Les biotechnologies sont arrivées au point ou la quantité d'information disponible permet de penser les objets biologiques comme des systèmes complexes. Dans ce contexte, les phénomènes qui émergent de ces systèmes sont intimement liés aux spécificités de leur organisation. Cela pose des problèmes computationnels et statistiques qui sont précisément l'objet d'étude de la communauté liée à l'apprentissage statistique. Cette thèse traite d'applications de méthodes d'apprentissage pour l'étude de phénomène biologique dans une perspective de système complexe. Ces méthodes sont appliquées dans le cadre de l'analyse d'interactions protéine-ligand et d'effets secondaires, du phenotypage de populations de cellules et du plan d'expérience pour des systèmes dynamiques non linéaires partiellement observés.D'importantes quantités de données sont désormais disponibles concernant les molécules mises sur le marché, tels que les profils d'interactions protéiques et d'effets secondaires. Cela pose le problème d'intégrer ces données et de trouver une forme de structure sous tendant ces observations à grandes échelles. Nous appliquons des méthodes récentes d'apprentissage non supervisé à l'analyse d'importants jeux de données sur des médicaments. Des exemples illustrent la pertinence de l'information extraite qui est ensuite validée dans un contexte de prédiction.Les variations de réponses à un traitement entre différents individus posent le problème de définir l'effet d'un stimulus à l'échelle d'une population d'individus. Par exemple, dans le contexte de la microscopie à haut débit, une population de cellules est exposée à différents stimuli. Les variations d'une cellule à l'autre rendent la comparaison de différents traitement non triviale. Un modèle génératif est proposé pour attaquer ce problème et ses propriétés sont étudiées sur la base de données expérimentales.A l'échelle moléculaire, des comportements complexes émergent de cascades d'interactions non linéaires entre différentes espèces moléculaires. Ces non linéarités engendrent des problèmes d'identifiabilité du système. Elles peuvent cependant être contournées par des plans expérimentaux spécifiques, un des champs de recherche de la biologie des systèmes. Une stratégie Bayésienne itérative de plan expérimental est proposée est des résultats numériques basés sur des simulations in silico d'un réseau biologique sont présentées. / Biotechnologies came to an era where the amount of information one has access to allows to think about biological objects as complex systems. In this context, the phenomena emerging from those systems are tightly linked to their organizational properties. This raises computational and statistical challenges which are precisely the focus of study of the machine learning community. This thesis is about applications of machine learning methods to study biological phenomena from a complex systems viewpoint. We apply machine learning methods in the context of protein-ligand interaction and side effect analysis, cell population phenotyping and experimental design for partially observed non linear dynamical systems.Large amount of data is available about marketed molecules, such as protein target interaction profiles and side effect profiles. This raises the issue of making sense of this data and finding structure and patterns that underlie these observations at a large scale. We apply recent unsupervised learning methods to the analysis of large datasets of marketed drugs. Examples show the relevance of extracted information which is further validated in a prediction context.The variability of the response to a treatment between different individuals poses the challenge of defining the effect of this stimulus at the level of a population of individuals. For example in the context High Content Screening, a population of cells is exposed to different stimuli. Between cell variability within a population renders the comparison of different treatments difficult. A generative model is proposed to overcome this issue and properties of the model are investigated based on experimental data.At the molecular scale, complex behaviour emerge from cascades of non linear interaction between molecular species. These non linearities leads to system identifiability issues. These can be overcome by specific experimental plan, one of the field of research in systems biology. A Bayesian iterative experimental design strategy is proposed and numerical results based on in silico biological network simulations are presented.

Un crible biologique a pour objectif de tester en parallèle l'impact de nombreuses conditions expérimentales sur un processus biologique d'un organisme modèle. Le progrès technique et computationnel a rendu possible la réalisation de tels cribles à grande échelle - jusqu'à des centaines de milliers de conditions. L'imagerie sur cellules vivantes est un excellent outil pour étudier en détail les conséquences d'une perturbation chimique sur un processus biologique. L'analyse des cribles sur cellules vivantes demande toutefois la combinaison de méthodes robustes d'imagerie par ordinateur et de contrôle qualité, et d'approches statistiques efficaces pour la détection des effets significatifs. La présente thèse répond à ces défis par le développement de méthodes analytiques pour les images de cribles temporels à haut débit. Les cadres qui y sont développés sont appliqués à des données publiées, démontrant par là leur applicabilité ainsi que les bénéfices d'une ré-analyse des données de cribles à haut contenu (HCS). Le premier workflow pour l'étude de la motilité cellulaire à l'échelle d'une cellule dans de telles données constitue le chapitre 2. Le chapitre 3 applique ce workflow à des données publiées et présente une nouvelle distance pour l'inférence de cible thérapeutique à partir d'images de cribles temporels. Enfin, le chapitre 4 présente une pipeline méthodologique complète pour la conduite de cribles temporels à haut débit en toxicologie environnementale. / Biological screens test large sets of experimental conditions with respect to their specific biological effect on living systems. Technical and computational progresses have made it possible to perform such screens at a large scale - up to hundreds of thousands of experiments. Live cell imaging is an excellent tool to study in detail the consequences of chemical perturbation on a given biological process. However, the analysis of live cell screens demands the combination of robust computer vision methods, efficient statistical methods for the detection of significant effects and robust procedures for quality control. This thesis addresses these challenges by developing analytical methods for the analysis of High Throughput time-lapse microscopy screening data. The developed frameworks are applied to publicly available HCS data, demonstrating their applicability and the benefits of HCS data remining. The first multivariate workflow for the study of single cell motility in such large-scale data is detailed in Chapter 2. Chapter 3 presents this workflow application to previously published data, and the development of a new distance for drug target inference by in silico comparisons of parallel siRNA and drug screens. Finally, chapter 4 presents a complete methodological pipeline for performing HT time-lapse screens in Environmental Toxicology.

The regulation of muscle stem cell (MuSC) asymmetric division plays an essential role in controlling the growth and repair of skeletal muscle. Perturbations in MuSC function have been demonstrated in disease and aging contexts such as Duchenne’s Muscular Dystrophy (DMD) and sarcopenia. We developed and optimized a high content analysis platform combining lineage tracing, myofiber culture, imaging, and bioinformatic analysis to determine modulators of muscle stem cell division. We discover kinase domain receptor (KDR) as a positive modulator of MuSC asymmetric division and confirmed its expression in satellite cells by ddPCR and immunofluorescence. Knockdown of KDR significantly reduces the numbers of asymmetric divisions, whereas ligand stimulation of KDR increases the numbers of asymmetric divisions. KDR signaling is impaired in dystrophin- deficient satellite cells and requires a polarized cell environment established by the dystrophin glycoprotein complex (DGC) to direct asymmetric division. Mice lacking KDR in MuSCs exhibit reduced numbers of satellite cells due to precocious differentiation, and deficits in regeneration consistent with impaired asymmetric division and reduced generation of progenitors. Therefore, our experiments identify KDR signaling as playing an essential role in MuSC function in muscle regeneration. These findings further our understanding of muscle stem cell biology, and in particular, the role of asymmetric division under homeostatic and regenerative conditions.

Introduction : Le virus respiratoire syncytial (RSV) est associé à des taux élevés de morbidité et de mortalité non seulement chez les jeunes enfants, en particulier les nourrissons et ceux atteints de cardiomyopathie congénitale, mais aussi chez les personnes de tous âges immunodéprimées et chez les personnes âgées. Les options thérapeutiques actuelles se limitent à une prophylaxie par anticorps monoclonaux réservée aux nourrissons à haut risque de maladie grave associée au RSV. Le développement de nouveaux antiviraux est donc urgent. Les antiviraux ciblant les protéines de l'hôte constituent une alternative émergente aux antiviraux classiques ciblant les protéines virales qui présentent des risques de développement de résistances. L'autophagie est un mécanisme cellulaire qui peut favoriser ou limiter la réplication virale. Nos travaux en cours suggèrent que l'autophagie dans les cellules épithéliales des voies respiratoires humaines (AECs) offre une protection antivirale contre RSV. Objectif : L'objectif de cette étude est d'évaluer la capacité de divers molécules induisant l'autophagie (AID) approuvés par la FDA à inhiber la réplication du RSV dans les AECs. Méthodes : Afin de quantifier l'induction de l'autophagie, la réplication du RSV et la viabilité cellulaire à l'aide d'un système d'imagerie à haut-débit, nous avons développé un essai utilisant des cellules A549, une lignée cellulaire modèle de cellules épithéliales respiratoires, la protéine LC3-RFP comme marqueur d’autophagie, un virus RSV recombinant exprimant la protéine GFP, et un marquage avec le SYTOX-Orange et le DAPI pour évaluer la viabilité cellulaire. Résultats et discussion : En utilisant la Torin-1, un AID caractérisé qui agit de manière mTOR -dépendante, nous avons confirmé que notre essai permet de mesurer l’induction de l’autophagie. De plus, nous avons constaté que la Torin-1 diminue significativement la réplication du RSV-GFP de manière dose-dépendante. Conclusion : En résumé, notre étude a permis de mettre en place un système expérimental à haut débit pour la caractérisation de l’effet des AIDs sur l’autophagie et leur impact sur la réplication du RSV. Nos résultats permettent de montrer que l’induction de l’autophagie corrèle avec la diminution de la réplication de RSV. Ces données devront être complétées par l’utilisation d’autres AIDs pour identifier des molécules approuvées par la FDA qui présentent une activité anti-RSV in vitro. / Introduction: Respiratory syncytial virus (RSV) is associated with high rates of morbidity and mortality not only in young children, particularly infants and those with congenital cardiomyopathy, but also in immunocompromised people of all ages and in the elderly. Current treatment options are limited to monoclonal antibody prophylaxis reserved for infants at high risk of serious illness associated with RSV. The development of new antivirals is therefore urgent. Antivirals targeting host proteins are an emerging alternative to conventional antivirals targeting viral proteins that pose risks of resistance development. Autophagy is a cellular mechanism that can promote or limit viral replication. Our ongoing work suggests that autophagy in human airway epithelial cells (AECs) provides antiviral protection against RSV. Objective: The objective of this study is to evaluate the ability of various FDA-approved autophagy-inducing molecules (AIDs) to inhibit RSV replication in AECs. Methods: In order to quantify autophagy induction, RSV replication and cell viability using a high-throughput imaging system, we developed an assay using A549 cells, a cell line model of respiratory epithelial cells, the LC3-RFP protein as an autophagy marker, a recombinant RSV virus expressing the GFP protein, and labeling with SYTOX-Orange and DAPI to assess cell viability. Results and discussion: Using Torin-1, a characterized AID that acts in an mTOR-dependent manner, we confirmed that our assay can measure the induction of autophagy. Furthermore, we found that Torin-1 significantly decreases RSV-GFP replication in a dose-dependent manner. Conclusion: In summary, our study allowed to set up a high-throughput experimental system for the characterization of the effect of AIDs on autophagy and their impact on RSV replication. Our results show that the induction of autophagy correlates with the decrease in RSV replication. These data should be supplemented by the use of other AIDs to identify FDA-approved molecules that exhibit anti-RSV activity in vitro.

<p>High-content screening (HCS) produces an immense amount of data, often on the scale of Terabytes. This requires considerable processing power resulting in long analysis time. As a result, HCS with a single-core processor system is an inefficient option because it takes a huge amount of time, storage and processing power. The situation is even worse because most of the image processing software is developed in high-level languages which make customization, flexibility and multi-processing features very challenging. Therefore, the goal of the project is to develop a multithreading model in C language. This model will be used to extract subcellular localization features, such as threshold adjacency statistics (TAS) from the HCS data. The first step of the research was to identify an appropriate dye for use in staining the MCF-7 cell line. The cell line has been treated with staurosporin kinase inhibitor, which can provide important physiological and morphological imaging information. The process of identifying a suitable dye involves treating cells with different dye options, capturing the fluorescent images of the treated cells with the Opera microscope, and analyzing the imaging properties of the stained cells. Several dyes were tested, and the most suitable dye to stain the cellular membrane was determined to be Di4-Anepps. The second part of the thesis was to design and develop a parallel program in C that can extract TAS features from the stained cellular images. The program reads the input cell images captured by Opera microscopes, converts it to TIFF format from the proprietary Opera format, identifies the region-of-interest contours of each cell, and computes the TAS features. A significant increase in speed in the order of four fold was obtained using the customized program. Different scalability tests using the developed software were compared against software developed in Acapella scripting language. The result of the test shows that the computational time is proportional to number of cells in the image and is inversely proportional to number of cores in a processor.</p> / Master of Applied Science (MASc)

In 2001 the human proteome organization (HUPO) was created with the ambition to identify and characterize all proteins encoded in the human genome according to several criteria; their expression levels in different tissues and under different conditions; the sub-cellular localization; post-translational modifications; interactions, and if possible also the relationship between their structure and function.When the knowledge of different proteins and their potential interactions increases, so does the need for methods able to unravel the nature of molecular processes in cells and organized tissues, and ultimately for clinical use in samples obtained from patients. The in situ proximity ligation assay (in situ PLA) was developed to provide localized detection of proteins, post-translational modifications and protein-protein interactions in fixed cells and tissues. Dual recognition of the target or interacting targets is a prerequisite for the creation of a circular reporter DNA molecule, which subsequently is locally amplified for visualization of individual protein molecules in single cells. These features offer the high sensitivity and selectivity required for detection of even rare target molecules. Herein in situ PLA was first established and then employed as a tool for detection of both interactions and post-translational modifications in cultured cells and tissue samples. In situ PLA was also adapted to high content screening (HCS) for therapeutic effects, where it was applied for cell-based drug screening of inhibitors influencing post-translational modifications. This was performed using primary cells, paving the way for evaluation of drug effects on cells from patient as a diagnostic tool in personalized medicine. In conclusion, this thesis describes the development and applications of in situ PLA as a tool to study proteins, post-translational modifications and protein-protein interactions in genetically unmodified cells and tissues, and for clinical interactomics.

Target-based screening projects for bioactive (orphan) compounds have been shown in many cases to be insufficiently predictive for in vivo efficacy, leading to attrition in clinical trials. Phenotypic screening has hence undergone a renaissance in both academia and in the pharmaceutical industry, partly due to this reason. One key shortcoming of this paradigm shift is that the protein targets modulated need to be elucidated subsequently, which is often a costly and time-consuming procedure. In this work, we have explored both improved methods and real-world case studies of how computational methods can help in target elucidation of phenotypic screens. One limitation of previous methods has been the ability to assess the applicability domain of the models, that is, when the assumptions made by a model are fulfilled and which input chemicals are reliably appropriate for the models. Hence, a major focus of this work was to explore methods for calibration of machine learning algorithms using Platt Scaling, Isotonic Regression Scaling and Venn-Abers Predictors, since the probabilities from well calibrated classifiers can be interpreted at a confidence level and predictions specified at an acceptable error rate. Additionally, many current protocols only offer probabilities for affinity, thus another key area for development was to expand the target prediction models with functional prediction (activation or inhibition). This extra level of annotation is important since the activation or inhibition of a target may positively or negatively impact the phenotypic response in a biological system. Furthermore, many existing methods do not utilize the wealth of bioactivity information held for orthologue species. We therefore also focused on an in-depth analysis of orthologue bioactivity data and its relevance and applicability towards expanding compound and target bioactivity space for predictive studies. The realized protocol was trained with 13,918,879 compound-target pairs and comprises 1,651 targets, which has been made available for public use at GitHub. Consequently, the methodology was applied to aid with the target deconvolution of AstraZeneca phenotypic readouts, in particular for the rationalization of cytotoxicity and cytostaticity in the High-Throughput Screening (HTS) collection. Results from this work highlighted which targets are frequently linked to the cytotoxicity and cytostaticity of chemical structures, and provided insight into which compounds to select or remove from the collection for future screening projects. Overall, this project has furthered the field of in silico target deconvolution, by improving the performance and applicability of current protocols and by rationalizing cytotoxicity, which has been shown to influence attrition in clinical trials.