In recent years it has become evident that functional RNAs in living organisms are not just curious remnants from a primoridal RNA world but an ubiquitous phenomenon complementing protein enzyme based activity. Functional RNAs, just like proteins, depend in many cases upon their well-defined and evolutionarily conserved three-dimensional structure. In contrast to protein folds, however, RNA molecules have a biophysically important coarse-grained representation: their secondary structure. At this level of resolution at least, RNA structures can be efficiently predicted given only the sequence information. As a consequence, computational studies of RNA routinely incorporate structural information explicitly. RNA secondary structure prediction has proven useful in diverse fields ranging from theoretical models of sequence evolution and biopolymer folding, to genome analysis and even the design biotechnologically or pharmaceutically useful molecules.

A molecule can crystallise in more than one crystal structure, a common phenomenon in organic compounds known as polymorphism. Different polymorphic forms may have significantly different physical properties, and a reliable prediction would be beneficial to the pharmaceutical industry. However, crystal structure prediction (CSP) based on the knowledge of the chemical structure had long been considered impossible. Previous failures of some CSP attempts led to speculation that the thermodynamic calculations in CSP methodologies failed to predict the kinetically favoured structures. Similarly, regarding the stabilities of co-crystals relative to their pure components, the results from lattice energy calculations and full CSP studies were inconclusive. In this thesis, these problems are addressed using the state-of-the-art CSP methodology implemented in the GRACE software. Firstly, it is shown that the low-energy predicted structures of four organic molecules, which have previously been considered difficult for CSP, correspond to their experimental structures. The possible outcomes of crystallisation can be reliably predicted by sufficiently accurate thermodynamic calculations. Then, the polymorphism of 5- chloroaspirin is investigated theoretically. The order of polymorph stability is predicted correctly and the isostructural relationships between a number of predicted structures and the experimental structures of other aspirin derivatives are established. Regarding the stabilities of co-crystals, 99 out of 102 co-crystals and salts of nicotinamide, isonicotinamide and picolinamide reported in the Cambridge Structural Database (CSD) are found to be more stable than their corresponding co-formers. Finally, full CSP studies of two co-crystal systems are conducted to explain why the co-crystals are not easily obtained experimentally. / University of Bradford

<p></p>

De récentes découvertes montrent le rôle important que joue l’acide ribonucléique (ARN) au sein des cellules, que ce soit le contrôle de l’expression génétique, la régulation de plusieurs processus homéostasiques, en plus de la transcription et la traduction de l’acide désoxyribonucléique (ADN) en protéine. Si l’on veut comprendre comment la cellule fonctionne, nous devons d’abords comprendre ses composantes et comment ils interagissent, et en particulier chez l’ARN. La fonction d’une molécule est tributaire de sa structure tridimensionnelle (3D). Or, déterminer expérimentalement la structure 3D d’un ARN s’avère fort coûteux. Les méthodes courantes de prédiction par ordinateur de la structure d’un ARN ne tiennent compte que des appariements classiques ou canoniques, similaires à ceux de la fameuse structure en double-hélice de l’ADN. Ici, nous avons amélioré la prédiction de structures d’ARN en tenant compte de tous les types possibles d’appariements, dont ceux dits non-canoniques. Cela est rendu possible dans le contexte d’un nouveau paradigme pour le repliement des ARN, basé sur les motifs cycliques de nucléotides ; des blocs de bases pour la construction des ARN. De plus, nous avons dévelopées de nouvelles métriques pour quantifier la précision des méthodes de prédiction des structures 3D des ARN, vue l’introduction récente de plusieurs de ces méthodes. Enfin, nous avons évalué le pouvoir prédictif des nouvelles techniques de sondage de basse résolution des structures d’ARN. / Recent findings show the important role of ribonucleic acid (RNA) within the cell, be it the control of gene expression, the regulation of several homeostatic processes, in addition to the transcription and translation of deoxyribonucleic acid (DNA) into protein. If we wish to understand how the cell works, we first need to understand its components and how they interact, and in particular for RNA. The function of a molecule is tributary of its three-dimensional (3D) structure. However, experimental determination of RNA 3D structures imparts great costs. Current methods for RNA structure prediction by computers only take into account the classical or canonical base pairs, similar to those found in the well-celebrated DNA double helix. Here, we improved RNA structure prediction by taking into account all possible types of base pairs, even those said non-canonicals. This is made possible in the context of a new paradigm for the folding of RNA, based on nucleotide cyclic motifs (NCM): basic blocks for the construction of RNA. Furthermore, we have developed new metrics to quantify the precision of RNA 3D structure prediction methods, given the recent introduction of many of those methods. Finally, we have evaluated the predictive power of the latest low-resolution RNA structure probing techniques.

La flexibilité est une caractéristique intrinsèque des protéines qui doivent, dès le mo- ment de leur synthèse, passer d’un état de chaîne linéaire à un état de structure tridimen- sionnelle repliée et enzymatiquement active. Certaines protéines restent flexibles une fois repliées et subissent des changements de conformation de grande amplitude lors de leur cycle enzymatique. D’autres contiennent des segments si flexibles que leur structure ne peut être résolue par des méthodes expérimentales. Dans cette thèse, nous présentons notre application de méthodes in silico d’analyse de la flexibilité des protéines : • À l’aide des méthodes de dynamique moléculaire dirigée et d’échantillonnage pa- rapluie, nous avons caractérisé les trajectoires de liaison de l’inhibiteur Z-pro- prolinal à la protéine Prolyl oligopeptidase et identifié la trajectoire la plus pro- bable. Nos simulations ont aussi identifié un mode probable de recrutement des ligands utilisant une boucle flexible de 19 acides aminés à l’interface des deux domaines de la protéine. • En utilisant les méthodes de dynamique moléculaire traditionnelle et dirigée, nous avons examiné la stabilité de la protéine SAV1866 dans sa forme fermée insérée dans une membrane lipidique et étudié un des modes d’ouverture possibles par la séparation de ses domaines liant le nucléotide. • Nous avons adapté auproblème de la prédiction de la structure des longues boucles flexibles la méthode d’activation et de relaxation ART-nouveau précédemment uti- lisée dans l’étude du repliement et de l’agrégation de protéines. Appliqué au replie- ment de boucles de 8 à 20 acides aminés, la méthode démontre une dépendance quadratique du temps d’exécution sur la longueur des boucles, rendant possible l’étude de boucles encore plus longues. / Flexibility is an intrinsic characteristic of proteins who from the moment of synthesis into a linear chain of amino acids, have to adopt an enzymatically active tridimensionnel structure. Some proteins stay flexible once folded and display large amplitude confor- mational changes during their enzymatic cycles. Others contain parts that are so flexible that their structure can’t be resolved using experimental methods. In this thesis, we present our application of in silico methods to the study of protein flexibility. • Using steered molecular dynamics and umbrella sampling, we characterized the binding trajectories of the Z-pro-prolinal inhibiter to the Prolyl oligopeptidase pro- tein and we identified the most probable trajectory. Our simulations also found a possible ligand recrutement mechanism that involves a 19 amino acids flexible loop at the interface of the two domains of the protein. • Using traditional and steered molecular dynamics, we examined the stability of the SAV1866 protein in its closed conformation in a lipid membrane and we studied one of its proposed opening modes by separating its nucleotide binding domains. • We also adapted the activation-relaxation technique ART-nouveau which was pre- viously used to study protein folding and aggregation to the problem of structure prediction of large flexible loops. When tested on loops of 8 to 20 amino acids, the method demonstrate a quadratic execution time dependance on the loop length, which makes it possible to use the method on even larger loops.

A Determinação da Estrutura tridimensional de Proteínas (DEP) a partir da sua seqüência de aminoácidos é importante para a engenharia de proteínas e o desenvolvimento de novos fármacos. Uma alternativa para este problema tem sido a aplicação de técnicas de computação evolutiva. As abordagens utilizando Algoritmos Evolutivos (AEs) tem obtido resultados relevantes, porém estão restritas a pequenas proteínas, com dezenas de aminoácidos e a algumas classes de proteínas. Este trabalho propõe a investigação de uma abordagem utilizando AEs para a predição da estrutura terciária de proteínas independentemente do seu tamanho e classe. Os resultados obtidos demonstram que apesar das dificuldades encontradas a abordagem investigada constitue-se em uma alternativa em relação aos métodos clássicos de determinação da estrutura terciária das proteínas. / Protein structure determination (DEP) from aminoacid sequences is very importante to protein engineering and development of new drugs. Evolutionary computation has been aplied to this problem with relevant results. Nevertheless, Evolutionary Algorithms (EAs) can work with only proteins with few aminoacids and some protein classes. This work proposes an approach using AEs to predict protein tertiary structure independly from their size and class. The obtained results show that, despite of the difficulties that have been found, the investigate approach is a relevant alternative to classical methods to protein structure determination.

A predição de estruturas de proteínas é um campo de pesquisa que busca simular o enovelamento de cadeias de aminoácidos de forma a descobrir as funções das proteínas na natureza, um processo altamente dispendioso por meio de métodos in vivo. Inserida no contexto da Bioinformática, é uma das tarefas mais computacionalmente custosas e desafiadoras da atualidade. Devido à complexidade, muitas pesquisas se utilizam de gateways científicos para disponibilização de ferramentas de execução e análise desses experimentos, aliado ao uso de workflows científicos para organização de tarefas e disponibilização de informações. No entanto, esses gateways podem enfrentar gargalos de desempenho e falhas estruturais, produzindo resultados de baixa qualidade. Para atuar nesse contexto multifacetado e oferecer alternativas para algumas das limitações, esta tese propõe uma arquitetura modular baseada nos conceitos de Service Oriented Architecture (SOA) para oferta de recursos computacionais em gateways científicos, com foco nos experimentos de Protein Structure Prediction (PSP). A Arquitetura Modular Distribuída para auxílio à Predição de Estruturas de Proteínas (MDAPSP) é descrita conceitualmente e validada em um modelo de simulação computacional, no qual se pode identificar suas capacidades, detalhar o funcionamento de seus módulos e destacar seu potencial. A avaliação experimental demonstra a qualidade dos algoritmos propostos, ampliando a capacidade de atendimento de um gateway científico, reduzindo o tempo necessário para experimentos de predição e lançando as bases para o protótipo de uma arquitetura funcional. Os módulos desenvolvidos alcançam boa capacidade de otimização de experimentos de PSP em ambientes distribuídos e constituem uma novidade no modelo de provisionamento de recursos para gateways científicos. / PSP is a scientific process that simulates the folding of amino acid chains to discover the function of a protein in live organisms, considering that its an expensive process to be done by in vivo methods. PSP is a computationally demanding and challenging effort in the Bioinformatics stateof- the-art. Many works use scientific gateways to provide tools for execution and analysis of such experiments, along with scientific workflows to organize tasks and to share information. However, these gateways can suffer performance bottlenecks and structural failures, producing low quality results. With the goal of offering alternatives to some of the limitations and considering the complexity of the topics involved, this thesis proposes a modular architecture based on SOA concepts to provide computing resources to scientific gateways, with focus on PSP experiments. The Modular Distributed Architecture to support Protein Structure Prediction (MDAPSP) is described conceptually and validated in a computer simulation model that explain its capabilities, detail the modules operation and highlight its potential. The performance evaluation presents the quality of the proposed algorithms, a reduction of response time in PSP experiments and prove the benefits of the novel algorithms, establishing the basis for a prototype. The new modules can optmize the PSP experiments in distributed environments and are a innovation in the resource provisioning model for scientific gateways.

As proteínas são moléculas presentes nos seres vivos e essenciais para a vida deles. Para entender a função de uma proteína, devese conhecer sua estrutura tridimensional (o posicionamento correto de todos os seus átomos no espaço). A partir da estrutura de uma proteína vital de um organismo causador de uma doença é possível desenvolver fármacos para o tratamento da doença. Para encontrar a estrutura de uma proteína, métodos biofísicos, como Cristalografia de Raio-X e Ressonância Nuclear Magnética têm sido empregados. No entanto, o uso desses métodos tem restrições práticas que impedem a determinação de várias estruturas de proteínas. Para contornar essas limitações, métodos computacionais para o problema de predição da estrutura da proteína (PSP, Protein Structure Prediction) têm sido investigados. Várias classes de métodos computacionais têm sido desenvolvidas para o problema de PSP. Entre elas, as abordagens ab initio são muito importantes, pois não utilizam nenhuma informação prévia de outras estruturas de proteínas para fazer o PSP, apenas a sequência de aminoácidos da proteína e o gráfico de Ramachandran são empregados. O PSP ab initio é um problema combinatorial que envolve relativamente grandes instâncias na prática, por exemplo, as proteínas em geral têm centenas ou milhares de variáveis para determinar. Para vencer esse entrave, metaheurísticas como os Algoritmos Genéticos (AGs) têm sido investigados. As soluções geradas por um AG são avaliadas pelo cálculo da energia potencial da proteína. Entre elas, o cálculo da interação da energia de van der Waals é custoso computacionalmente tornando o processo evolutivo do AG muito lento mesmo para proteínas pequenas. Este trabalho investiga técnicas para reduzir significativamente o tempo de execução desse cálculo. Basicamente, foram propostas modificações de técnicas de paralelização utilizando MPI e OpenMP para os algoritmos resultantes. Os resultados mostram que o cálculo pode ser 1.500 vezes mais rápido para proteínas gigantes quando aplicadas as técnicas investigadas neste trabalho / Proteins are molecules present in the living organism and essential for their life. To understand the function of a protein, its threedimensional structure (the correct positions of all its atoms in the space) should be known. From the structure of a vital protein of an organism that causes a human disease, it is possible to develop medicines for treatment of the disease. To find a protein structure, biophysical methods, as Crystallography of X-Ray and Magnetic Nuclear Resonance, have been employed. However, the use of those methods have practical restrictions that impede the determination of several protein structures. Aiming to overcome such limitation, computational methods for the problem of protein structure prediction (PSP) has been investigated. Several classes of computational methods have been developed for PSP. Among them, ab initio approaches are very important since they use no previous information from other protein structure, only the sequence of amino acids of the protein and the Ramachandran graph are employed. The ab initio PSP is a combinatorial problem that involves relatively large instances in practice, i. e. proteins in general have hundreds or thousands of variables to be determined. To deal with such problem, metaheuristics as Genetic Algorithms (GAs) have been investigated. The solutions generated by a GA are evaluated by the calculus of the potencial energies of the protein. Among them, the calculation of the interaction of van der Waals energy is computationally intense making the evolutionary process of a GA very slow even for non-large proteins. This work investigated techniques to significantly reduce the running time for that calculus. Basically, we proposed modifications parallelization of the resultant algorithm using MPI and OpenMP techniques. The results show that such calculus can be 1.500 times faster when applying the techniques investigated in this work for large proteins

Este trabalho foca o desenvolvimento de algoritmos de otimização para o problema de PSP puramente ab initio. Algoritmos que melhor exploram o espaço de potencial de soluções podem, em geral, encontrar melhores soluções. Esses algoritmos podem beneficiar ambas abordagens de PSP, tanto o modelo ab initio quanto os baseados em conhecimento a priori. Pesquisadores tem mostrado que Algoritmos Evolutivos Multiobjetivo podem contribuir significativamente no contexto do problema de PSP puramente ab initio. Neste contexto, esta pesquisa investiga o Algoritmo Evolutivo Multiobjetivo baseado em Tabelas aplicado ao PSP puramente ab initio, que apresenta interessantes resultados para proteínas relativamente simples. Por exemplo, um desafio para o PSP puramente ab initio é a predição de estruturas com folhas-. Para trabalhar com tais proteínas, foi desenvolvido procedimentos computacionalmente eficientes para estimar energias de ligação de hidrogênio e solvatação. Em geral, estas não são consideradas no PSP por abordagens que combinam métodos de otimização e conhecimento a priori. Considerando somente van der Waals e eletrostática, as duas energias de interação que mais contribuem para a definição da estrutura de uma proteína, com as energias de ligação de hidrogênio e solvatação, o problema de PSP tem quatro objetivos. Problemas combinatórios (tais como o PSP), com mais de três objetivos, geralmente requerem métodos específicos capazes de lidar com muitos critérios. Para resolver essa limitação, este trabalho propõe um novo método para a otimização dos muitos objetivos, chamado Algoritmo Evolutivo Multiobjetivo com Muitas Tabelas (AEMMT). Esse método executa uma amostragem mais adequada do espaço de funções objetivo e, portanto, pode mapear melhor as regiões promissoras deste espaço. A capacidade de lidar com muitos objetivos capacita o AEMMT a utilizar melhor a informação oriunda das energias de solvatação e de ligação de hidrogênio, e então predizer estruturas com folhas- e algumas proteínas relativamente mais complexas. Do ponto de vista computacional, o AEMMT é um novo método que lida com muitos objetivos (mais de dez) encontrando soluções relevantes / This work focuses on the development of optimization algorithms for the purely ab initio Protein Structure Prediction (PSP) problem. Algorithms that better explore the space of potential solutions can in general find better solutions. Such algorithms can benefit both ab initio and template-based PSP, that uses priori knowledge. Researches have shown that Multiobjective evolutionary algorithms can contribute significantly in the context of purely ab initio PSP. In this context, this research investigates the Multiobjective Evolutionary Algorithm based on Tables applied to purely ab initio PSP, which has shown interesting results for relatively simple proteins. For example, one challenge for purely ab initio PSP is the prediction of structures with -sheets. To work with such proteins, this research has developed computationally efficient procedures to estimate hydrogen bond and solvation energies. In general, they are not considered by PSP approaches combining optimization methods with priori knowledge. Only by considering van der Waals and electrostatic, the two interaction energies that mostly contribute to defining a protein structure, and the hydrogen bond and solvation energies, the PSP problem has four objectives. Combinatorial problems (such as the PSP) with more than three objective usually require specific methods capable of dealing with many goals. To address this limitation, we propose a new method for many objective optimization, called Multiobjective Evolutionary Algorithm with Many Tables (MEAMT). This method performs a more adequate sampling of the space of objective functions and, therefore, can better map the promising regions of this space. The ability of dealing with many objectives enables the MEANT to better use information generated by solvation and hydrogen bond energies, and then predict structures with -sheets and some relatively complex proteins. From the computational point of view, the MEAMT is a new method for dealing with many objectives (more than ten) finding relevant solutions

Das Auffinden spezifischer Strukturmerkmale integraler Membranproteine ist eine wichtige Grundlage zum Verständnis der Stabilität und Faltung und die notwendige Voraussetzung zur Modellierung ihrer Raumstruktur. Die Aminosäurezusammensetzung der innerhalb des hydrophoben Bereichs der Membranabschnitte befindlichen alpha-Helices wird entscheidend durch dieses Milieu determiniert, wie der Vergleich mit beta-Barrel Membranproteinen und alpha-Helices globulärer Proteine zeigt. Die Untersuchung der Phi-, Psi- und Chi1-Winkel lieferte gleichwohl eine Reihe von signifikanten Besonderheiten welche speziell bei einigen polaren (Asn, Asp) bzw. aromatischen Aminosäuren (Trp, Tyr) deutlich miteinander korrelieren. Der Anteil kürzerer, der für alpha-Helices typischen i, i+4 Wasserstoffbrückenbindungen ist innerhalb der Membran insgesamt höher, was ebenfalls mit der Verschiebung der Phi- und Psi- Winkel korreliert. Die Geometrie von alpha-Helices integraler Membranproteine ist damit näherungsweise ideal. An den Enden von Membranhelices und in den Helixturns werden neben Pro und Gly häufig polare, amphiphile oder aromatische Aminosäuren gefunden. Die polaren und amphiphilen Aminosäuren liegen überwiegend in den Helixturns und im Bereich der polaren Lipidköpfchen auf der Seite der Helix, welche der Membran zugewandten ist. Zur Fettschicht im Zentrum der Membran hin ragen umgekehrt vorherrschend lipophile Aminosäureseitenketten. Dieser Gradient ist folgerichtig auch an der Aminosäurezusammensetzung der Helixcaps erkennbar, welche die transmembranen Helices zumeist intra- und extrazellulär abschließen. Helixcaps alpha-helikaler Membrandomänen sind somit spezifische, von den klassischen Caps globulärer Proteine unterscheidbare Strukturmotive. Die konsequente Trennung der Untersuchungen der alpha-helikalen Abschnitte innerhalb der hydrophoben Lipiddoppelschicht von den Helixenden im polar-wässrigen Milieu ermöglicht außerdem die Identifikation vieler Aminosäurepräferenzen für exponierte (Leu, Ile), oder verdeckte Positionen (Ser, Asn, Cys). Umgekehrt wie bei den alpha-Helices globulärer Proteine ist die Atomzusammensetzung der Solvent exponierten Aminosäureseitenketten im Zentrum der Membran doppelt so hydrophob wie im Proteininnern. / Sorting out structural patterns that are specific for integral membrane proteins is a crucial base for understanding their stability and folding and a valuable source for the modelling of their tertiary structure. The detailed comparison of alpha-helical with beta-barrel membrane proteins and alpha-helices of globular proteins pointed out, that the amino acid composition of integral membrane proteins is overwhelmingly directed by the influence of the surrounding hydrophobic milieu. Nevertheless the investigation of the phi-, psi und chi1-angles yielded remarkable peculiarities of alpha-helical membrane proteins that correlate strongly in particular for some polar (asn, asp) and aromatic (trp, tyr) amino acids. The portion of the shorter and stronger i, i+4 H-bonds, that is typically found in alpha-helices is higher within the borders of the membrane. This observation confirms the observed shift of the phi- and psi- angles as well. Consequently the geometry of alpha-helices in membrane proteins is nearly ideal. At the ends of alpha-helices and in the turns of integral membrane proteins pro, gly and those amino acids are predominant that contain polar, amphiphilic or aromatic side chains. Whilst the aromats are equally positioned at the protein inside, the polar or amphiphilic amino acids are largely found at the helix turns, or at the side of the helix that contacts the polar lipid head groups. In contrast merely hydrophobic side chains face the lipophilic tails of the fatty acids in the core of the membrane. This gradient consequently shapes the amino acid composition of the helix caps that frequently coat the transmembrane helices on both sides of the membrane, too. Hence helix caps of transmembrane domains are substantially specific structural patterns that must be distinguished from the classical caps, known from alpha-helices of globular proteins. The clear distinction of the alpha-helical parts that are in contact with the lipophilic tails of the fatty acids, from the helix ends that stretch out into the polar aqueous solvent, advances the identification of amino acid preferences either lipid-exposed (Leu, Ile) or protein-buried (Ser, Asn, Cys). Finally, in sharp contrast to the helices of globular proteins, in the core of the membrane, the atomic composition of the solvent exposed amino acid side chains is twice as hydrophobic as inside the protein.