An increasing number of proteins with unknown function have their three-dimensional structure solved at high resolution. This situation, largely due to structural genomics initiatives, has been stimulating the development of automated structure-based function prediction methods. Knowledge of residues important for function – and more particularly – for binding can help automated prediction of function in different ways. The properties of a binding site such as its shape or amino acid composition can provide clues on the ligand that may bind to it. Also, having information on functionally important regions in similar proteins can refine the process of annotation transfer between homologues.<p>Experimental results indicate that functional residues often have an unfavourable contribution to the stability of the folded state of a protein. This observation is the underlying principle of several computational methods for predicting the location of functional sites in protein structures. These methods search protein structures for destabilising residues, with the assumption that these are likely to be important for function.<p>We have developed a method to detect clusters of destabilising residues which are in close spatial proximity within a protein structure. Individual residue contributions to protein stability are evaluated using detailed atomic models and an energy function based on fundamental physico-chemical principles.<p>Our overall aim in this work was to evaluate the overlap between these clusters of destabilising residues and known binding sites in proteins.<p>Unfortunately, reliable benchmark datasets of known binding sites in proteins are sorely lacking. Therefore, we have undertaken a comprehensive approach to define binding sites unambiguously from structural data. We have rigorously identified seven issues which should be considered when constructing datasets of binding sites to validate prediction methods, and we present the construction of two new datasets in which these problems are handled. In this regard, our work constitute a major improvement over previous studies in the field.<p>Our first dataset consists of 70 proteins with binding sites for diverse types of ligands (e.g. nucleic acids, metal ions) and was constructed using all available data, including literature curation. The second dataset contains 192 proteins with binding sites for small ligands and polysaccharides, does not require literature curation, and can therefore be automatically updated.<p>We have used our dataset of 70 proteins to evaluate the overlap between destabilising regions and binding sites (the second dataset of 192 proteins was not used for that evaluation as it constitutes a later improvement). The overlap is on average limited but significantly larger than random. The extent of the overlap varies with the type of bound ligand. Significant overlap is obtained for most polysaccharide- and small ligand-binding sites, whereas no overlap is observed for nucleic acid-binding sites. These differences are rationalised in terms of the geometry and energetics of the binding sites.<p>Although destabilising regions, as detected in this work, can in general not be used to predict all types of binding sites in protein structures, they can provide useful information, particularly on the location of binding sites for polysaccharides and small ligands.<p>In addition, our datasets of binding sites in proteins should help other researchers to derive and validate new function prediction methods. We also hope that the criteria which we use to define binding sites may be useful in setting future standards in other analyses. / Doctorat en Sciences / info:eu-repo/semantics/nonPublished

Dans les cellules eucaryotes, l’expression d’un gène peut être régulée à de nombreux niveaux. Les études réalisées sur le contrôle de l’expression génique se sont généralement intéressées aux mécanismes de contrôle transcriptionnel. Cependant de nombreux exemples mettent de plus en plus en évidence l’importance des mécanismes post-transcriptionnels dans cette régulation. Les contrôles post-transcriptionnels de l’expression génique reposent essentiellement sur des interactions spécifiques entre les régions 5’ et 3’ non traduites de l’ARNm et des protéines agissant en trans qui contrôlent spécifiquement la maturation des ARNs messagers (ARNms), leur localisation cytoplasmique, leur stabilité et/ou leur traduction. Les éléments riches en adénine et en uridine (ARE), localisés dans la région 3’ non traduite de nombreux ARNms, font partie des séquences régulatrices les plus étudiées. Elles sont notamment présentes dans les ARNms codant pour des cytokines et des proto-oncogènes. Les protéines de liaison à l’ARN jouent donc un rôle central dans la régulation de l’expression des gènes. Les protéines TIA-1 et TIAR appartiennent à la famille des protéines qui fixent l’ARN et qui contiennent des domaines RRM (RNA Recognition Motif). Elles sont impliquées dans des mécanismes permettant la régulation de l’expression génique tels que l’épissage alternatif et la traduction. En particulier, elles participent à l’arrêt général de la traduction qui accompagne un stress environnemental en séquestrant les ARNms poly(A)+ non traduits dans des foci cytoplasmiques appelés granules de stress (SGs). Elles sont également impliquées dans la répression traductionnelle d’ARNms spécifiques en liant les ARE présents dans les extrémités 3’ non traduites de certains ARNms, et notamment des ARNs messagers codant pour le TNF-α et la cyclooxygénase-2 (Cox-2). L’invalidation des gènes tia-1 et tiar chez la souris conduit à une létalité embryonnaire élevée suggérant que ces protéines jouent également un rôle important au cours de l’embryogenèse. Afin de comprendre les mécanismes par lesquels les protéines TIA-1 et TIAR remplissent leurs différentes fonctions, nous avons réalisé un criblage par la technique du double hybride en levure afin d'identifier des partenaires d’interaction de ces deux protéines. Les protéines TIA-1 et TIAR interagissent avec les protéines FBPs (Fuse Binding Proteins). Celles-ci participent notamment à la maturation et à la dégradation des ARNs. Nous avons montré que les protéines FBPs co-localisent parfaitement avec TIA-1 dans le noyau et migrent dans les granules de stress en réponse à un stress oxydatif. De plus, des expériences de retard de migration sur gel réalisées à partir d’extrait cytosolique de macrophages ont montré que les protéines FBPs sont présentes dans le même complexe liant l’ARE du TNF-α que TIA-1. Enfin, la surexpression du domaine de liaison à l’ARN KH3 de FBP2 en fusion à l’EGFP induit la séquestration spécifique des protéines TIA-1 et TIAR dans des foci cytoplasmiques, empêchant ainsi leur accumulation nucléaire. Nos résultats indiquent que les protéines TIA-1/R et FBPs pourraient être fonctionnellement impliquées dans des étapes communes du métabolisme de l’ARN dans le noyau et/ou le cytoplasme. / Doctorat en sciences, Spécialisation biologie moléculaire / info:eu-repo/semantics/nonPublished

Cette thèse de doctorat s’inscrit dans le cadre de l’étude in silico des relations qui lient la séquence d’une protéine à sa structure, sa stabilité et sa fonction. Elle a pour objectif de permettre à terme la conception rationnelle de protéines modifiées qui restent actives dans des conditions physico chimiques non physiologiques. Nous nous sommes plus particulièrement penchés sur la stabilité thermique des protéines, qui est définie par leur température de fusion Tm au delà de laquelle leur structure n’est thermodynamiquement plus stable. Notre travail s’articule en trois grandes parties :la recherche de facteurs favorisant la thermostabilité des protéines parmi des familles de protéines homologues, la mise sur pied d’une base de données de protéines de structure et de Tm déterminées expérimentalement, de laquelle sont dérivés des potentiels statistiques dépendant de la température, et enfin la mise au point de deux outils bioinformatiques visant à prédire d’une part la Tm d’une protéine à partir de la Tm de protéines homologues et d’autre part les changements de thermostabilité d’une protéine (Tm) engendrés par l’introduction d’une mutation ponctuelle.<p><p>La première partie a pour objectif l’identification des facteurs de séquence et de structure (e.g. fréquence de ponts salins, d’interactions cation-{pi}) responsables des différentes stabilités thermiques de protéines homologues au sein de huit familles (chapitre 2). La spécificité de chaque famille ne nous a pas permis de généraliser l’impact de ces différents facteurs sur la stabilité thermique des protéines. Cependant, cette approche nous a permis de constater la multitude de stratégies différentes suivies par les protéines pour atteindre une plus grande thermostabilité.<p><p>La deuxième partie concerne le développement d’une approche originale pour évaluer l’influence de la température sur la contribution de différents types d’interactions à l’énergie libre de repliement des protéines (chapitres 3 et 4). Cette approche repose sur la dérivation de potentiels statistiques à partir d’ensembles de protéines de thermostabilité moyenne distincte. Nous avons d’une part collecté le plus grand nombre possible de protéines de structure et de Tm déterminées expérimentalement, et d’autre part développé des potentiels tenant compte de l’adaptation des protéines aux températures extrêmes au cours de leur évolution. Cette méthode originale a mis en évidence la dépendance en la température d’interactions protéiques tels les ponts salins, les interactions cation-{pi}, certains empilements hydrophobes .Elle nous a en outre permis de mettre le doigt sur l’importance de considérer la dépendance en la température non seulement des interactions attractives mais également des interactions répulsives, ainsi que sur l’importance de décrire la résistance thermique par la Tm plutôt que la Tenv, température de l’environnement de l’organisme dont elle provient (chapitre 5).<p><p>La dernière partie de cette thèse concerne l’utilisation des profils énergétiques dans un but prédictif. Tout d’abord, nous avons développé un logiciel bioinformatique pour prédire la thermostabilité d’une protéine sur la base de la thermostabilité de protéines homologues. Cet outil s’est avéré prometteur après l’avoir testé sur huit familles de protéines homologues. Nous avons également développé un deuxième outil bioinformatique pour prédire les changements de thermostabilité d’une protéine engendrés par l’introduction d’une mutation ponctuelle, en s’inspirant d’un logiciel de prédiction des changements de stabilité thermodynamique des protéines développé au sein de notre équipe de recherche. Ce deuxième algorithme de prédiction repose sur le développement d’une grande base de données de mutants caractérisés expérimentalement, d’une combinaison linéaire de potentiels pour évaluer la Tm, et d’un réseau de neurones pour identifier les coefficients de la combinaison. Les prédictions générées par notre logiciel ont été comparées à celles obtenues via la corrélation qui existe entre stabilités thermique et thermodynamique, et se sont avérées plus fiables.<p><p>Les travaux décrits dans notre thèse, et en particulier le développement de potentiels statistiques dépendant de la température, constituent une nouvelle approche très prometteuse pour comprendre et prédire la thermostabilité des protéines. En outre, nos travaux de recherche ont permis de développer une méthodologie qui pourra être adaptée à l’étude et à la prédiction d’autres propriétés physico chimiques des protéines comme leur solubilité, leur stabilité vis à vis de l’acidité, de la pression, de la salinité .lorsque suffisamment de données expérimentales seront disponibles.<p> / Doctorat en Sciences agronomiques et ingénierie biologique / info:eu-repo/semantics/nonPublished

Domain swapping is a wide spread phenomenon which involves the association between two or more protein subunits such that intra-molecular interactions between domains in each subunit are replaced by equivalent inter-molecular interactions between the same domains in different subunits. This thesis is devoted to the analysis of the factors that drive proteins to undergo such association modes. The specific system analyzed is the monomer to swapped dimer formation of the B1 domain of the immunoglobulin G binding protein (GB1). The formation of this dimer was shown to be fostered by 4 amino acid substitutions (L5V, F30V, Y33F, A34F) (Byeon et al. 2003). In this work, computational protein design and molecular dynamics simulations, both with detailed atomic models, were used to gain insight into how these 4 mutations may promote the domain swapping reaction.<p>The stability of the wt and quadruple mutant GB1 monomers was assessed using the software DESIGNER, a fully automatic procedure that selects amino acid sequences likely to stabilize a given backbone structure (Wernisch et al. 2000). Results suggest that 3 of the mutations (L5V, F30V, A34F) have a destabilizing effect. The first mutation (L5V) forms destabilizing interactions with surrounding residues, while the second (F30V) is engaged in unfavorable interactions with the protein backbone, consequently causing local strain. Although the A34F substitution itself is found to contribute favorably to the stability of the monomer, this is achieved only at the expense of forcing the wild type W43 into a highly strained conformation concomitant with the formation of unfavorable interactions with both W43 and V54.<p>Finally, we also provide evidence that A34F mutation stabilizes the swapped dimer structure. Although we were unable to perform detailed protein design calculations on the dimer, due to the lower accuracy of the model, inspection of its 3D structure reveals that the 34F side chains pack against one another in the core of the swapped structure, thereby forming extensive non-native interactions that have no counterparts in the individual monomers. Their replacement by the much smaller Ala residue is suggested to be significantly destabilizing by creating a large internal cavity, a phenomenon, well known to be destabilizing in other proteins. Our analysis hence proposes that the A34F mutation plays a dual role, that of destabilizing the GB1 monomer structure while stabilizing the swapped dimer conformation.<p>In addition to the above study, molecular dynamics simulations of the wild type and modeled quadruple mutant GB1 structures were carried out at room and elevated temperatures (450 K) in order to sample the conformational landscape of the protein near its native monomeric state, and to characterize the deformations that occur during early unfolding. This part of the study was aimed at investigating the influence of the amino acid sequence on the conformational properties of the GB1 monomer and the possible link between these properties and the swapping process. Analysis of the room temperature simulations indicates that the mutant GB1 monomer fluctuates more than its wild type counter part. In addition, we find that the C-terminal beta-hairpin is pushed away from the remainder of the structure, in agreement with the fact that this hairpin is the structural element that is exchanged upon domain swapping. The simulations at 450 K reveal that the mutant protein unfolds more readily than the wt, in agreement with its decreased stability. Also, among the regions that unfold early is the alpha-helix C-terminus, where 2 out of the 4 mutations reside. NMR experiments by our collaborators have shown this region to display increased flexibility in the monomeric state of the quadruple mutant.<p>Our atomic scale investigation has thus provided insights into how sequence modifications can foster domain swapping of GB1. Our findings indicate that the role of the amino acid substitutions is to decrease the stability of individual monomers while at the same time increase the stability of the swapped dimer, through the formation of non-native interactions. Both roles cooperate to foster swapping. / Doctorat en sciences, Spécialisation biologie moléculaire / info:eu-repo/semantics/nonPublished

Living cells and their components play a key role within biotechnology industry. Cell cultures and their products of interest are used for the design of vaccines as well as in the agro-alimentary field. In order to ensure optimal working of such bioprocesses, the understanding of the complex mechanisms which rule them is fundamental. Mathematical models may be helpful to grasp the biological phenomena which intervene in a bioprocess. Moreover, they allow prediction of system behaviour and are frequently used within engineering tools to ensure, for instance, product quality and reproducibility.<p> <p>Mathematical models of cell cultures may come in various shapes and be phrased with varying degrees of mathematical formalism. Typically, three main model classes are available to describe the nonlinear dynamic behaviour of such biological systems. They consist of macroscopic models which only describe the main phenomena appearing in a culture. Indeed, a high model complexity may lead to long numerical computation time incompatible with engineering tools like software sensors or controllers. The first model class is composed of the first principles or white box models. They consist of the system of mass balances for the main species (biomass, substrates, and products of interest) involved in a reaction scheme, i.e. a set of irreversible reactions which represent the main biological phenomena occurring in the considered culture. Whereas transport phenomena inside and outside the cell culture are often well known, the reaction scheme and associated kinetics are usually a priori unknown, and require special care for their modelling and identification. The second kind of commonly used models belongs to black box modelling. Black boxes consider the system to be modelled in terms of its input and output characteristics. They consist of mathematical function combinations which do not allow any physical interpretation. They are usually used when no a priori information about the system is available. Finally, hybrid or grey box modelling combines the principles of white and black box models. Typically, a hybrid model uses the available prior knowledge while the reaction scheme and/or the kinetics are replaced by a black box, an Artificial Neural Network for instance.<p><p>Among these numerous models, which one has to be used to obtain the best possible representation of a bioprocess? We attempt to answer this question in the first part of this work. On the basis of two simulated bioprocesses and a real experimental one, two model kinds are analysed. First principles models whose reaction scheme and kinetics can be determined thanks to systematic procedures are compared with hybrid model structures where neural networks are used to describe the kinetics or the whole reaction term (i.e. kinetics and reaction scheme). The most common artificial neural networks, the MultiLayer Perceptron and the Radial Basis Function network, are tested. In this work, pure black box modelling is however not considered. Indeed, numerous papers already compare different neural networks with hybrid models. The results of these previous studies converge to the same conclusion: hybrid models, which combine the available prior knowledge with the neural network nonlinear mapping capabilities, provide better results.<p><p>From this model comparison and the fact that a physical kinetic model structure may be viewed as a combination of basis functions such as a neural network, kinetic model structures allowing biological interpretation should be preferred. This is why the second part of this work is dedicated to the improvement of the general kinetic model structure used in the previous study. Indeed, in spite of its good performance (largely due to the associated systematic identification procedure), this kinetic model which represents activation and/or inhibition effects by every culture component suffers from some limitations: it does not explicitely address saturation by a culture component. The structure models this kind of behaviour by an inhibition which compensates a strong activation. Note that the generalization of this kinetic model is a challenging task as physical interpretation has to be improved while a systematic identification procedure has to be maintained.<p><p>The last part of this work is devoted to another kind of biological systems: proteins. Such macromolecules, which are essential parts of all living organisms and consist of combinations of only 20 different basis molecules called amino acids, are currently used in the industrial world. In order to allow their functioning in non-physiological conditions, industrials are open to modify protein amino acid sequence. However, substitutions of an amino acid by another involve thermodynamic stability changes which may lead to the loss of the biological protein functionality. Among several theoretical methods predicting stability changes caused by mutations, the PoPMuSiC (Prediction Of Proteins Mutations Stability Changes) program has been developed within the Genomic and Structural Bioinformatics Group of the Université Libre de Bruxelles. This software allows to predict, in silico, changes in thermodynamic stability of a given protein under all possible single-site mutations, either in the whole sequence or in a region specified by the user. However, PoPMuSiC suffers from limitations and should be improved thanks to recently developed techniques of protein stability evaluation like the statistical mean force potentials of Dehouck et al. (2006). Our work proposes to enhance the performances of PoPMuSiC by the combination of the new energy functions of Dehouck et al. (2006) and the well known artificial neural networks, MultiLayer Perceptron or Radial Basis Function network. This time, we attempt to obtain models physically interpretable thanks to an appropriate use of the neural networks.<p> / Doctorat en sciences appliquées / info:eu-repo/semantics/nonPublished

Molecular Biology has allowed the characterization and manipulation of the molecules of life in the wet lab. Also the structures of those macromolecules are being continuously elucidated. During the last decades of the past century, there was an increasing interest to study how the different genes are organized into different organisms (‘genomes’) and how those genes are expressed into proteins to achieve their functions. Currently the sequences for many genes over several genomes have been determined. In parallel, the efforts to have the structure of the proteins coded by those genes go on. However it is experimentally much harder to obtain the structure of a protein, rather than just its sequence. For this reason, the number of protein structures available in databases is an order of magnitude or so lower than protein sequences. Furthermore, in order to understand how living organisms work at molecular level we need the information about the interaction of those proteins. Elucidating the structure of protein macromolecular assemblies is still more difficult. To that end, the use of computers to predict the structure of these complexes has gained interest over the last decades.<p>The main subject of this thesis is the evaluation of current available computational methods to predict protein – protein interactions and build an atomic model of the complex. The core of the thesis is the evaluation protocol I have developed at Service de Conformation des Macromolécules Biologiques et de Bioinformatique, Université Libre de Bruxelles, and its computer implementation. This method has been massively used to evaluate the results on blind protein – protein interaction prediction in the context of the world-wide experiment CAPRI, which have been thoroughly reviewed in several publications [1-3]. In this experiment the structure of a protein complex (‘the target’) had to be modeled starting from the coordinates of the isolated molecules, prior to the release of the structure of the complex (this is commonly referred as ‘docking’).<p>The assessment protocol let us compute some parameters to rank docking models according to their quality, into 3 main categories: ‘Highly Accurate’, ‘Medium Accurate’, ‘Acceptable’ and ‘Incorrect’. The efficiency of our evaluation and ranking is clearly shown, even for borderline cases between categories. The correlation of the ranking parameters is analyzed further. In the same section where the evaluation protocol is presented, the ranking participants give to their predictions is also studied, since often, good solutions are not easily recognized among the pool of computer generated decoys.<p>An overview of the CAPRI results made per target structure and per participant regarding the computational method they used and the difficulty of the complex. Also in CAPRI there is a new ongoing experiment about scoring previously and anonymously generated models by other participants (the ‘Scoring’ experiment). Its promising results are also analyzed, in respect of the original CAPRI experiment. The Scoring experiment was a step towards the use of combine methods to predict the structure of protein – protein complexes. We discuss here its possible application to predict the structure of protein complexes, from a clustering study on the different results.<p>In the last chapter of the thesis, I present the preliminary results of an ongoing study on the conformational changes in protein structures upon complexation, as those rearrangements pose serious limitations to current computational methods predicting the structure protein complexes. Protein structures are classified according to the magnitude of its conformational re-arrangement and the involvement of interfaces and particular secondary structure elements is discussed. At the end of the chapter, some guidelines and future work is proposed to complete the survey. / Doctorat en Sciences / info:eu-repo/semantics/nonPublished

Cette thèse se place dans le cadre de l'étude in silico, c'est-à-dire assistée par ordinateur, des liens qui unissent la séquence d'une protéine à la (ou aux) structure(s) tri-dimensionnelle(s) qu'elle adopte. Le décryptage de ces liens présente de nombreuses applications dans divers domaines et constitue sans doute l'une des problématiques les plus fascinantes de la recherche en biologie moléculaire.<p><p>Le premier aspect de notre travail concerne le développement de potentiels statistiques dérivés de bases de données de protéines dont les structures sont connues. Ces potentiels présentent plusieurs avantages: ils peuvent être aisément adaptés à des représentations structurales simplifiées, et permettent de définir un nombre limité de fonctions énergétiques qui incarnent l'ensemble complexe d'interactions gouvernant la structure et la stabilité des protéines, et qui incluent également certaines contributions entropiques. Cependant, leur signification physique reste assez nébuleuse, car l'impact des diverses hypothèses nécessaires à leur dérivation est loin d'être clairement établi. Nous nous sommes attachés à l'étude de certaines limitations des ces potentiels: leur dépendance en la taille des protéines incluses dans la base de données, la non-additivité des termes de potentiels, et l'importance souvent négligée de l'environnement protéique spécifique ressenti par chaque résidu. Nous avons ainsi mis en évidence que l'influence de la taille des protéines de la base de données sur les potentiels de distance entre résidus est spécifique à chaque paire d'acides aminés, peut être relativement importante, et résulte essentiellement de la répartition inhomogène des résidus hydrophobes et hydrophiles entre le coeur et la surface des protéines. Ces résultats ont guidé la mise au point de fonctions correctives qui permettent de tenir compte de cette influence lors de la dérivation des potentiels. Par ailleurs, la définition d'une procédure générale de dérivation de potentiels et de termes de couplage a rendu possible la création d'une fonction énergétique qui tient compte simultanément de plusieurs descripteurs de séquence et de structure (la nature des résidus, leurs conformations, leurs accessibilités au solvant, ainsi que les distances qui les séparent dans l'espace et le long de la séquence). Cette fonction énergétique présente des performances nettement améliorées par rapport aux potentiels originaux, et par rapport à d'autres potentiels décrits dans la littérature.<p><p>Le deuxième aspect de notre travail concerne l'application de programmes basés sur des potentiels statistiques à l'étude de protéines qui adoptent des structures alternatives. La permutation de domaines est un phénomène qui affecte diverses protéines et qui implique la génération d'un oligomère suite à l'échange de fragments structuraux entre monomères identiques. Nos résultats suggèrent que la présence de "faiblesses structurales", c'est-à-dire de régions qui ne sont pas optimales vis-à-vis de la stabilité de la structure native ou qui présentent une préférence marquée pour une conformation non-native en absence d'interactions tertiaires, est intimement liée aux mécanismes de permutation. Nous avons également mis en évidence l'importance des interactions de type cation-{pi}, qui sont fréquemment observées dans certaines zones clés de la permutation. Finalement, nous avons sélectionné un ensemble de mutations susceptibles de modifier sensiblement la propension de diverses protéines à permuter. L'étude expérimentale de ces mutations devrait permettre de valider, ou de raffiner, les hypothèses que nous avons proposées quant au rôle joué par les faiblesses structurales et les interactions de type cation-{pi}. Nous avons également analysé une autre protéine soumise à d'importants réarrangements conformationnels: l'{alpha}1-antitrypsine. Dans le cas de cette protéine, les modifications structurales sont indispensables à l'exécution de l'activité biologique normale, mais peuvent sous certaines conditions mener à la formation de polymères insolubles et au développement de maladies. Afin de contribuer à une meilleure compréhension des mécanismes responsables de la polymérisation, nous avons cherché à concevoir rationnellement des protéines mutantes qui présentent une propension à polymériser contrôlée. Des tests expérimentaux ont été réalisés par le groupe australien du Professeur S.P. Bottomley, et ont permis de valider nos prédictions de manière assez remarquable.<p><p><p><p>The work presented in this thesis concerns the computational study of the relationships between the sequence of a protein and its three-dimensional structure(s). The unravelling of these relationships has many applications in different domains and is probably one of the most fascinating issues in molecular biology.<p><p>The first part of our work is devoted to the development of statistical potentials derived from databases of known protein structures. These potentials allow to define a limited number of energetic functions embodying the complex ensemble of interactions that rule protein folding and stability (including some entropic contributions), and can be easily adapted to simplified representations of protein structures. However, their physical meaning remains unclear since several hypotheses and approximations are necessary, whose impact is far from clearly understood. We studied some of the limitations of these potentials: their dependence on the size of the proteins included in the database, the non-additivity of the different potential terms, and the importance of the specific environment of each residue. Our results show that residue-based distance potentials are affected by the size of the database proteins, and that this effect can be quite strong, is residue-specific, and seems to result mostly from the inhomogeneous partition of hydrophobic and hydrophilic residues between the surface and the core of proteins. On the basis of these observations, we defined a set of corrective functions in order to take protein size into account while deriving the potentials. On the other hand, we developed a general procedure of derivation of potentials and coupling terms and consequently created an energetic function describing the correlations between several sequence and structure descriptors (the nature of each residue, the conformation of its main chain, its solvent accessibility, and the distances that separate it from other residues, in space and along the sequence). This energetic function presents a strongly improved predictive power, in comparison with the original potentials and with other potentials described in the literature.<p><p>The second part describes the application of different programs, based on statistical potentials, to the study of proteins that adopt alternative structures. Domain swapping involves the exchange of a structural element between identical proteins, and leads to the generation of an oligomeric unit. We showed that the presence of “structural weaknesses”, regions that are not optimal with respect to the folding mechanisms or to the stability of the native structure, seems to be intimately linked with the swapping mechanisms. In addition, cation-{pi} interactions were frequently detected in some key locations and might also play an important role. Finally, we designed a set of mutations that are likely to affect the swapping propensities of different proteins. The experimental study of these mutations should allow to validate, or refine, our hypotheses concerning the importance of structural weaknesses and cation-{pi} interactions. We also analysed another protein that undergoes large conformational changes: {alpha}1-antitrypsin. In this case, the structural modifications are necessary to the proper execution of the biological activity. However, under certain circumstances, they lead to the formation of insoluble polymers and the development of diseases. With the aim of reaching a better understanding of the mechanisms that are responsible for this polymerisation, we tried to design mutant proteins that display a controlled polymerisation propensity. An experimental study of these mutants was conducted by the group of Prof. S.P. Bottomley, and remarkably confirmed our predictions.<p> / Doctorat en sciences appliquées / info:eu-repo/semantics/nonPublished