Texte remanié: Diss.--Munich--Ludwig-Maximilians-Universität, 1993. / Bibliogr. p. 195-209.

abstract: The human transcriptional regulatory machine utilizes hundreds of transcription factors which bind to specific genic sites resulting in either activation or repression of targeted genes. Networks comprised of nodes and edges can be constructed to model the relationships of regulators and their targets. Within these biological networks small enriched structural patterns containing at least three nodes can be identified as potential building blocks from which a network is organized. A first iteration computational pipeline was designed to generate a disease specific gene regulatory network for motif detection using established computational tools. The first goal was to identify motifs that can express themselves in a state that results in differential patient survival in one of the 32 different cancer types studied. This study identified issues for detecting strongly correlated motifs that also effect patient survival, yielding preliminary results for possible driving cancer etiology. Second, a comparison was performed for the topology of network motifs across multiple different data types to identify possible divergence from a conserved enrichment pattern in network perturbing diseases. The topology of enriched motifs across all the datasets converged upon a single conserved pattern reported in a previous study which did not appear to diverge dependent upon the type of disease. This report highlights possible methods to improve detection of disease driving motifs that can aid in identifying possible treatment targets in cancer. Finally, networks where only minimally perturbed, suggesting that regulatory programs were run from evolved circuits into a cancer context. / Dissertation/Thesis / Masters Thesis Biomedical Engineering 2020

Protein regulatory networks are the hallmark of many important biological functionalities. Two of these functionalities are mammalian cell cycle progression and near-perfect adaptive responses. Modeling and simulating these functionalities are crucial stages to understanding and predicting them as systems-level properties of cells. In the context of the mammalian cell cycle, the timing of DNA synthesis, mitosis and cell division is regulated by a complex network of biochemical reactions that control the activities of a family of cyclin-dependent kinases. The temporal dynamics of this reaction network is typically modeled by nonlinear differential equations describing the rates of the component reactions. This approach provides exquisite details about molecular regulatory processes but is hampered by the need to estimate realistic values for the many kinetic constants that determine the reaction rates. To avoid this problem, modelers often resort to "qualitative" modeling strategies, such as Boolean switching networks, but these models describe only the coarsest features of cell cycle regulation. In this work, we describe a hybrid approach that combines features of continuous and discrete networks. The model is evaluated in terms of flow cytometry measurements of cyclin proteins in asynchronous populations of human cell lines. Using our hybrid approach, modelers can quickly create quantitatively accurate, computational models of protein regulatory networks found in various contexts within cells. Large-scale protein regulatory networks, such as the one that controls the progression of the mammalian cell cycle, also contain small-scale motifs or modules that carry out specific dynamical functions. Systematic characterization of smaller, interacting, network motifs whose individual behavior is well known under certain conditions is therefore of great interest to systems biologists. We model and simulate various 3-node network motifs to find near-perfect adaptation behavior. This behavior entails that a system responds to a change in its environmental cues, or signals, by coming back nearly to its pre-signal state even in the continued presence of the signal. We let various topologies evolve in their parameter space such that they eventually stumble upon a region where they score well under a pre-defined scoring metric. We find many such parameter sample sets across various classes of topologies. / Ph. D.

Ce mémoire propose des outils d'analyse des données fonctionnelles (ADF). Cette analyse s'applique aux données dont la structure peut être considérée comme une fonction. Dans ce travail, nous avons fait, avec des exemples pratiques sous R, un parcours sur une partie des méthodes existantes. D'abord, nous avons commencé par une petite introduction générale sur l'ADF où nous avons parlé non seulement des différentes bases classiques (base de Fourier, base polynomiale, base spline) qui permettent de représenter les données fonctionnelles en les rapprochant avec une combinaison linéaire de fonctions de base, mais aussi du lissage qui permet de convertir des données observées discrètes en courbes coutinues (par les moindres carrés ou par la pénalisation). Au vu du problème de variabilité en phase des données fonctionnelles, nous avons parlé des différentes méthodes d'alignement des courbes (par points de repère et alignement continu). Ensuite, nous avons étudié le clustering de données fonctionnelles. Il s'agit du processus qui permet de regrouper en cluster des fonctions similaires. Plusieurs méthodes de clustering exitent, mais nous nous sommes focalisés principalement sur les méthodes basées sur la distance, plus particulièrement sur la méthode K-moyennes. Également, le K-moyennes combinant simultanément l'alignement de manière globale et le clustering des courbes a été présentée. Et à travers des exemples pratiques, nous avons pu trouver la structure sous-jacente de données (identification de certaines tendances des données). Enfin, nous avons terminé par présenter le K-moyennes probabiliste avec alignement local qui prend en compte la forme locale des courbes, ce qui n'est pas le cas avec les méthodes susmentionnées qui traitent les courbes dans leur intégralité. Cette méthode, développée par Cremona et Chiaromonte, permet le regroupement local des courbes et la découverte de motifs fonctionnels ("formes" typiques) dans ces courbes. Sous R, cette méthode est fournit par le package probKMA.FMD (ProbKMA-based Functional Motif Discovery) qui est en construction. Des exemples pratiques ont été montrés afin de bien tester les codes et les fonctions du package. / This thesis proposes tools for functional data analysis (FDA). FDA is used on data whose structure can be considered as a function. In this work, we have looked at some of the existing methods using practical examples with R. First, we started with a short general introduction on FDA in which we discussed the different classical bases (Fourier basis, polynomial basis, Spline basis) which allow to represent functional data through a linear combination of basis functions. We also considered smoothing, which allows the conversion of discrete observed data into continuous curves (by least squares or by rougghness penalty). Regarding the problem of phase variability of functional data, we discussed different methods of curve alignment (by landmarks and continuous alignment). Second, we studied the clustering of functional data, which permits to group similar functions into clusters. Several clustering methods exist, but we mainly focused on distance-based methods, particularly on the K-means. Moreover, we considered a K-means combining simultaneously global alignment and clustering of curves. Through practical examples, we were able to find the underlying structure of the data (identification of certain patterns in the data). Finally, we presented the probabilistic K-means with local alignment which takes into account the local shape of the curves,as opposed to the previously mentioned methods which treat the curves in their entirety. This method, developed by Cremona and Chiaromonte, allows the local clustering of curves and the discovery of functional motifs ("shapes") in these curves. In R, this method is provided by the probKMA.FMD (ProbKMA-based Functional Motif Discovery) package which is under construction. Practical examples have been shown to test the codes and functions of the package.

Découvert en 1977, l'épissage est une étape de maturation post-transcriptionnelle consistant à rabouter les exons et éliminer les introns d'un ARN pré-messager. Pour que l'épissage soit correctement pris en charge par l'épisome et ses protéines auxiliaires, différents signaux sont présents le long de la séquence de l'ARN pré-messager. Il est maintenant reconnu que près de la moitié des mutations pathogènes chez l'homme impactent l'épissage, aboutissant à un dysfonctionnement du gène. Il est ainsi indispensable pour les biologistes d'être capables de détecter ces signaux sur une séquence génomique.Cette thèse a donc pour but de concevoir de nouveaux algorithmes permettant d'apporter la puissance de calcul des ordinateurs au service de la biologie de l'épissage. La solution proposée, Human Splicing Finder (HSF), est capable de prédire les trois types de signaux d'épissage à partir d'une séquence quelconque extraite du génome humain. Nous avons évalué l'efficacité de prédiction d'HSF dans l'ensemble des situations associées à des mutations pathogènes pour lesquelles il a été démontré expérimentalement leur impact sur l'épissage et par rapport aux autres algorithmes de prédiction. Parallèlement à ces apports directs tant pour la connaissance des processus biologiques de l'épissage que pour le diagnostic, les nouvelles approches thérapeutiques génotype-spécifiques peuvent également bénéficier de ces nouveaux algorithmes. Ainsi HSF permet de mieux cibler les oligonucléotides anti-sens utilisés pour induire le saut d'exon dans la myopathie de Duchenne et les dysferlinopathies.La reconnaissance récente de l'intérêt majeur de l'épissage dans des domaines aussi variés que la recherche fondamentale, la thérapeutique et le diagnostic nécessitaient un point central d'accès aux signaux d'épissage. HSF a pour objet de remplir ce rôle, en étant régulièrement mis à jour pour intégrer de nouvelles connaissances, et est d'ores et déjà reconnu comme un outil de référence. / Discovered in 1977, splicing is a post-transcriptional maturation process that consists in link-ing exons together and removing introns from a pre-messanger RNA. For splicing to be cor-rectly undertaken by the spliceosome and its auxiliary proteins, several signals are located along the pre-messanger RNA sequence. Nearly half of pathogenous mutations in humans are now recognized to impact splicing and leading to a gene dysfunction. Therefore it is es-sential for biologists to detect those signals in any genomic sequence.Thus, the goals of this thesis were to conceive new algorithms: i) to identify splicing signals; ii) to predict the impact of mutations on these signals and iii) to give access to this information to researchers thanks to the power of bioinformatics. The proposed solution, Human Splicing Finder (HSF), is a web application able to predict all types of splicing signals hidden in any sequence extracted from the human genome. We demonstrated the prediction's efficiency of HSF for all situations associated with pathogenous mutations for which an impact on splicing has been experimentally demonstrated. Along with these direct benefits for the knowledge of biological processes for splicing and diagnosis, new genotype-specific therapeutic approaches can also benefit from these new algorithms. Thus, HSF allows to better target antisense olignucleotides used to induce exon skipping in Duchenne myopathy and dysferlinopathies.The recent recognition of the major interest of splicing in various domains such as fundamen-tal research, therapeutics and diagnosis needed a one stop shop for splicing signals. HSF has for object to fulfill this need, being regularly updated to integrate new knowledge and is already recognized as an international reference tool.

Des millions de réactions chimiques sont décrites dans des bases de données sous la forme de transformations de graphes moléculaires. Cette thèse propose différentes méthodes de fouille de donnés pour extraire des motifs pertinents contenus dans ces graphes et ainsi aider les chimistes à améliorer leurs connaissances des réactions chimiques et des molécules. Ainsi on commence par montrer comment le problème central de la recherche des schémas de réactions fréquents peut se résoudre à l'aide de méthodes existantes de recherche de sous-graphes fréquents. L'introduction du modèle général des motifs les plus informatifs permet ensuite de restreindre l'analyse de ces motifs fréquents à un nombre réduit de motifs peu redondants et représentatifs des données. Si l'application du modèle aux bases de réactions permet d'identifier de grandes familles de réactions, le modèle est inadapté pour extraire les schémas caractéristiques de méthodes de synthèse (schémas CMS) dont la fréquence est trop faible. Afin de surmonter cet obstacle, est ensuite introduite une méthode de recherche heuristique fondée sur une contrainte d’intervalle entre graphes et adaptée à l’extraction de motifs de très faible fréquence. Cette méthode permet ainsi de déterminer à partir d'exemples de réactions et sous certaines conditions le schéma CMS sous-jacent à une réaction donnée. La même approche est ensuite utilisée pour traiter le problème de la classification supervisée de sommets ou d'arêtes fondée sur leurs environnements puis exploitée pour évaluer la formabilité des liaisons d'une molécule. Les résultats produits ont pu être analysés par des experts de la synthèse organique et sont très encourageants. / Millions of chemical reactions are described in databases as transformations of molecular graphs. This thesis proposes different data-mining methods to extract relevant patterns included in those graphs and therefore to help chemists in improving knowledge about chemical reactions and molecules. One first shows how the central problem of searching frequent reaction patterns can be solved using existing graph-mining methods. Introducing the general model of most informative patterns then allows experts to reduce the analysis of these frequent patterns to a very small set of non-redundant patterns characteristic of data. If the application of this model to reaction database identifies large and characteristic families of reactions, the model doesn't allow in practice the extraction of reaction patterns characteristic of synthesis methods (abbr. CSM patterns) as their frequencies are far too low. In order to overcome this problem, is introduced a heuristic search algorithm based on a graph interval constraint and able to extract patterns with very low frequency. Thus this method determines from examples of chemical reactions and under some conditions the CSM pattern underlying a given input reaction. The same approach is then used to address the problem of supervised classification of vertices or edges based on their environment and then applied to evaluate formability of bonds in molecules. Experimental results have been analyzed by experts and are very encouraging.

Ferramentas bioinformática têm se tornado a escolha para auxiliar pesquisadores tanto para a anotação de novos genes, como para estudar genes em condições fisiológicas de interesse. Entre essas ferramentas destacam-se os algoritmos de agrupamento filogenético e os algoritmos de predição de padrões curtos de DNA, como, por exemplo, predições de sítios para ligação de fatores de transcrição. Desenvolver uma abordagem mista com o objetivo de agrupar genes baseando-se unicamente nos sinais transcricionais preditos em suas seqüências é um desafio de difícil transposição. No presente trabalho, apresentamos nossos resultados para tentar superar tal limitação que podem ser subdividos em duas seções: a primeira aonde desenvolvemos uma abordagem para a melhoria das predições computacionais de sítios de ligação e a segunda, onde passamos a agrupar genes com base nos seus sinais transcricionais preditos em seqüências conservadas flanqueadoras. A primeira seção de nosso trabalho foi focada no estudo de uma seqüência de indução de transcrição próxima ao gene Aldh1a2 de camundongo aonde foram preditos sítios para fatores de transcrição que foram posteriormente testados biologicamente e se mostraram associados ao controle da expressão desse gene. A partir de uma profunda pesquisa bibliográfica, nós determinamos um grupo de 57 fatores de transcrição já associados com a especialização de subpopulações de neurônios durante o desenvolvimento neuroembrionário de vertebrados. Nossa abordagem de seleção de sítios de alto valor biológico foi agora testada em seqüências conservadas próximas a cada um desses genes que codificam esses fatores de transcrição associados e os sítios de ligação para fatores de transcrição foram preditos. Tais sítios foram contabilizados e utilizados com entrada para nossa abordagem de agrupamento. A análise dos resultados do agrupamento determinou que, nossa abordagem se mostrou suficientemente sensível para construir uma árvore solução com boas relações com os padrões, já conhecidos, de expressão para esses genes agrupados. Essa abordagem poderá ser utilizada tanto para anotar funcionalmente genes de interesse quanto para minerar informações dentro de um grupo de genes previamente selecionado. / Bioinformatics tools are becoming the choice for aiding scientists for gene annotation and for studying gene in physiological conditions of interest. Among those efforts, phylogenetics clustering algorithms and tools for predicting short DNA patterns, such as binding sites for transcription factor, are outlined as essential. To develop a mixture procedure merging this two distant fields of bioinformatics research is a challenge hard to overcome. In the present study, we present our results of trying to overcome such limitation and it be easily subdivided in two distinct sections: initially we develop a procedure to improve the computational prediction of binding site for transcription factors and the second one where genes were grouped based solely in their transcriptional patterns predicted in conserved flanking sequences. The first section of the present study was focused in the study of an enhancer near Aldh1a2 gene in mouse where binding sites were predicted and latter biologically tested and showed strong influence in expression control of this gene. By a comprehensive bibliographic research we determined a group of 57 transcription factors which were already associated with neuron subpopulations specialization during the neuroembryonary development in vertebrates. Our computational procedure for selection of high biological value binding sites was applied in conserved flanking sequence in each of these genes encoding these associated transcription factors and a large group of binding sites were predicted. This sites were counted and use as an input for our clustering procedure. Clustering results analyses determined that our procedure showed to be sufficiently sensible to construct a solution tree showing good relations with, already determined, expression patterns of grouped genes. This procedure could be for functionally annotation of genes and for data mining in a group of already determined genes of interest.

Mineração de dados tem sido cada vez mais aplicada em distintas áreas com o objetivo de extrair conhecimento interessante e relevante de grandes conjuntos de dados. Nesse contexto, aprendizado de máquina fornece alguns dos principais métodos utilizados em mineração de dados. Dentre os métodos empregados em aprendizado de máquina destacam-se os simbólicos que possuem como principal contribuição a interpretabilidade. Entretanto, os métodos de aprendizado de máquina tradicionais, como árvores e regras de decisão, não consideram a informação temporal presente nesses dados. Este trabalho propõe uma metodologia para extração de conhecimento de séries temporais por meio da extração de características e da identificação de motifs. Características e motifs são utilizados como atributos para a extração de conhecimento por métodos de aprendizado de máquina. Essa metodologia foi avaliada utilizando conjuntos de dados conhecidos na área. Foi realizada uma análise comparativa entre a metodologia e a aplicação direta de métodos de aprendizado de máquina sobre as séries temporais. Os resultados mostram que existe diferença estatística significativa para a maioria dos conjuntos de dados avaliados. Finalmente, foi realizado um estudo de caso preliminar referente ao monitoramento ambiental do reservatório da Usina Hidrelétrica Itaipu Binacional. Nesse estudo somente a identificação de motifs foi aplicada. Foram utilizadas séries temporais referentes à temperatura da água coletadas em distintas regiões do reservatório. Nesse estudo observou-se a existência de um padrão na distribuição dos motifs identificados para cada região do reservatório, corroborando com resultados consagrados na literatura / Data mining has been applied to several areas with the objective of extracting interesting and relevant knowledge from large data bases. In this scenario, machine learning provides some of the main methods employed in data mining. Symbolic learning are among the most used machine learning methods since these methods can provide models that can be interpreted by domain experts. However, traditional machine learning methods, such as decision trees and decision rules, do not take into account the temporal information present into data. This work proposes a methodology to extract knowledge from time series data using feature extraction and motif identification. Features and motifs are used as attributes for knowledge extraction performed by machine learning methods. This methodology was evaluated using some well-known data sets. In addition, we compared the proposed methodology to the approach that feeds machine learning algorithms with raw time series data. Results show that there are statistically significant differences for most of the data sets employed in the study. Finally, it is presented a preliminary study with environmental monitoring data from the Itaipu reservoir, made available by Itaipu Binacional. This study is restricted to the application of motif identification. We have used time series of water temperature collected from several regions of the reservoir. In this study, a pattern in motif distribution was observed for each region of the reservoir, agreeing with some well-known literature results

Ce travail de thèse a pour objectif d’étudier l’emploi des motifs géométriques, végétaux ou empruntés à l’architecture, dans le décor peint des monuments funéraires de Macédoine antique. Il s’agit d’une approche intégrale qui est effectuée à travers la documentation et l’analyse stylistique, technique et physico-chimique des motifs tels qu’ils apparaissent dans un riche corpus de soixante-huit monuments datés de l’époque hellénistique et le début de l’époque romaine. La thèse est organisée en trois parties. La première partie présente la documentation des motifs rencontrés, leur occurrence dans les monuments avec des observations sur le rendu, les combinaisons chromatiques et les pigments employés. Cette partie comprend une étude comparative des motifs dans l’art de l’antiquité en Grèce et ailleurs, et des commentaires sur l’origine et le symbolisme des motifs et leur transposition de la réalité ou de l’architecture dans la peinture.La deuxième partie concerne l’étude archéométrique des matériaux picturaux employés. Des prélèvements ponctuels ont été effectués sur le décor de vingt tombes et analysés au moyen de méthodes photographiques et physico-chimiques, in situ et en laboratoire. La présentation de la méthodologie et les résultats obtenus au cours de cette étude se fait de façon analytique et systématique. La troisième partie présente un catalogue illustré des tombes peintes de Macédoine classé selon leur type architectural, comprenant les données de fouille, la description du décor peint des surfaces pariétales et de l’aménagement intérieur et des informations concernant l’architecture et les matériaux picturaux. Les différents types de motifs ornementaux rencontrés et les types de moulures sur lesquelles ils sont réalisés sont présentés dans un lexique illustré. / The objective of this thesis is the study of the architectural, geometric and vegetal motifs that decorate the funerary monuments of Macedonia. An integral approach has been applied, combining detailed documentation with the stylistic, technical and physicochemical analysis of the motifs, as they appear in a rich corpus of sixty-eight monuments that date in the Hellenistic and early Roman period. The thesis is organised in three parts. The first part provides the documentation of the motifs, their occurrence in the monuments, along with observations on their design, the chromatic combinations and the pigments employed. This part includes a comparative study of the motifs in the art of ancient Greece and elsewhere, as well as commentaries on their origin and symbolism or the transformation from real objects or architectural elements into painted motifs. The second part concerns the archaeometrical study of the painting materials used. Samples from twenty tombs have been collected and analysed by means of photographical and physicochemical methods, both in situ and in the laboratory. The methodology and the results obtained during this study are presented systematically and in detail. The third part consists of an illustrated catalogue of the painted tombs of Macedonia, according to their architectural structure, which comprises the excavation data, the description of the painted decoration on the walls surface and on the funerary pieces of furniture as well as information on the architecture and the painting materials. The different ornamental motifs and types of moulding on which they are painted, are presented in an illustrated glossary of terms.

Die vorliegende Arbeit wurde im Rahmen einer kooperativen Promotion zwischen der TU Dresden und der Hochschule Mittweida angefertigt. In dieser werden neuartige, computerorientierte Ansätze für die Analyse von Membranproteinen vorgestellt. Membranproteine sind von essentieller Bedeutung für eine Vielzahl biologischer Prozesse innerhalb eines Organismus und stellen wichtige Zielmoleküle für eine breite Palette von Pharmazeutika dar. Ihre Sequenzen liefern wertvolle und teilweise noch nicht entschlüsselte Informationen über die dreidimensionale Struktur und funktionale Eigenschaften. Innerhalb der Proteomik und Genomik stellen Analysen an Membranproteinstrukturen einen wichtigen Teil für das Verständnis komplexer biologischer Prozesse dar. Im Zuge von Untersuchungen an Membranproteinen konnte eine Vielzahl kurzer wiederkehrender Muster, sogenannte Motive, in den Sequenzen von Membranproteinen beobachtet werden. Diese Motive unterstützen das Verständnis, wie sich Membranproteine in der Zellmembran falten. Im Fokus dieser Arbeit stehen derartige Sequenzmotive. Innerhalb von drei Projekten bilden ausschließlich sequenzmotivbasierte Ansätze die Grundlage für nähere Untersuchungen an Membranproteinstrukturen. Letztendlich liefern die in dieser Arbeit postulierten Methoden wertvolle Erkenntnisse über die strukturelle und funktionelle Rolle von Sequenzmotiven, auf deren Grundlage dazu beigetragen wird, den komplexen Aufbau von Membranproteinen besser verstehen zu können. Generell wird die Zusammenführung proteomischer und mutagener Informationen intensiviert. Nicht zuletzt wird dazu beigetragen, die in dieser Arbeit zusammengetragenen Ergebnisse, für die Planung von in vitro Experimenten sowie weiterführenden Arbeiten auf dem Gebiet der Membranproteinanalyse, der Wissenschaft zur Verfügung zu stellen. / The present work was written as part of a cooperative doctorate between the TU Dresden and the University of Applied Sciences Mittweida. In the doctoral thesis, novel, computer-oriented approaches for the analysis of membrane proteins are presented. Membrane proteins are essential for many cellular processes and are important targets for a wide range of pharmaceuticals. Their sequences provide valuable and partly not yet decoded information about their three-dimensional structure and functional characteristics. The analysis of membrane proteins is an important part for the understanding of complex biological processes in the context of proteomics and genomics. Research of membrane proteins revealed a large number of short, distinct sequence motifs. The motifs found so far support the understanding of the folded protein in the Membrane environment. In this dissertation, in three different approaches it is shown how the output of sequence motif-based methods can support the understanding of structural and functional properties of membrane proteins. In general, the junction of proteomic and mutagenic information is intensified. Last but not least, the results of this work are made available for the planning of in vitro experiments as well as for further works in the field of membrane Protein analysis.