertation also presents the development of a novel feature selection algorithm called Predictive Power Estimate Analysis (PPEA) to computationally tackle on overfitting. The algorithm iteratively apply a two-way bootstrapping procedure to estimate predictive power of each individual gene, and make it possible to construct a predictive model from a much smaller set of genes with the highest predictive power. Using DrugMatrix™ rat liver data, we identified genomic biomarkers of hepatic specific injury for inflammation, cell death, and bile duct hyperplasia. We demonstrated that the signature genes were mechanistically related to the phenotype the signature intended to predict (e.g. 17 out of top 20 genes for inflammation selected by PPEA were members of NF-kB pathway, which is a key pre-inflammatory pathway for a xenobiotic response). The top 4 gene signature for BDH has been further validated by QPCR in a toxicology lab. This is important because our results suggest that the PPEA model not largely deters the over-fitting problem, but also has the capability to elucidate mechanism(s) of drug action and / or of toxicity.

Avec les nouvelles avancées en biologie et toxicologie, on constate de plus en plus la complexité des mécanismes et le grand nombre de voies de toxicité. Les concepts de ‘biologie systémique’ (SB) et de ‘voies des effets indésirables’ (adverse outcome pathway, AOP) pourraient être des outils appropriés pour l’étude de la toxicologie à ces niveaux de complexité élevés. Le point central du travail de cette thèse est le développement d’un modèle de SB du rôle de la voie de signalisation Nrf2 dans le contrôle du stress oxydant. Pour la calibration de ce modèle avec des données expérimentales (exposition des cellules rénales RPTEC/TERT1 à différentes doses de bromate de potassium), plusieurs cycles de proposition/vérification d’hypothèses ont progressivement contribué à l’ajout de nouvelles réactions. Ces nouvelles hypothèses (par exemple : action directe du bromate de potassium sur le DCF, atténuation de la fluorescence du DCF avec le temps, etc.) devraient être confirmées par de futures expérimentations. Ce modèle de SB a été ensuite utilisé pour la quantification d’un AOP de l’insuffisance rénale chronique et comparé à deux autres approches: l’utilisation de modèles statistiques empiriques et celle d’un réseau Bayésien dynamique. Les calibrations des paramètres ont été effectuées par chaînes de Markov simulées MCMC avec le logiciel GNU MCSim avec une quantification des incertitudes associées aux prédictions. Même si la mise au point du modèle SB a été une tâche complexe, la compréhension de la biologie qu’offre ce modèle n’est pas accessible aux deux autres approches. Nous avons aussi évalué les interactions entre Nrf2 et deux autres voies de toxicité, AhR et ATF4, dans le cadre d’une analyse utilisant des données de toxico-génomique provenant de trois projets différents. Les résultats de cette dernière analyse suggèrent d’ajouter au modèle SB de Nrf2 la co-activation par AhR de plusieurs gènes (par exemple, HMOX1, SRXN1 et GCLM) ainsi que d’associer (au moins partiellement) à ce modèle la voie ATF4. Malgré leur complexité, les modèles SB constituent un investissement intéressant pour le développement de la toxicologie prédictive. / New understanding of biology shows more and more that the mechanisms that underlie toxicity are complex and involve multiple biological processes and pathways. Adverse outcome pathways (AOPs) and systems biology (SB) can be appropriate tools for studying toxicology at this level of complexity. This PhD thesis focuses on the elaboration of a SB model of the role of the Nrf2 pathway in the control of oxidative stress. The model’s calibration with experimental data (obtained with RPTEC/TERT1 renal cells exposed to various doses of potassium bromate) comprised several rounds of hypotheses stating/verification, through which new reactions were progressively added to the model. Some of these new hypotheses (e.g., direct action of potassium bromate on DCF, bleaching of DCF with time, etc.) could be confirmed by future experiments. Considered in a wider framework, this SB model was then evaluated and compared to two other computational models (i.e., an empirical dose-response statistical model and a dynamic Bayesian model) for the quantification of a ‘chronic kidney disease’ AOP. All parameter calibrations were done by MCMC simulations with the GNU MCSim software with a quantification of uncertainties associated with predictions. Even though the SB model was indeed complex to conceive, it offers insight in biology that the other approaches could not afford. In addition, using multiple toxicogenomic databases; interactions and cross-talks of the Nrf2 pathway with two other toxicity pathways (i.e., AhR and ATF4) were examined. The results of this last analysis suggest adding new AhR contribution to the control of some of the Nrf2 genes in our SB model (e.g., HMOX1, SRXN1 and GCLM), and integrating in it description of the ATF4 pathway (partially at least). Despites their complexity, precise SB models are precious investments for future developments in predictive toxicology.

General toxicity evaluation and risk assessment for human exposure is essential when developing new pharmaceuticals and chemicals. Developmental toxicology is an important part of this risk assessment which consumes large resources and many laboratory animals. The prediction of developmental toxicity could potentially be assessed in vitro using embryo-derived pluripotent stem cells for lead characterization and optimization. This thesis explored the potential of short-time assays with pluripotent stem cells of embryonic origin using toxicogenomics. Three established pluripotent stem cell lines; P19 mouse embryonal carcinoma (EC) cells, R1 mouse embryonic stem (mES) cells, and SA002 human embryonic stem (hES) cells were used in the studies. Valproic acid (VPA), an antiepileptic drug which can cause the neural tube defects spina bifida in human and exencephaly in mouse, was used together with microarrays to investigate the global transcriptional response in pluripotent stem cells using short-time exposures (1.5 - 24 h). In addition to VPA, three closely related VPA analogs were tested, one of which was not teratogenic in mice. These analogs also differed in their ability to inhibit histone deacetylase (HDAC) allowing this potential mechanism of VPA teratogenicity to be investigated. The results in EC cells indicated a large number of genes to be putative VPA targets, many of which are known to be involved in neural tube morphogenesis. When compared with data generated in mouse embryos, a number of genes emerged as candidate in vitro markers of VPA-induced teratogenicity. VPA and its teratogenic HDAC inhibiting analog induced major and often overlapping deregulation of genes in mES cells and hES cells. On the other hand, the two non-HDAC inhibiting analogs (one teratogenic and one not) had only minor effects on gene expression. This indicated that HDAC inhibition is likely to be the major mechanism of gene deregulation induced by VPA. In addition, a comparison between human and mouse ES cells revealed an overlap of deregulated genes as well as species specific deregulated genes.

The conventional battery for genotoxicity testing is not well-suited to assessing the large number of chemicals needing evaluation. Traditional in vitro tests lack throughput capacity, provide little mechanistic information, and have poor specificity in predicting in vivo genotoxicity. The Health Canada GeneTox21 research program is developing a multi-endpoint platform for modernized in vitro genotoxicity assessment. The GeneTox21 assays include the TGx-DDI transcriptomic biomarker (i.e., 64-gene expression signature to identify DNA damage-inducing (DDI) substances), the MicroFlow® assay (i.e., a flow cytometry-based micronucleus (MN) test), and the MultiFlow® assay (i.e., a multiplexed flow cytometry-based reporter assay that yields mechanism-of-action (MoA) information). As part of GeneTox21 development, the objective of this study was to investigate the utility of the TGx-DDI transcriptomic biomarker, multiplexed with the MicroFlow® and MultiFlow® assays, as an integrated testing strategy for screening data-poor substances prioritized by Health Canada’s New Substances Assessment and Control Bureau. Human lymphoblastoid TK6 cells were exposed to 3 control and 10 data-poor substances, using a 6-point concentration range. Cells were exposed for 4 hours with or without exogenous metabolic activation. Gene expression profiling was conducted using the targeted TempO-SeqTM assay, and the TGx-DDI classifier was applied to the dataset. Classifications were compared with those based on the MicroFlow® and MultiFlow® assays. Benchmark Concentration (BMC) modeling was used for potency ranking. The results of the integrated hazard calls indicate that five data-poor compounds are genotoxic in vitro, causing DNA damage via a clastogenic MoA, and one is positive via a pan-genotoxic MoA. Two compounds are likely irrelevant positives in the MN test; two are considered possibly genotoxic causing DNA damage via an ambiguous MoA. From quantitative analyses of concentration-response data, we observed nearly identical potency rankings for each assay with two main potency groups being observed. This ranking was maintained when all endpoint BMCs were converted into a single score using the Toxicological Prioritization (ToxPi) approach. Overall, this study contributes to the establishment of a modernized approach for effective genotoxicity assessment and chemical prioritization for further regulatory scrutiny. We conclude that integration of the TGx-DDI biomarker with other GeneTox21 assays is an effective NAM-based strategy for genotoxicity assessment of data-poor compounds.

<p>Developmental toxicology is an important part of preclinical drug toxicology as well as environmental toxicology. Assessing reproductive and developmental toxicity is especially expensive and time demanding, since at least two generations of animals are needed in the tests. In light of this there is a great need for alternative test methods in many areas of developmental toxicity testing.</p><p>The complete set of RNA transcripts in any given organism is called the transcriptome. Proteomics refers to the study of the proteins in a given organism or cell population. The work of this thesis has focused on the use of high throughput screening methods in transcriptomics and proteomics to search for molecular markers of developmental toxicity.</p><p>We have studied the global gene expression effects of the developmentally toxic substance valproic acid (VPA) using microarray technology. Several genes were found that display the same gene expression pattern <i>in vivo</i> using mouse embryos as the pattern seen <i>in vitro</i> using the embryocarcinoma cell line P19. Based on these observations, the gene Gja1 was suggested as one potential molecular marker of VPA induced developmental toxicity and potential marker of histone deacetylase (HDAC) inhibition <i>in vitro</i>. </p><p>Using 2D-DIGE technology, which measures relative protein abundances, the effect of neonatal exposure to the flame retardant PBDE-99 was studied in mouse brain (cortex, hippocampus and striatum) 24 hr after exposure. Differentially expressed proteins in the cortex and the striatum indicate that PBDE-99 may alter neurite outgrowth.</p><p>Finally, we have suggested several improvements in the use of the 2D-DIGE technology. Novel methods for normalizing data were presented, with several advantages compared to existing methods. We have presented a method named DEPPS that makes use of all identified proteins in a dataset to make comprehensive remarks about biological processes affected.</p>

Developmental toxicology is an important part of preclinical drug toxicology as well as environmental toxicology. Assessing reproductive and developmental toxicity is especially expensive and time demanding, since at least two generations of animals are needed in the tests. In light of this there is a great need for alternative test methods in many areas of developmental toxicity testing. The complete set of RNA transcripts in any given organism is called the transcriptome. Proteomics refers to the study of the proteins in a given organism or cell population. The work of this thesis has focused on the use of high throughput screening methods in transcriptomics and proteomics to search for molecular markers of developmental toxicity. We have studied the global gene expression effects of the developmentally toxic substance valproic acid (VPA) using microarray technology. Several genes were found that display the same gene expression pattern in vivo using mouse embryos as the pattern seen in vitro using the embryocarcinoma cell line P19. Based on these observations, the gene Gja1 was suggested as one potential molecular marker of VPA induced developmental toxicity and potential marker of histone deacetylase (HDAC) inhibition in vitro. Using 2D-DIGE technology, which measures relative protein abundances, the effect of neonatal exposure to the flame retardant PBDE-99 was studied in mouse brain (cortex, hippocampus and striatum) 24 hr after exposure. Differentially expressed proteins in the cortex and the striatum indicate that PBDE-99 may alter neurite outgrowth. Finally, we have suggested several improvements in the use of the 2D-DIGE technology. Novel methods for normalizing data were presented, with several advantages compared to existing methods. We have presented a method named DEPPS that makes use of all identified proteins in a dataset to make comprehensive remarks about biological processes affected.

CoordenaÃÃo de AperfeiÃoamento de Pessoal de NÃvel Superior / Atualmente sÃo liberadas para comercializaÃÃo no mercado agrobiotecnolÃgico brasileiro, pelo menos, 12 variedades de algodÃo expressando proteÃnas Cry de Bacillus thruringiensis (Bt) para conferir resistÃncia ao ataque de insetos. Contudo, essas proteÃnas sÃo ativas contra lepidÃpteros, enquanto que a principal praga da cotonicultura brasileira à um coleÃptero, o bicudo-do-algodoeiro (Anthonomus grandis). Neste contexto, tÃm-se feito um grande esforÃo na busca por novas molÃculas Cry ativas contra coleÃpteros. Utilizando tÃcnicas de evoluÃÃo molecular dirigida in vitro, foi desenvolvida a proteÃna mutante Cry8Ka5, que à vÃrias vezes mais tÃxica contra o bicudo do que a proteÃna original, Cry8Ka1, inicialmente descoberta. Dada a sua atividade promissora contra coleÃpteros, Cry8Ka5 està sendo utilizada para o desenvolvimento de um novo algodÃo Bt, bem como de outras culturas Bt de importÃncia econÃmica. No entanto, como parte das exigÃncias legais para a liberaÃÃo comercial de uma planta transgÃnica, essa proteÃna recombinante deve ser testada quanto a sua biosseguranÃa alimentar (BA) a fim de detectar potenciais efeitos alergÃnicos, tÃxicos e/ou antimetabÃlicos. Assim, este trabalho objetivou realizar uma ampla avaliaÃÃo de BA da proteÃna mutante Cry8Ka5 e da proteÃna Cry1Ac, jà incorporada em vÃrias culturas Bt comercializadas. A Ãltima foi utilizada como proteÃna controle ou teste dependendo do ineditismo da abordagem empregada em cada seÃÃo do trabalho. Nesse Ãmbito, a avaliaÃÃo de BA proposta cobriu desde as exigÃncias legais atà a execuÃÃo de metodologias, atà entÃo, nunca propostas para avaliaÃÃo de seguranÃa de consumo de produtos de Bt. Para tanto, este trabalho estruturou-se em quatro capÃtulos. O CapÃtulo 1 objetivou realizar uma revisÃo da literatura acerca do estado da arte das plantas Bt e proteÃnas Cry, bem como fornecer um entendimento geral sobre os conceitos e a legislaÃÃo vigente em BA de plantas GM. Jà o CapÃtulo 2 objetivou avaliar a BA da entomotoxina mutante Cry8Ka5 atravÃs do mÃtodo de duas etapas baseado em pesos de evidÃncia que, por sua vez, atende a todas as recomendaÃÃes da CTNBio. Neste caso, a proteÃna Cry1Ac foi utilizada como controle experimental, pois abordagens similares jà foram realizadas com essa molÃcula. Baseando-se nos mÃtodos oficiais, à possÃvel concluir que nÃo à esperado qualquer risco associado ao consumo da proteÃna Cry8Ka5 devido aos seguintes resultados encontrados: 1. Produtos Bt mostraram ter longo histÃrico de uso seguro; 2. A proteÃna Cry8Ka5 nÃo mostrou similaridade significativa de sua sequÃncia de aminoÃcidos primÃria com proteÃnas alergÃnicas, tÃxicas e/ou antinutricionais; 3. As proteÃnas Cry mostraram possuir modo de aÃÃo bastante compreendido e elevada especificidade contra insetos; 4. Cry8Ka5 foi altamente susceptÃvel à digestÃo em fluÃdo gÃstrico simulado; 5. A proteÃna Cry8Ka5 nÃo causou efeitos adversos relevantes em camundongos (5.000 mg de proteÃna/Kg de peso corpÃreo, via oral, dose Ãnica). Por sua vez, o CapÃtulo 3 objetivou avaliar a proteÃna Cry8Ka5 e a Cry1Ac quanto aos seus efeitos cito- e genotÃxicos em uma sÃrie de testes clÃssicos e alternativos, alÃm de pesquisar seus efeitos antimicrobianos. As proteÃnas nÃo causaram efeitos cito- ou genotÃxicos em linfÃcitos humanos (ambas com CI50 > 1.000 Âg/mL em todos os testes), nÃo promoveram hemÃlise em suspensÃes de eritrÃcitos de humanos (tipos A, B, AB e O), de coelho e rato (ambas com CI50 > 1.000 Âg/mL para todas as espÃcies), enquanto somente a Cry8Ka5 causou alteraÃÃes na topografia de membrana celular de eritrÃcitos humanos do tipo O, na concentraÃÃo de 1.000 Âg/mL, detectadas via microscopia de forÃa atÃmica. AlÃm disso, as proteÃnas Cry8Ka5 e Cry1Ac nÃo apresentaram citotoxicidade elevada contra naÃplios de Artemia sp. (CL50 > 700 e > 1000 Âg/mL, respectivamente) e nÃo inibiram o crescimento de quatro cepas de bactÃrias e quatro de leveduras (ambas com CIM > 1.000 Âg/mL para todos os microrganismos testados). Por fim, o CapÃtulo 4 objetivou avaliar os efeitos das proteÃnas Cry8Ka5 e Cry1Ac e de um âpoolâ de peptÃdeos de cada uma (Cry8Ka5-pep e Cry1Ac-pep, respectivamente), obtidos por digestÃo sequencial in vitro, sobre o perfil de expressÃo gÃnica de cÃlulas de carcinoma de mama MCF-7 e de carcinoma de cÃlon Caco-2, ambas de origem humana. Nenhuma das amostras testadas causou alteraÃÃes no perfil de expressÃo gÃnica das cÃlulas Caco-2 diferenciadas (utilizadas como um modelo de epitÃlio intestinal humano), mesmo quando testadas em alta concentraÃÃo (100 Âg/mL). Nos testes de exposiÃÃo com cÃlulas MCF-7 indiferenciadas, apenas a proteÃna Cry1Ac causou alteraÃÃes significativas no padrÃo de expressÃo gÃnica. Contudo, esse efeito nÃo foi detectado apÃs tratamento dessas cÃlulas com a amostra digerida de Cry1Ac (Cry1Ac-pep), que, por sua vez, mimetiza melhor as condiÃÃes reais de exposiÃÃo Ãs molÃculas Cry. As abordagens experimentais utilizadas nos trÃs Ãltimos capÃtulos mostraram-se bastante Ãteis e, apÃs tudo que foi exposto, foi possÃvel confirmar o carÃter inÃcuo da proteÃna Cry1Ac, bem como concluir que a proteÃna Cry8Ka5 à uma ferramenta biotecnolÃgica segura para o desenvolvimento de plantas de algodÃo Bt para conferir resistÃncia ao ataque do bicudo-do-algodoeiro. / Currently at least 12 varieties of cotton expressing Cry proteins from Bacillus thruringiensis (Bt) are released for sale in Brazil to confer resistance to insect attack. However, these proteins are active against lepidopteran, whereas the major pest of Brazilian cotton is a beetle, the cotton boll weevil (Anthonomus grandis). In this context, great efforts have been dedicated in the search for new Cry molecules active against coleopteran. Using techniques of in vitro directed molecular evolution, the Cry8Ka5 mutant protein was developed, which is several times more toxic against the boll weevil than the original protein, Cry8Ka1, initially discovered. Given its promising activity against coleopteran, Cry8Ka5 is being used for the development of a new Bt cotton, as well as of other Bt crops of economic importance. However, as part of legal requirements for the commercial release of a transgenic plant, the recombinant products should be tested for their food safety (FS) in order to detect potential allergenic, toxic and antimetabolic effects. This study aimed to perform a comprehensive FS assessment of the Cry8Ka5 mutant protein and of the Cry1Ac protein, already incorporated in several commercialized Bt crops. The latter was used as a protein control or test depending on the originality of the approach used in each section of this work. The FS evaluation performed covered not only the legal requirements but also the implementation of methodologies hitherto never proposed to assess the safety of consumption of Bt. For that, this study was structured in four chapters. Chapter 1 aimed to conduct a literature review on the state of the art of Bt plants and Cry proteins, as well as provide a general understanding of the concepts and current legislation in GM plants FS. Chapter 2 aimed to evaluate the FS of Cry8Ka5 mutant entomotoxin through the two-step method based on weights of evidence that, in turn, meets the recommendations of CTNBio. In this section, the Cry1Ac protein was used as experimental control, since similar approaches have already been performed with this molecule. Based on official methods, we conclude that it is not expected any risk associated with consumption of protein Cry8Ka5 due to the following results found: 1. Bt products were shown to have a long history of safe use, 2. The protein showed no significant similarity in its primary amino acid sequence with known allergenic, toxic and/or antinutritional proteins; 3. The mode of action and high specificity against insects of Cry proteins are well known; 4. Cry8Ka5 was highly susceptible to digestion in simulated gastric fluid; 5. The protein Cry8Ka5 caused no significant adverse effects in mice (5,000 mg protein/kg body weight, orally, single dose). Chapter 3 aimed to evaluate the Cry8Ka5 and Cry1Ac proteins as to their cyto- and genotoxicity in a serie of conventional and alternative tests as well as antimicrobial effects. The proteins did not cause cyto- or genotoxic effects in human lymphocytes (both with IC50 > 1,000 Âg/mL in all tests), did not cause hemolysis in human (types A, B, AB and O), rabbit and rat (both with IC50 > 1,000 Âg/mL for all species) rythrocytes, but only Cry8Ka5 caused changes in the topography of the cell membrane of human O type erythrocytes at a concentration of 1,000 Âg/mL, detected by atomic force microscopy. Furthermore, the Cry8Ka5 and Cry1Ac proteins showed no cytotoxicity against Artemia sp. nauplii (LC50 > 700 and >1000 Âg/mL, respectively), and did not inhibit the growth of four strains of bacteria and four of yeasts (both with MIC > 1,000 Âg/mL for all microrganisms tested). Finally, Chapter 4 aimed to evaluate the effects of Cry8Ka5 and Cry1Ac proteins and of a "pool" of peptides of each one (Cry8Ka5-pep and Cry1Ac-pep, respectively), obtained by in vitro sequential digestion, on the gene expression profile of MCF-7 breast carcinoma cells and Caco-2 colon carcinoma cells, both human. No sample caused changes in the gene expression profile of differentiated Caco-2 cells (used as a model for human intestinal epithelium), even when tested at high concentration (100 Âg/mL). In the exposure tests with undifferentiated MCF-7 cells, only Cry1Ac (at 100 Âg/mL) caused significant changes in the gene expression pattern. However, this effect was not detected after treatment of these cells with the sample digested of Cry1Ac Cry1Ac-pep), which, in turn, better mimic actual conditions of exposure to Cry molecules. The experimental approaches used in the last three chapters were quite useful and, after all the above, it was possible to confirm the harmless nature of the Cry1Ac protein and conclude that the Cry8Ka5 protein is a safe biotechnological tool for the development of Bt cotton plants to confer resistance against the cotton boll weevil attack.

L’émergence des Mégadonnées (« Big Data ») en biologie moléculaire, surtout à travers la transcriptomique, a révolutionné la façon dont nous étudions diverses disciplines telles que le processus de développement du médicament ou la recherche sur le cancer. Ceci fut associé à un nouveau concept, la médecine de précision, dont le principal but est de comprendre les mécanismes moléculaires entraînant une meilleure réponse thérapeutique chez le patient. Cette thèse est à mi-chemin entre les études pharmaco — et toxicogénomiques expérimentales, et les études cliniques et translationnelles. Le but de cette thèse est surtout de montrer le potentiel et les limites de ces jeux de données et leur pertinence pour la découverte de biomarqueurs de réponse ainsi que la compréhension des mécanismes d’action/toxicité de médicaments, en vue d’utiliser ces informations à des fins thérapeutiques. L’originalité de cette thèse réside dans son approche globale pour analyser les plus larges jeux de données pharmaco/toxicogénomiques publiés à ce jour et ceci pour : 1) Aborder la notion de biomarqueurs de réponse aux médicaments en pharmacogénomique du cancer, en étudiant les facteurs discordants entre deux grandes études publiées en 2012; 2) Comprendre le mécanisme d’action des médicaments et construire une taxonomie performante en utilisant une approche intégrative; et 3) Créer un répertoire toxicogénomique à partir des hépatocytes humains, exposés à différentes classes de médicaments et composés chimiques. Mes contributions principales sont les suivantes : • J’ai développé une approche bioinformatique pour étudier les facteurs discordants entre deux grandes études pharmacogénomiques et suggérées que les différences observées émergeaient plutôt de l’absence de standardisation des mesures pharmacologiques qui pourrait limiter la validation de biomarqueurs de réponse aux médicaments. • J’ai implémenté une approche bioinformatique qui montre la supériorité de l’intégration tenant en compte des différents paramètres pour les médicaments (structure, cytotoxicité, perturbation du transcriptome) afin d’élucider leur mécanisme d’action (MoA). • J’ai développé un pipeline bioinformatique pour étudier le niveau de conservation des mécanismes moléculaires entre les études toxicogénomiques in vivo et in vitro démontrant que les hépatocytes humains sont un modèle fiable pour détecter les produits toxiques hépatocarcinogènes. Au total, nos études ont permis de fournir un cadre de travail original pour l’exploitation de différents types de données transcriptomiques pour comprendre l’impact des produits chimiques sur la biologie cellulaire. / The emergence of Big Data in molecular biology, especially through the study of transcriptomics, has revolutionized the way we look at various disciplines, such as drug development and cancer research. Big data analysis is an important part of the concept of precision medicine, which primary purpose is to understand the molecular mechanisms leading to better therapeutic response in patients. This thesis is halfway between pharmaco-toxicogenomics experimental studies, and clinical and translational studies. The aim of this thesis is mainly to show the potential and limitations of these studies and their relevance, especially for the discovery of drug response biomarkers and understanding the drug mechanisms (targets, toxicities). This thesis is an original work since it proposes a global approach to analyzing the largest pharmaco-toxicogenomic datasets available to date. The key aims were: 1) Addressing the challenge of reproducibility for biomarker discovery in cancer pharmacogenomics, by comparing two large pharmacogenomics studies published in 2012; 2) Understanding drugs mechanism of action using an integrative approach to generate a superior drug-taxonomy; and 3) Evaluating the conservation of toxicogenomic responses in primary hepatocytes vs. in vivo liver samples in order to check the feasability of cell models in toxicology studies. My main contributions can be summarized as follow: - I developed a bioinformatics pipeline to study the factors that trigger (in)consistency between two major pharmacogenomic studies. I suggested that the observed differences emerged from the non-standardization of pharmacological measurements, which could limit the validation of drug response biomarker. - I implemented a bioinformatics pipeline that demonstrated the superiority of the integrative approach, since it takes into account different parameters for the drug (structure, cytotoxicity, transcriptional perturbation) to elucidate the mechanism of action (MoA). - I developed a bioinformatics pipeline to study the level of conservation of toxicity mechanisms between the in vivo and in vitro system, showing that human hepatocytes is a reliable model for hepatocarcinogens testing. Overall, our studies have provided a unique framework to leverage various types of transcriptomic data in order to understand the impact of chemicals on cell biology.