Carbohydrates are ubiquitous components in nature involved in a range of tasks. They cover every cell and contribute both structural stability as well as identity. Lipopolysaccharides are the outermost exposed part of the bacterial cell wall and the primary target for host-pathogen recognition. Understanding the structure and biosynthesis of these polysaccharides is crucial to combat disease and develop new medicine. Structural determinations can be carried out using NMR spectroscopy, a powerful tool giving information on an atomistic scale. This thesis is focused on method development to study polysaccharide structures as well as application on bacterial lipopolysaccharides. The focus has been to incorporate a bioinformatics approach prior to analysis by NMR spectroscopy, and then computer assisted methods to aid in the subsequent analysis of the spectra. The third chapter deals with the recent developments of ECODAB, a tool that can help predict structural fragments in Escherichia coli O-antigens. It was migrated to a relational database and the aforementioned predictions can now be made automatically by ECODAB. The fourth chapter gives insight into the program CASPER, a computer program that helps with structure determination of oligo- and polysaccharides. An approach to determine substituent positions in polysaccharides was investigated. The underlying database was also expanded and the improved capabilities were demonstrated by determining O-antigenic structures that could not previously be solved. The fifth chapter is an application to O‑antigen structures of E. coli strains. This is done by a combination of NMR spectroscopy and bioinformatics to predict components as well as linkages prior to spectra analysis. In the first case, a full structure elucidation was performed on E. coli serogroup O63, and in the second case a demonstration of the bioinformatics approach is done to E. coli serogroup O93. In the sixth chapter, a new version of the CarbBuilder software is presented. This includes a more robust building algorithm that helps build sterically crowded polysaccharide structures, as well as a general expansion of possible components. / <p>At the time of the doctoral defense, the following papers were unpublished and had a status as follows: Paper 3: Manuscript. Paper 5: Manuscript.</p>

Ce mémoire traite du développement d'outils informatiques d'aide à l'analyse structurale des petites molécules organiques par Résonance Magnétique Nucléaire (RMN). Il comprend deux volets axés sur l'automatisation de tâches auxquelles les chimistes des laboratoires de synthèse organique ou d'isolement de substances naturelles sont confrontés au quotidien, à savoir l'élucidation et la vérification de structure. Le premier volet concerne des améliorations apportées au logiciel de génération de structure LSD (Logic for Structure Determination). Ce logiciel est basé sur l'interprétation des corrélations des spectres de RMN 2D pour la détermination de structures complètement ou partiellement inconnues. Les progrès récents ont pour but d'augmenter la diversité des molécules analysables et d'améliorer le traitement des corrélations ainsi que la présentation des résultats. L'intégration de LSD avec la base de données SISTEMAT permet de bénéficier de la source d'information supplémentaire que constitue la valeur des déplacements chimiques. Cet avantage se traduit par un filtrage des solutions en fonction d'éléments de sous-structure sélectionnés dans une collection de squelettes de produits naturels. Le second volet présente le développement du logiciel CASA (Computer-Aided Spectral Assignment) dont le rôle est de réaliser une vérification automatique de structure par l'attribution des résonances. Il s'appuie sur des contraintes issues des spectres de RMN 2D et sur un module de prédiction des déplacements chimiques 13C. / This thesis deals with the development of computational tools for structural analysis of small organic molecules by Nuclear Magnetic Resonance (NMR). It consists of two parts that focus on the automation of tasks that chemists working in the fields of organic synthesis or natural substance isolation daily face, namely structure elucidation and verification. The first part reports improvements of the structure generation software LSD (Logic for Structure Determination). This software is designed for the determination of completely or partially unknown structures from the interpretation of 2D NMR correlation spectra. The present work aims to increase the diversity of the molecules that can be analyzed and to improve the processing of correlation data as well as the presentation of results. The integration of LSD with the SISTEMAT database introduces chemical shifts values as an additional information source. It allows the chemist to sort the possible solutions of a problem according to the presence of known natural product skeletons. The second part presents the development of the CASA (Computer-Aided Spectral Assignment) software as a tool for automatic structure assignment. It jointly makes use of constraints from 2D NMR spectra and from the matching between experimental and predicted 13C chemical shifts.