No description available.

Background: To allow an immediate treatment of an infection with suitable antibiotics and bactericides or fungicides, there is an urgent need for fast and precise identification of the causative human pathogens. Methods based on DNA sequence comparison like 16S rRNA analysis have become standard tools for pathogen verification. However, the distinction of closely related organisms remains a challenging task. To overcome such limitations, we identified a new genomic target sequence located in the single copy gene for tRNA nucleotidyltransferase fulfilling the requirements for a ubiquitous, yet highly specific DNA marker. In the present study, we demonstrate that this sequence marker has a higher discriminating potential than commonly used genotyping markers in pro- as well as eukaryotes, underscoring its applicability as an excellent diagnostic tool in infectology. Results: Based on phylogenetic analyses, a region within the gene for tRNA nucleotidyltransferase (CCA-adding enzyme) was identified as highly heterogeneous. As prominent examples for pro- and eukaryotic pathogens, several Vibrio and Aspergillus species were used for genotyping and identification in a multiplex PCR approach followed by gel electrophoresis and fluorescence-based product detection. Compared to rRNA analysis, the selected gene region of the tRNA nucleotidyltransferase revealed a seven to 30-fold higher distinction potential between closely related Vibrio or Aspergillus species, respectively. The obtained data exhibit a superb genome specificity in the diagnostic analysis. Even in the presence of a 1,000-fold excess of human genomic DNA, no unspecific amplicons were produced. Conclusions: These results indicate that a relatively short segment of the coding region for tRNA nucleotidyltransferase has a higher discriminatory potential than most established diagnostic DNA markers. Besides identifying microbial pathogens in infections, further possible applications of this new marker are food hygiene controls or metagenome analyses.

While it is now widely accepted that new firms growth is essential for the foundation of economic dynamism, knowledge about this early growth is still scattered. Indeed, very little is known about how new firms grow and develop over time. What types of distinct growth patterns do those firms exhibit? How do these growth patterns and corresponding firms differ from each others in terms of development and strategic choices? To better understand the process of new firm growth, recent entrepreneurship research stresses that there is a strong need for a new conceptual scheme and new longitudinal research methods. This is actually one of the main entrepreneurship research challenges. In this context, our aim is to provide new insights regarding the process of new firm growth. In this research, we develop and test an original methodology allowing the empirical taxonomy of early growth trajectories across multiple sectors, integrating both the multidimensional and dynamic aspects of growth. Our approach applies principal component and cluster analysis to a large sample of firms, using financial and demographic data collected over time to identify in a systematic way distinct growth stages. We use then sequence analysis and a Markov chain approach to extract and compare the trajectories of individual firms over time. This allows the identification of a limited number of typical growth trajectories, which are adopted by the majority of firms in our sample. Finally, internal replication is performed to validate the growth trajectories identified and bivariate analysis is used to examine the link between the identified growth trajectories and the demographic characteristics of the corresponding firms. We have applied our methodology to a sample of 741 Belgian firms created between 1992 and 2002 and which have grown above micro-firm size. Our approach allowed identifying four distinct growth stages and seven typical growth trajectories, which remain valid for the six first years of the majority of the firms in our sample. This taxonomy of early growth trajectories is consistent with individual patterns already identified in the literature and appears not to be sector-dependent. The major contribution of this doctoral thesis is that, based on empirical evidence, early growth appears to be neither a continuous (or life cycle based) nor idiosyncratic (or completely random) process. It can be adequately described through a limited number of typical growth trajectories, valid across sectors. Thus, our research brings insight regarding how new firm evolve over time and therefore contributes to our understanding and appreciation of the heterogeneity of the growth trajectory phenomenon. Next, our research provides also an original methodological approach allowing the systematic analysis of growth trajectories, which deals with key limitations identified in the literature regarding the need for a multidimensional and dynamic study of growth across multiple sectors. Our findings indicate that this novel systematic approach is useful for taxonomy development and therefore contributes to reduce the gap between the complexity of new firm growth process and the standard approaches often mobilised to deal with it. Finally, while our findings provide empirical and methodological support in early development of new firms study, they also provide many implications to entrepreneurial research and practices. Further researches are needed to improve our understanding of the dynamic growth process of new ventures. It should explore which endogenous and exogenous factors might explain why a majority of start-ups follow the seven identified typical growth trajectories. It could be also highly relevant to refine our taxonomy by examining the relationship between innovative and technological sources and growth trajectories, both in high and low technological industries. Finally, we should test the accuracy of the proposed taxonomy across countries as well as beyond the early stage of new firm development.

近年來隨著電腦科技的進步，發展出了各種人機介面的應用。不同於傳統的滑鼠與鍵盤，新的人機介面往往在其他不同的領域能夠得到更好的發揮，眼動（eye movement）即是一個例子，利用眼球移動在螢幕上的眼動軌跡來操控電腦。另一方面，螢幕所呈現的內容也會影響使用者的眼動行為，例如閱讀文章時不同的文章內容就會影響使用者在看文章時的眼動行為，因此我們就可以從眼動軌跡去推測出使用者對於螢幕的內容（如文章或圖片）的反應。本論文的目的就是要找出閱讀文章時的眼動軌跡和閱讀者對該文章的理解程度之間的關連性，最後發展出一個系統，利用閱讀文章時的眼動軌跡就能去推測閱讀者對於文章的理解程度，而無需後續的閱讀測驗。 / 一般人通常對自身的眼動軌跡的進行方式不甚了解，認為閱讀文章的眼動應該是順著字句的進行而移動。其實仔細觀察細小的眼動軌跡會發現比想像中複雜的多，例如當我們閱讀到文意較為困難的片段時，通常都會放慢速度，或是反覆閱讀，又例如看到自己有興趣的主題時會凝視在某些區塊較久的時間等等行為，這些都是在各種不同的情境之下做出相對的反應。本論文提出了資料探勘的方法來分析複雜的眼動軌跡，能針對每個人不同的閱讀模式找出不同的眼動規則。此方法能適應各種使用者的閱讀習性，對閱讀理解程度的預測能達到更精確的效果。 / With the advance of the computer technology, there are many kinds of HCI (Human Computer interface) applications been developed in recent years. Different from mouse and keyboard, new interface will bring more benefits in different area, and eye-movement is an example. It extracts the trace of eye movement to control the computer. Furthermore, the screen content will also affect the eye-movement. For example, when the user reads the articles, the content will affect the eye-movement, so we can speculate the reaction of the user after seen the screen content (such as article or picture) by eye-movement. The goal of the paper is to find out the relationship between eye-movement and the degree of comprehension, and develop a system which can automatically measure reading comprehension without any follow-up comprehension test. / Most people don’t realize how there eye move. They think eye movements should be carried out along the words. In fact, eye movement will find more complex than imagined, for example, when we read the context of the article is more difficult, we often slow down or read over and over again. In another example, people will gaze at some of the subjects that they are interest in for a longer period of time. These reactions correspond to a variety of different situations. This paper presents a data mining approach to analyze complex eye movement. It can find out the different rules of eye movement for users who have different reading strategy. This method will be able to adapt to a variety of users reading habits, and make more accurate prediction with the degree of the comprehension.

Aplicando tecnologia de microarray, objetivamos traçar o perfil do programa de maturação das Mo-DC pulsadas com HIV autólogo inativado por AT-2, a fim de identificar marcadores específicos de ativação funcional e sugerir um perfil de expressão de genes úteis na identificação de respostas ao modelo in vitro das Mo-vacina DC. Essas informações podem ajudar a estabelecer assinaturas moleculares das funções celulares mais relevante para a melhoria das vacinas terapêuticas. O perfil transcricional foi analisado com base das vias celulares moduladas das Mo-DCs no estado imaturo, transitório e maduro. O HIV-1 inativado por AT-2 induz ativação de genes associados à apresentação de antígenos. Os conjuntos de genes do citoesqueleto podem influenciar a mudança de comportamento migratório das Mo-DCs ativadas. O aumento na expressão dos receptores celulares contribuem para o recrutamento de monócitos, DCs e macrófagos para o local da infecção. Além disso, modulam a resposta imune inata e adaptativa, incluindo a polarização das células Th e sub-regulação da resposta inflamatória, que pode interferir significativamente com a resposta imune. Coletivamente, o perfil transcricional das Mo-DCs induzido pelo HIV-1 inativado com AT-2 reflete uma significativa reprogramação imunológica e celular das células envolvidas na resposta imune do hospedeiro. Os resultados deste estudo focaram na interpretação de genes específicos dos perfis de transcrição das Mo-DCs como modelo terapêutico utilizado na vacina anti-HIV. As análises de assinaturas gene associado e sua correlação as respostas funcionais simplificam a identificação de indivíduos susceptíveis a vacina e a compreensão de eventuais falhas em ensaios clínicos. Microarray permitiu a análise quantitativa e simultânea da expressão de um elevado número de genes. Os estudos do perfil de expressão foram extremamente úteis para identificar os eventos moleculares e vias envolvidas nas funções de celular induzida por estímulos específicos. Em particular, os resultados sobre o padrão global da expressão dos genes subjacentes as modificações induzidas pelo HIV-1 inativado por AT-2, na fase inicial da administração do antígeno, pôde ser extremamente útil para a identificarmos marcadores de ativação e avaliar os efeitos biológicos que poderiam estar envolvidos para modificação e otimização de estratégias vacinação com Mo-DC / Applying microarray technology, we intend to profile the program to mature Mo-DC pulsed with autologous inactivated HIV by AT-2, in order to identify specific markers of functional activation and suggest a profile of expression of specific genes, useful identification of responders to in vitro model of Mo-DC vaccine. Such information may help to establish detailed molecular signatures of cellular functions most relevant to improving the therapeutic vaccines. The transcriptional profile was analyzed on the basis of the cellular pathways modulated in immature MoDC, transitional MoDC and mature MoDC. The AT-2-inactivated HIV-1 induction of MoDC results in the activation of genes associated with antigen presentation functions. A set of cytoskeletal genes that may potentially mediate shape change and migratory behavior of activated MoDC is also observed. The increase in the expression of immune receptors contribute to the recruitment of monocytes, DCs, and macrophages to the site of infection. Moreover, they modulate both innate and adaptive immune response, including the polarization of Th cells, and the down-regulation of the inflammatory response, which may significantly interfere with the immune response. Collectively, the transcriptional profile induced by AT-2-inactivated HIV-1 in MoDc reflects a significant cellular and immunological reprogramming of cells directly involved in the host immune response. The results of this study focused on the interpretation of specific genes of transcription profile of MoDC used in therapeutic HIV vaccine model. Supplementing the analyses with examination of associated gene signatures and their correlation to functional responses will simplify the identification of responsive vaccine individuals and the understanding of eventual failures in individuals enrolled in clinical trials. Microarray approach allows quantitative and simultaneous analysis of gene expression of a large amount of genes and the systematic studies of expression patterns are extremely useful for identify molecular events and key pathways involved in cellular functions induced by specific stimuli. In particular, data on the global pattern of gene expression underlying the modifications induced by AT-2-inactivated HIV-1 in MoDC, at early stages of antigen administration, may be extremely helpful for the identification of exclusive activation markers to trace the biological effects of modifications/optimizations of the MoDc vaccination strategy

Die vorliegende Arbeit untersucht mit bioinformatischen Methoden die biologische Bedeutung von Ähnlichkeiten in Kleinstrukturen und peptidischen Sequenzmotiven sowie lokaler und globaler Sequenzähnlichkeit. Der erste Teil der Arbeit behandelt chemische Ähnlichkeiten. Ausgehend von bekannten Inhibitoren der Fehlfaltung des Prionproteins wurde eine Datenbank pharmakologischer Wirkstoffe nach chemisch und strukturell ähnlichen Substanzen durchsucht und 16 Substanzen als neue potentielle Inhibitoren der Fehlfaltung vorgeschlagen. Der nächste Teil untersucht Ähnlichkeiten in Sequenzmotiven, die eine Interaktion mit Pex19, dem Importrezeptor für peroxisomale Membranproteine, vermitteln. In Zusammenarbeit mit einer experimentellen Arbeitsgruppe konnte die Bindestelle charakterisiert und Präferenzen für bestimmte Aminosäuren herausgearbeitet werden. Das Bindemotiv ist eine vermutlich helikale Region mit verzweigtkettigen aliphatischen und basischen Aminosäuren. Aus experimentellen Daten konnte eine positionsabhängige Vorhersagematrix erstellt und validiert werden. Die Beziehung zwischen lokalen Sequenzähnlichkeiten und der Konformation von Prolylbindungen in Proteinen ist Thema des dritten Teils. Die Aminosäurepräferenzen in der Nachbarschaft von cis- und trans-Prolylresten unterscheiden sich, und beide zeigen unterschiedliche Austauschpräferenzen bei Mutationen. Im Gegensatz zu lokaler Sequenzähnlichkeit ist eine globale Sequenzähnlichkeit von nur 20% ein wesentlich besserer Indikator für das Auftreten von cis-Prolylbindungen. Der letzte Teil befaßt sich mit inverser Sequenzähnlichkeit zwischen Proteinen, die wesentlich öfter auftritt als erwartet. Proteine aus einem nichtredundanten Datensatz wurden gleich- und gegenläufig aligniert und strukturelle Ähnlichkeiten zwischen den aufgefundenen Proteinpaaren untersucht. Es konnte gezeigt werden, daß bis auf kurze Sekundärstruktur-Einheiten eine inverse Sequenzähnlichkeit zwischen Proteinen keine strukturelle Ähnlichkeit impliziert. / This work is dealing with the biological impact of similarities between chemical structures, protein sequence motifs and local sequence surrounding as well as global sequence similarity. All four aspects are analyzed by computational methods. The first part is dealing with chemical similarities. Based on a recently published set of prion protein misfolding inhibitors, a data base of approved drugs has been screened for compounds with chemical and structural similarities to these substances. 16 drugs are proposed as new potential inhibitors of prion protein aggregation. The next part addresses similarities of sequence motifs which mediate the interaction with the peroxisomal membrane protein import receptor Pex19. In cooperation with an experimental group, the binding site could be characterized, and amino acid preferences of the different positions of the motif have been determined. The binding motif is a probably helical region of target proteins bearing branched aliphatic and basic residues. A position specific scoring matrix for the prediction of Pex19 binding sites could be generated and validated. The relation between local sequence similarity and prolyl bond conformation is examined in the third part. Amino acid preferences of neighboring residues differ between cis and trans prolyl residues, and both species show different amino acid exchange patterns upon mutation. In contrast to local sequence similarity, overall sequence similarity between proteins as low as 20% is a much better indicator for the occurrence of cis prolyl bonds. The last part focuses on inverse sequence similarity between proteins which occurs far more often than expected by chance. Proteins from a nonredundant data set have been aligned in parallel and antiparallel, and structural similarities between the detected protein pairs have been examined. It could be shown that, with the exception of short secondary structural elements, inverse sequence similarity does not imply structural similarity.

Resumo: A meningite é uma inflamação das membranas que revestem o sistema nervoso central. As principais etiologias desta doença são de origem infecciosa, podendo ser bacteriana, fúngica ou viral. A meningite pode ocorrer como uma infecção hospitalar e pode ser associada a trauma ou neurocirurgia. Quando diagnosticada após um procedimento neurocirúrgico, a maior parte dos agentes infecciosos causadores da meningite provem da microbiota endógena da pele e do cabelo. No entanto, existem casos no qual o agente etiológico não é diagnosticado pelas técnicas laboratoriais convencionais, como a cultura microbiológica e bacterioscopia, dificultando a prescrição de terapias adequadas. O objetivo deste estudo foi identificar os agentes causadores de meningite após neurocirurgia (MAN) utilizando técnicas de biologia molecular e comparando-as com a cultura microbiológica. Foram incluídas amostras de líquido cefalorraquidiano (LCR) de pacientes submetidos a neurocirurgia e pacientes submetidos a cirurgias eletivas com uso de raquianestesia durante o período de 2015 a 2016. A reação em cadeia da polimerase (PCR) foi utilizada para avaliação da presença do gene 16S do DNA ribossômico, comum em microrganismos de origem bacteriana, e o sequenciamento do mesmo para a identificação do agente etiológico. As amostras foram classificadas em 5 grupos de acordo com a suspeita clínica e dados quimiocitológicos do LCR: meningite bacteriana (MB) confirmada, MB possível, MB provável, MB improvável e sem MB, neste último grupo estão apenas pacientes submetidos a cirurgias eletivas. Das 51 amostras de LCR incluídas (43 pós-neurocirurgia e 8 pré-anestésica), 21 (41,2%) apresentaram cultura microbiológica negativa com PCR positiva, sendo: 3 (14,2%) MB possível, 4 (19,0%) MB provável, 13 (62,0%) MB improvável, 1 (4,8%) sem MB. Do total de 15 amostras positivas para PCR foi identificada ao menos a família filogenética, houve predomínio de microrganismos Gram negativos, somando 11 contra 4 Gram positivos. A identificação dos agentes etiológicos na MAN, incluindo os não detectados por métodos convencionais de identificação laboratorial, demonstraram que a biologia molecular pode complementar o diagnóstico colaborando de forma positiva, guiando o tratamento para o microrganismo específico ou sua família / Abstract: Meningitis is an inflammation of the membranes covering the central nervous system. The main causes of this disease are bacterial, fungal or viral agents. Meningitis may be associated with trauma or neurosurgery. When meningitis is diagnosed after a neurosurgical procedure, the most common microorganisms belong to skin and hair microbiota. However, there are cases in which the etiological agent is not diagnosed by conventional laboratory techniques, such as microbial culture and bacterioscopy, which makes it difficult to establish adequate therapies. The objective of this study is to identify agents causing meningitis after neurosurgery (MAN) using polymerase chain reaction (PCR) and sequencing of the 16S rRNA gene, compared to the conventional microbiological culture. Cerebrospinal fluid (CSF) was collected from 43 patients who had undergone neurosurgery and 8 patients during spinal anesthesia were included during the period from 2015 to 2016. Polymerase chain reaction (PCR) was used to evaluate the presence of the 16S ribosomal RNA gene, common in microorganisms of bacterial origin, and the sequencing of the same for the identification of the etiological agent. Samples were classified into five groups according to clinical data and CSF analysis: confirmed bacterial meningitis (MB), probable MB, possible MB, unlikely MB, no meningitis, in this last group are only patients submitted to elective surgeries. There were 51 CSF samples included (43 post neurosurgery and 8 pre-spinal anesthesia), 21 samples (41.2%) presented negative microbial culture and were PCR-positive, divided as: 3 (14.2%) probable MB, 4 (19%) possible MB, 13 (62%) unlikely MB and 1 (4.8%) no meningitis. From the total of 15 PCR-positive samples at least the phylogenetic family was identified with a predominance of Gram negative microorganisms, (11). The identification of etiologic agents in MAN, including those not detected by conventional laboratory identification methods, suggests molecular biology can complement the diagnosis and collaborate in guiding the treatment for the specific microorganism or its family

Diese Dissertation beschäftigt sich mit der Komplexität von Familienverläufen und beinhaltet vier empirische Studien. Die erste Studie untersucht, ob sich Familienverläufe in Geburtskohorten und Ländern unterscheiden, und ob sie mehr über die Zeit oder die Länder hinweg variieren. Anhand der SHARELIFE Daten wird ein Verfahren entwickelt, indem Komplexitätsmaße aus der Sequenzanalyse mit der Mehrebenenmodellierung zusammengeführt werden. Die zweite Studie untersucht, ebenfalls auf Basis der SHARELIFE Daten, folgende Fragen: Wie hängen Familienpolitik und -komplexität zusammen, und wird dieser Zusammenhang durch den Zeitpunkt des Eintreffens der jeweiligen Familienpolitik im Lebensverlauf moderiert? Um den Zusammenhang zwischen Familisierungs-, Defamilisierungs-, und Liberalisierungsindezes und Komplexität zu schätzen, werden weitere Datenquellen herangezogen. Die Zusammenhänge zwischen den Indizes und Komplexität werden mit Länder- und Kohorten Fixed-Effects geschätzt. Das dritte Kapitel untersucht auf Basis der NLSY79 und NLSY97 Daten den Zusammenhang zwischen elterlichen Ressourcen und Komplexität im jungen Erwachsenenalter, und ob sich dieser über Kohorten hinweg verändert. Die Ergebnisse zeigen, dass die Komplexität eher bei jungen Erwachsenen aus benachteiligten Familien angestiegen ist. Das vierte Kapitel verwendet Lebensverlaufs- und genetische Daten aus den USA (HRS) und ermittelt die Erblichkeit von Komplexität mittels GCTA. Es wird geschlussfolgert, dass die Zunahme von Komplexität, auch in den USA, relativ gering ist und Länderunterschiede viel bedeutsamer sind. Nicht kulturelle Veränderungen, sondern zunehmende ökonomische Unsicherheit und sozialpolitische Institutionen scheinen die wichtigsten Faktoren für kohorten- und länderspezifische Komplexitätsunterschiede zu sein. Abschließend lässt sich festhalten, dass genetische Faktoren die Komplexität ebenfalls beeinflussen und ihre Berücksichtigung die Vorhersagekraft von statistischen Modellen erhöhen kann. / This dissertation on family life course complexity revolves around four empirical studies. The first chapter investigates how family life courses vary across birth cohorts, how family life courses vary across countries, and whether family life courses vary more across birth cohorts or across countries. This study uses SHARELIFE and combines sequence complexity metrics with cross-classified multilevel modeling to quantify the proportion of variance attributable to cohort and country differences. The second chapter also uses SHARELIFE to address two research questions: what is the association between family policies and complexity and does the timing of family policies within the life course moderate this association? Data sources are combined to estimate the relationships between three family policy dimensions - familization, defamilization, and liberalization - and complexity. The associations between my policy indexes and complexity are estimated using country and time fixed effects regression models. The third chapter asks what is the association between parental resources and the early family life course complexity and has the association between parental resources and complexity changed across birth cohorts. NLSY79 and NLSY97 data show that complexity is higher among disadvantaged young adults. The fourth chapter applies life history and genetic data from the HRS to a GCTA to study the heritability of complexity. It is concluded that the increase in complexity, even in the United States, is relatively small and cross-national variation seems to be much more important. Rather than ideational change, increasing economic uncertainty and differences in national institutional arrangements are the most important factors for cross-national and cross-cohort differences in complexity. Finally, genetic factors matter for the complexity of individuals’ family life courses and could likely contribute to the predictive power of future statistical models.

Les Machines de Boltzmann restreintes (RBM) sont des modèles graphiques capables d’apprendre simultanément une distribution de probabilité et une représentation des données. Malgré leur architecture relativement simple, les RBM peuvent reproduire très fidèlement des données complexes telles que la base de données de chiffres écrits à la main MNIST. Il a par ailleurs été montré empiriquement qu’elles peuvent produire des représentations compositionnelles des données, i.e. qui décomposent les configurations en leurs différentes parties constitutives. Cependant, toutes les variantes de ce modèle ne sont pas aussi performantes les unes que les autres, et il n’y a pas d’explication théorique justifiant ces observations empiriques. Dans la première partie de ma thèse, nous avons cherché à comprendre comment un modèle si simple peut produire des distributions de probabilité si complexes. Pour cela, nous avons analysé un modèle simplifié de RBM à poids aléatoires à l’aide de la méthode des répliques. Nous avons pu caractériser théoriquement un régime compositionnel pour les RBM, et montré sous quelles conditions (statistique des poids, choix de la fonction de transfert) ce régime peut ou ne peut pas émerger. Les prédictions qualitatives et quantitatives de cette analyse théorique sont en accord avec les observations réalisées sur des RBM entraînées sur des données réelles. Nous avons ensuite appliqué les RBM à l’analyse et à la conception de séquences de protéines. De part leur grande taille, il est en effet très difficile de simuler physiquement les protéines, et donc de prédire leur structure et leur fonction. Il est cependant possible d’obtenir des informations sur la structure d’une protéine en étudiant la façon dont sa séquence varie selon les organismes. Par exemple, deux sites présentant des corrélations de mutations importantes sont souvent physiquement proches sur la structure. A l’aide de modèles graphiques tels que les Machine de Boltzmann, on peut exploiter ces signaux pour prédire la proximité spatiale des acides-aminés d’une séquence. Dans le même esprit, nous avons montré sur plusieurs familles de protéines que les RBM peuvent aller au-delà de la structure, et extraire des motifs étendus d’acides aminés en coévolution qui reflètent les contraintes phylogénétiques, structurelles et fonctionnelles des protéines. De plus, on peut utiliser les RBM pour concevoir de nouvelles séquences avec des propriétés fonctionnelles putatives par recombinaison de ces motifs. Enfin, nous avons développé de nouveaux algorithmes d’entraînement et des nouvelles formes paramétriques qui améliorent significativement la performance générative des RBM. Ces améliorations les rendent compétitives avec l’état de l’art des modèles génératifs tels que les réseaux génératifs adversariaux ou les auto-encodeurs variationnels pour des données de taille intermédiaires. / Restricted Boltzmann machines (RBM) are graphical models that learn jointly a probability distribution and a representation of data. Despite their simple architecture, they can learn very well complex data distributions such the handwritten digits data base MNIST. Moreover, they are empirically known to learn compositional representations of data, i.e. representations that effectively decompose configurations into their constitutive parts. However, not all variants of RBM perform equally well, and little theoretical arguments exist for these empirical observations. In the first part of this thesis, we ask how come such a simple model can learn such complex probability distributions and representations. By analyzing an ensemble of RBM with random weights using the replica method, we have characterised a compositional regime for RBM, and shown under which conditions (statistics of weights, choice of transfer function) it can and cannot arise. Both qualitative and quantitative predictions obtained with our theoretical analysis are in agreement with observations from RBM trained on real data. In a second part, we present an application of RBM to protein sequence analysis and design. Owe to their large size, it is very difficult to run physical simulations of proteins, and to predict their structure and function. It is however possible to infer information about a protein structure from the way its sequence varies across organisms. For instance, Boltzmann Machines can leverage correlations of mutations to predict spatial proximity of the sequence amino-acids. Here, we have shown on several synthetic and real protein families that provided a compositional regime is enforced, RBM can go beyond structure and extract extended motifs of coevolving amino-acids that reflect phylogenic, structural and functional constraints within proteins. Moreover, RBM can be used to design new protein sequences with putative functional properties by recombining these motifs at will. Lastly, we have designed new training algorithms and model parametrizations that significantly improve RBM generative performance, to the point where it can compete with state-of-the-art generative models such as Generative Adversarial Networks or Variational Autoencoders on medium-scale data.

O receptor ativado do peroxissomo γ (PPAR-γ) é regulador do metabolismo e diferenciação do tecido adiposo, sendo um alvo conhecido das tiazolidinedionas (TZD), utilizadas para o tratamento do diabetes tipo 2 (DM2). As TZD agem como um agente sensibilizador da ação da insulina nos tecidos periféricos e tem sido especulado que as TZDs podem ter um papel na função da célula , prevenindo perda de massa e melhorando a sua viabilidade a longo prazo. Este efeito seria supostamente mediado pela transcrição de genes que favoreceriam a lipólise, diminuindo o conteúdo intracelular de triglicérides e, portanto, diminuindo a lipotoxicidade. Entretanto, alguns estudos também mostraram efeito nulo ou mesmo deletério das TZDs sobre as ilhotas pancreáticas. Na realidade, o papel de genes-alvo para o PPAR- nas ilhotas pancreáticas é ainda pouco conhecido. Estudamos o perfil de expressão gênica induzido pelo tratamento com Pioglitazona (Pio), uma TZD aprovada e disponível para uso clínico no tratamento do DM2, em ilhotas pancreáticas murídeas em cultura primária, com concentrações normal e suprafisiológica de glicose no meio de cultura. As ilhotas foram obtidas de ratos wistar machos de dois meses de idade e isoladas pelo método do gradiente de Ficoll e então cultivadas em 5,6 mM ou 23 mM de glicose por 24h, sendo tratadas com Pio 10 M ou DMSO 0,1% (veículo). A Pioglitazona foi cedida pela Takeda Farmacêutica, Osaka, Japão. O RNA foi extraído com Trizol e purificado com o kit RNeasy (Qiagen). As amostras foram marcadas e hibridizadas no microarranjo de cDNA Mouse Panchip 13k, usando-se cinco replicatas biológicas diferentes para cada condição. A análise estatística dos dados do microarranjo foi feita com o uso do programa significance analysis of microarrays (SAM) com uso de taxa de descobrimento falso (FDR) de 20%. A análise das vias acometidas foi feita com o Ingenuity Pathway Analysis (www.ingenuity.com). Os resultados de expressão gênica foram confirmados por RT-qPCR. Em concentração de 5,6 mM de glicose no meio de cultura, 101 genes foram modulados pela Pio, sendo 49 regulados para cima, com aumento de sua expressão na presença da droga e 52 genes regulados para baixo. Em 23 mM de glicose, 1.235 genes foram afetados, sendo 621 para cima e 623 para baixo. A comparação entre as duas condições revelou 74 genes que foram modulados em ambas as concentrações de glicose. A análise das vias biológicas alteradas mostrou que genes relacionados ao metabolismo de lípides foram modulados em ambas as concentrações de glicose. Em 23 mM foi ainda significativo o grupo de genes relacionados a ciclo celular e morte celular que tiveram sua expressão modificada pela presença da droga na cultura. Este dado demonstrou que além de seus efeitos conhecidos nos adipócitos, o sensibilizador de insulina Pioglitazona modula a expressão de genes nas ilhotas pancreáticas, especialmente na presença de concentrações suprafisiológicas de glicose, afetando notadamente genes relacionados ao metabolismo lipídico, sendo vários deles ligados a lipogênese, como Srebf1, Scd2 e Fabp4 cujas expressões aumentaram em ambas as concentrações de glicose. Além disso foi observado aumento na expressão de genes com atividade pró-apoptótica como Tnf, Bad, Bax, Caspase4, Fadd e Myc. A Pioglitazona parece induzir um perfil gênico desfavorável em ilhotas pancreáticas mantidas em cultura em concentrações suprafisiológicas de glicose. / Peroxisome proliferator-activator receptor-γ (PPAR-γ) is a target for thiazolidinedione (TZD) antidiabetic drugs and a regulator of adipose tissue differentiation and metabolism. TZD act as an insulin sensitizing agent on peripheral tissues. It has been speculated that TZD could play a role on beta-cell function, preventing loss and improving viability in the long-term. This effect is supposed to be mediated through a potential benefit against lipotoxicity, favouring lypolisis and decreasing intracellular tryglicerides content. Nevertheless some studies also showed a lack or even a potential deleterious effect of TZD on islets. The role of PPAR-γ target genes in pancreatic islets is actually still largely unclear. We studied the gene expression profile induced by the treatment with Pioglitazone (Pio), an approved TZD for T2DM therapy, on rat pancreatic islets primary culture both at normal and supraphysiological glucose medium concentrations. Islets were obtained from 2 month-old, male, wistar rats and isolated through the Ficoll gradient method and then cultured with 5.6 mM or 23 mM of glucose concentration for 24h, being treated with Pio 10 µM or DMSO 0.1% (vehicle). Pioglitazone was provided by Takeda Pharmaceuticals, Osaka, Japan. RNA was extracted with Trizol (Sigma) and purified with RNeasy kit (Qiagen). Samples were labeled and then hybridized on the Mouse PanChip 13k cDNA microarray, using 5 different biological replicates for each test condition. Statistical Analysis of the microarray data was performed using significance analysis of microarrays (SAM) with a false discovery rate of 20%. Pathways assessment was performed through Ingenuity Pathway Analysis (www.ingenuity.com). Gene expression results were confirmed through RT-qPCR. At 5.6 mM glucose 101 genes were modulated by Pio, 49 upregulated and 52 downregulated. At 23 mM, 1,235 genes were affected, 612 upregulated and 623 downregulated. Comparison between both conditions revealed 74 genes that were similarly modulated at both glucose concentrations. Pathway analysis of perturbed genes revealed biologically relevant networks related to lipid metabolism at both glucose medium concentrations. At 23 mM, cell cycle and cell death pathways were significant modulated as well. These data demonstrates that in addition to known effect in adipocytes, the insulin sensitizing agent Pioglitazone modulates gene expression in pancreatic islets, especially in the presence of supraphysiological glucose concentrations, affecting especially lipid metabolism and mechanisms of cell death and cell cycle. Considering the ontology of modulated genes it seems to be a trend towards lypogenesis (increased Srebf1, Scd2 and Fabp4 RNA expressions) with Pio treatment also enhancing the abundance of some genes considered to be pro apoptotic like Tnf, Bad, Bax, Caspase4, Fadd and Myc. Pioglitazone seems to induce a negative gene expression profile in islets cultured at high glucose concentrations.