Complex systems are those which exhibit one or more of the following inter-related behaviours: 1. Nonlinear behaviour: the component parts do not act in linear ways, that is the superposition of the actions of the parts is not the output of the system. 2. Emergent behaviour: the output of the system may be inexpressible in terms of the rules or equations of the component parts. 3. Self-organisation: order appears from the chaotic interactions of individuals and the rules they obey. 4. Layers of description: in which a rule may apply at some higher levels of description but not at lower layers. 5. Adaptation: in which the environment becomes encoded in the rules governing the structure and/or behaviour of the parts (in this case strictly agents) that undergo selection in which those that are by some measure better become more numerous than those that are not as “fit”. A single cell is a complex system: we cannot explain all of its behaviour as simply the sum of its parts. Similarly, DNA structures, social networks, cancers, the brain, and living beings are intricate complex systems. This thesis tackles all of these topics from a complex systems approach. I have skirted some of the philosophical issues of complex systems and mainly focussed on appropriate tools to analyse these systems, addressing important questions such as: • What is the best way to extract information from DNA? • How can we model and analyse mutations in DNA? • Can we determine the likely spread of both viruses and ideas in social networks? • How can we model the growth of cancer? • How can we model and analyse interactions between genes in such living systems as the fruit fly, cancers, and humans? • Can complex systems techniques give us some insight into the human brain? / http://proxy.library.adelaide.edu.au/login?url= http://library.adelaide.edu.au/cgi-bin/Pwebrecon.cgi?BBID=1290759 / Thesis (Ph.D.)-- School of Electrical and Electronic Engineering, 2007

In biology, regulatory networks are sets of macromolecules, mostly proteins and RNAs that interact to execute task. The main players in regulatory networks are DNAbinding proteins, also called transcription factors as they modulate the first step in gene expression. A gene regulatory network (GRN) is a set of genes or proteins that interact with each other to control a specific cell function. Gene regulatory networks are important in development, differentiation and to respond to environmental cues. Gene regulatory networks (GRNs) are the on-off switches of a cell operating at the gene and/or protein level. The modeling methods can be broadly categorized into continuous and discrete. In this work , we dedicate attention to discrete models on cell senescence models for Astrocyte [35], the modelling of drug synergies to control gastric cancer [38], and we also wrote a paper about Discrete and Continuous Model, advantage or disadvantage of these models and a list of available softwares for using these kind of approaches. / Em biologia, redes regulatórias são conjuntos de macromoléculas, principalmente proteínas e RNAs que interagem para executar uma tarefa. As proteínas de ligação de DNA, também chamadas de fatores de transcrição, são as principais executoras nas redes regulatórias, visto que modulam o primeiro passo na expressão gênica. Uma rede genética regulatória (RRG) é um conjunto de genes ou proteínas que interagem uns com os outros para controlar uma função celular específica. Redes regulatórias são importantes no desenvolvimento, diferenciação e para responder aos sinais ambientais. Elas são os botões de liga/desliga de uma célula operando no nível do gene e/ou proteína. Seus métodos de modelagem podem ser geralmente classificados em contínuos e discretos. Neste trabalho, dedicamos atenção aos modelos discretos em senescência celular para astrócitos [35], a modelagem de sinergias de drogas para controle do câncer gástrico [38] e também escrevemos um artigo sobre Modelos Discretos e Contínuos, vantagens e desvantagens desses modelos e listagem dos softwares disponíveis para uso nesse tipo de abordagem.

Cette thèse concerne l'ingénierie des systèmes complexes à partir d'une dynamique souhaitée. En particulier, nous nous intéressons aux populations d'oscillateurs et aux réseaux de régulation génétique. Dans une première partie, nous nous fondons sur une hypothèse, introduite en neurosciences, qui souligne le rôle de la synchronisation neuronale dans le traitement de l'information cognitive. Nous proposons de l'utiliser sur un plan plus large pour étudier le traitement de l'information par des populations d'oscillateurs. Nous discutons des isochrons de quelques oscillateurs classés selon leurs symétries dans l'espace des états. Cela nous permet d'avoir un critère qualitatif pour choisir un oscillateur. Par la suite, nous définissons des procédures d'impression, de lecture et de réorganisation de l'information sur une population d'oscillateurs. En perspective, nous proposons un système à couches d'oscillateurs de Wilson-Cowan. Ce système juxtapose convenablement synchronisation et désynchronisation à travers l'utilisation de deux formes de couplage: un couplage continu et un couplage par pulsation. Nous finissons en proposant une application de ce système: la détection de contours dans une image. En deuxième partie, nous proposons d'utiliser une approche par contraintes pour identifier des réseaux de régulation génétique à partir de connaissances partielles sur leur dynamique et leur structure. Le formalisme que nous utilisons est connu sous le nom de réseaux d'automates booléens à seuil ou réseaux Hopfield-semblables. Nous appliquons cette méthode, afin de déterminer le réseau de régulation de la morphogenèse florale d'Arabidopsis thaliana. Nous montrons l'absence d'unicité des solutions dans l'ensemble des modèles valides (ici, 532 modèles). Nous montrons le potentiel de cette approche dans la détermination et la classification de modèles de réseaux de régulation génétique. L'ensemble de ces travaux mène à un certain nombre d'applications, en particulier dans le développement de nouvelles méthodes de stockage de l'information et dans le design de systèmes de calcul non conventionnel. / This thesis is concerned by the engineering of complex systems from a desired dynamics. Particularly, we are interested by populations of oscillators and genetical regulatory networks. In a first part, we start from a hypothesis introduced in neuroscience, which highlight the role of neural synchronization in the cognitive processing. We propose to use this hypothesis in a more general panorama to investigate the computing with populations of oscillators. We discuss about the isochrons of few oscillators selected according to their symmetry in the state space. Therefore, we define procedures for making footprints, for reading and for reorganizing information by a population of oscillators. As a perspective, we propose a system of lattices of Wilson-Cowan oscillators organized in several interconnected layers. This system properly mixes synchronization and desynchronization by using two types of coupling : pulsed and continuous coupling. At the end of this part, we propose to use this system in order to detect the edges of an image. In the second part, we propose a constraint-based approach to determine the structure of genetic regulatory networks starting from incomplete knowledge on their structure and their dynamics. The formalism we use is widely called thresholded Boolean automata networks or Hopfield-like networks. As an proof of concept, we apply this method to determine the regulatory network of Arabidopsis thaliana flower morphogenesis. We obtain 532 valid models instead of one unique solution and then classify them by using structural robustness criteria. By this way, we showed the potential of this approach in determining and classifying thresholded Boolean automata networks like genetic regulatory networks or neural networks. This works leads to many applications, in particular the developpement and the design of new methods for processing information and the design of systems of unconventional computing.

L'étude des grands modèles de réseaux biologiques par l'utilisation d'outils d'analyse et de simulation conduit à un grand nombre de prédictions. Cela soulève la question de savoir comment identifier les prédictions intéressantes de nouveaux phénomènes, qui peuvent être confrontés à des données expérimentales. Les techniques de vérification formelle basées sur le model checking constituent une technologie puissante pour faire face à cette augmentation d'échelle et de complexité pour l'analyse de ces réseaux. L'application de ces techniques est par contre difficile, pour plusieurs raisons. Premièrement, le domaine de la biologie des systèmes a mis en évidence quelques propriétés dynamiques du réseau, comme la multi-stabilité et les oscillations, qui ne sont pas facilement exprimables avec les logiques temporelles classiques. Deuxièmement, la difficulté de poser des questions pertinentes et intéressantes en logique temporelle est difficile pour les utilisateurs non-experts. Enfin, la plupart des modèles existants et des outils de simulation ne sont pas capables d'appliquer des techniques de model checking d'une manière transparente. La mise en œuvre des approches développées dans ce travail contribue à enlever des obstacles pour l'utilisation de la technologie de vérification formelle en biologie. Leur application a été validée sur l'analyse et la simulation de deux modèles biologiques complexes. / The study of large models of biological networks by means of analysis and simulation tools leads to large amounts of predictions. This raises the question of how to identify interesting predictions of novel phenomena that can be confronted with experimental data. Formal verification techniques based on model-checking have recently been used to the analysis of these networks, providing a powerful technology to keep up with this increase in scale and complexity. The application of these techniques is hampered, however, by several key issues. First, the systems biology domain brought to the fore a few properties of the network dynamics like multistability and oscillations, that are not easily expressed using classical temporal logics. Second, the problem of posing relevant and interesting questions in temporal logic, is difficult for non-expert users. Finally, most of the existing modeling and simulation tools are not capable of applying model-checking techniques in a transparent way. The approaches developed in this work lower the obstacles to the use of formal verification in systems biology. They have been validated on the analysis and simulation of two real and complex biological models. / O estudo de redes biológicas tem originado o desenvolvimento de modelos cada vez mais complexos e detalhados. O estudo de redes biológicas complexas utilizando ferramentas de análise e simulação origina grandes quantidades de previsões. Isto levanta a questão de como identificar previsões interessantes de novos fenómenos que possam ser comparados com dados experimentais. As técnicas de verificação formal baseadas em model-checking têm sido usadas na análise destas redes, fornecendo uma tecnologia poderosa para acompanhar o aumento de escala e complexidade do problema. A aplicação destas técnicas tem sido dificultada por um conjunto importante de factores. Em primeiro lugar, em biologia de sistemas têm sido tratadas diversas questões acerca da dinâmica da rede, como a multi-estabilidade e oscilações, que não são facilmente expressas usando lógicas temporais clássicas. Em segundo lugar, o problema de como elaborar perguntas relevantes em lógica temporal, é difícil para o utilizador comum. Por último, a maioria das ferramentas de modelação e simulação não estão preparadas para a aplicação de técnicas de model-checking de forma transparente. Os métodos desenvolvidos nesta tese aliviam os obstáculos no uso da verificação formal em biologia de sistemas. Estes métodos foram validados através da análise e simulação de dois modelos biológicos complexos.

A cana-de-açúcar é uma gramínea C4 usada por séculos como a principal fonte de açúcar e mais recentemente para obtenção de etanol. Devido a sua grande importância no cenário econômico mundial, estudos em cana-de-açúcar são cada vez mais importantes no sentido de prover informações que possam levar ao aumento de produtividade para suprir tanto a demanda interna quanto externa. No entanto, a quantidade de dados moleculares e biotecnológicos disponíveis está muito aquém do necessário, e investimentos na obtenção de novos conhecimentos serão necessários se quisermos evoluir neste campo, assim como manter o nosso país como líder na produção de etanol. Neste trabalho, conduzimos experimentos em campo para comparar variedades contrastantes para a tolerância ao déficit hídrico e realizamos diagnósticos fisiológicos e moleculares para diferenciar as variedades. O estresse hídrico levou à diminuição do crescimento e desenvolvimento de todas as três variedades analisadas. A variedade RB855536 foi identificada como a menos produtiva das três, visto que em condições de déficit hídrico ela diminui mais seu crescimento, sofre mais efeitos do estresse oxidativo, acumula mais osmólitos e tem o ciclo de Calvin menos ativo. A variedade RB867515 teve um melhor desempenho, não acumulou osmólitos, não teve aumento na concentração de prolina, teve sinais menores de estresse oxidativo e um melhor funcionamento do ciclo de Calvin. Além disto, nós observamos a indução de transcritos na via de resposta do ABA e proteínas relacionadas com a fotossíntese, transporte de água e dobramento protéico. A variedade RB92579 teve um padrão similar ao encontrado para a RB867515, mas teve uma maior fotossíntese, um menor estresse oxidativo e uma menor perda de pigmentos. Nossos dados avançaram na identificação de genes envolvidos na resposta ao déficit hídrico em cana-de-açúcar assim como permitiram distinguir as variedades utilizadas no estudo. A partir de uma prospecção inicial de dados de transcriptoma previamente obtidos pelo grupo, foram selecionados fatores de transcrição associados a características agronômicas de interesse. Produzimos anticorpos para 5 TFs de cana-de-açúcar e padronizamos a metodologia de ChIP-Seq utilizando a plataforma de sequenciamento Roche 454. Uma análise dos dados de ChIP-Seq usando anticorpos para a RNA Polimerase II nos permitiu detectar contaminações com DNA humano, provavelmente pela utilização de gDNA como controle das reamplificações, assim como a detecção de mapeamento de muitas regiões repetitivas, o que pode ser normal, podendo indicar locais de ligação da Polimerase II ou então background. O mapeamento de praticamente todos os dados de sorgo em cana evidenciou a similaridade entre estes organismos, mas a maior quantidade de mapeamento em cana evidencia uma maior complexidade de seu genoma. Os resultados foram importantes por permitir o estabelecimento de uma metodologia de mapeamento de regiões regulatórias no genoma da cana-de-açúcar e significa um importante passo no estabelecimento de redes regulatórias associadas a características agronômicas de interesse / Sugarcane is a C4 grass used for centuries as the main source of sugar and more recently for ethanol production. We have seen an increasing interest in recent years to understand sugarcane to improve yield and supply the increasing world demand. However, the amount of molecular data available is far from ideal, and investments in acquiring new knowledge will be necessary to progress in this field if we want to maintain our country as a pioneer in ethanol production. In this work, we conducted field experiments to compare varieties contrasting to drought tolerance and performed physiological and molecular analysis. The stress condition led to reduced growth and development of all three varieties. The variety RB855536 was identified as the least productive, suffered more effects of oxidative stress, has accumulated more osmolytes and has the Calvin cycle less active. The variety RB867515 had a better performance, did not accumulate osmolytes, had no increase in the concentration of proline, had minor signs of oxidative stress and a better functioning of the Calvin cycle. In addition, we observed an induction of transcripts in the ABA response pathway and proteins related to photosynthesis, water transport and protein folding. The variety RB92579 had a pattern similar to that found in RB867515, but presented increased photosynthesis, lower oxidative stress and a lower loss of pigment. We succeded in identifying genes involved in the response to water stress in sugarcane as well as in distinguishing the varieties used in the study. We selected transcription factors associated with agronomical traits from the transcriptome data previously obtained by the group. We produced antibodies for 5 sugarcane TFs and standardized the methodology of ChIP-Seq using the 454 sequencing platform. Data analysis of ChIP-Seq using RNA Pol II antibody allowed us to detect contamination with the human genome, probably due to the use of gDNA in the reamplification control, as well as to map the detection repetitive regions, which may be binding sites of Pol II or background. The data reveals that sorghum and the sugarcane genome are similar despite the complexity of sugarcane. The results were important since they allow the beggining of studies to map regulatory regions in the sugarcane genome, as well as the uncovery of regulatory networks associated with agronomic traits of interest

Um problema importante na área de Biologia Sistêmica é o de inferência de redes de regulação gênica. Os avanços científicos e tecnológicos nos permitem analisar a expressão gênica de milhares de genes simultaneamente. Por \"expressão gênica\'\', estamos nos referindo ao nível de mRNA dentro de uma célula. Devido a esta grande quantidade de dados, métodos matemáticos, estatísticos e computacionais têm sido desenvolvidos com o objetivo de elucidar os mecanismos de regulação gênica presentes nos organismos vivos. Para isso, modelos matemáticos de redes de regulação gênica têm sido propostos, assim como algoritmos para inferir estas redes. Neste trabalho, focamos nestes dois aspectos: modelagem e inferência. Com relação à modelagem, estudamos modelos existentes para o ciclo celular da levedura (Saccharomyces cerevisiae). Após este estudo, propomos um modelo baseado em redes Booleanas probabilísticas sensíveis ao contexto, e em seguida, um aprimoramento deste modelo, utilizando cadeias de Markov não homogêneas. Mostramos os resultados, comparando os nossos modelos com os modelos estudados. Com relação à inferência, propomos um novo algoritmo utilizando o paradigma de crescimento de semente de genes. Neste contexto, uma semente é um pequeno subconjunto de genes de interesse. Nosso algoritmo é baseado em dois passos: passo de crescimento de semente e passo de amostragem. No primeiro passo, o algoritmo adiciona outros genes à esta semente, seguindo algum critério. No segundo, o algoritmo realiza uma amostragem de redes, definindo como saída um conjunto de redes potencialmente interessantes. Aplicamos o algoritmo em dados artificiais e dados biológicos de células HeLa, mostrando resultados satisfatórios. / A key problem in Systems Biology is the inference of gene regulatory networks. The scientific and technological advancement allow us to analyze the gene expression of thousands of genes, simultaneously. By \"gene expression\'\' we refer to the mRNA concentration level inside a cell. Due to this large amount of data, mathematical, statistical and computational methods have been developed in order to elucidate the gene regulatory mechanisms that take part of every living organism. To this end, mathematical models of gene regulatory networks have been proposed, along with algorithms to infer these networks. In this work, we focus in two aspects: modeling and inference. Regarding the modeling, we studied existing models for the yeast (Saccharomyces cerevisiae) cell cycle. After that, we proposed a model based on context sensitive probabilistic Boolean networks, and then, an improvement of this model, using nonhomogeneous Markov chain. We show the results, comparing our models against the studied models. Regarding the inference, we proposed a new algorithm using the seed growing paradigm. In this context, a seed is a small subset of genes. Our algorithm is based in two main steps: seed growing step and sampling step. In the first step, the algorithm adds genes into the seed, according to some criterion. In the second step, the algorithm performs a sampling process on the space of networks, defining as its output a set of potentially interesting networks. We applied the algorithm on artificial and biological HeLa cells data, showing satisfactory results.

I extended an artificial life platform called GReaNs (the name stands for Gene Regulatory evolving artificial Networks) to explore the evolutionary abilities of biologically inspired Spiking Neural Network (SNN) model. The encoding of SNNs in GReaNs was inspired by the encoding of gene regulatory networks. As proof-of-principle, I used GReaNs to evolve SNNs to obtain a network with an output neuron which generates a predefined spike train in response to a specific input. Temporal pattern recognition was one of the main tasks during my studies. It is widely believed that nervous systems of biological organisms use temporal patterns of inputs to encode information. The learning technique used for temporal pattern recognition is not clear yet. I studied the ability to evolve spiking networks with different numbers of interneurons in the absence and the presence of noise to recognize predefined temporal patterns of inputs. Results showed, that in the presence of noise, it was possible to evolve successful networks. However, the networks with only one interneuron were not robust to noise. The foraging behaviour of many small animals depends mainly on their olfactory system. I explored whether it was possible to evolve SNNs able to control an agent to find food particles on 2-dimensional maps. Using ring rate encoding to encode the sensory information in the olfactory input neurons, I managed to obtain SNNs able to control an agent that could detect the position of the food particles and move toward it. Furthermore, I did unsuccessful attempts to use GReaNs to evolve an SNN able to control an agent able to collect sound sources from one type out of several sound types. Each sound type is represented as a pattern of different frequencies. In order to use the computational power of neuromorphic hardware, I integrated GReaNs with the SpiNNaker hardware system. Only the simulation part was carried out using SpiNNaker, but the rest steps of the genetic algorithm were done with GReaNs.

Cells must regulate their metabolism in order to grow, adapt to changes in nutrient availability and maintain homeostasis. Flux, or the turnover of metabolites, through the metabolic network can be regulated at the allosteric and transcriptional levels. While study of allosteric regulation is limited to biochemical examination of individual proteins, transcriptional control of metabolism can be explored at a systems level. We endeavored to elucidate transcriptional mechanisms of metabolic flux regulation in the model organism Caenorhabditis elegans (C. elegans). We also worked to create a visual tool to explore metabolic pathways that will support future efforts in the research of metabolic gene regulation. C. elegans is a small, free-living nematode that feeds on bacteria and experiences a high level of diversity in nutrient level and composition. Previously, we identified a mechanism by which the essential cofactor, vitamin B12, regulates the expression of genes involved in the degradation of propionate, referred to as B12‑mechanism‑I. This mechanism functions to prevent the toxic accumulation of propionate and requires the TFs NHR-10 and NHR-68. Using genetic screens as well as transcriptomic and metabolomic approaches, we discover a second mechanism by which vitamin B12 regulates metabolic gene expression: B12-mechanism-II. Unlike B12-mechanism-I, B12-mechanism-II is independent of propionate, requires the transcription factor NHR-114 and functions to maintain the metabolic activity of the Methionine/S-adenosylmethionine cycle in a tightly regulated regime. We also present WormPaths, an online resource that allows visualization of C. elegans metabolic pathways and enables metabolic pathway enrichment of user-uploaded transcriptomic data.

La modélisation de systèmes biologiques est devenue indispensable pour comprendre les phénomènes complexes et émergents issus d'influences partiellement connues, et pour envisager de contrôler un système altéré dans le but de restaurer un comportement physiologique. Tout modèle, quel que soit son paradigme sous-jacent, fait intervenir des paramètres gouvernant sa dynamique mais les mesures expérimentales ne permettent généralement pas de les identifier et cela reste l'un des problèmes majeurs de la modélisation. Cette thèse propose une méthode automatique d'identification des paramètres dynamiques de systèmes biologiques dans un cadre de modélisation hybride. Le cadre hybride choisi découpe l'espace des phases selon l'activité des entités biologiques, et associe à chacun de ces sous-espaces une vitesse d'évolution de chacun des composants. Nous proposons une logique de Hoare en temps continu ainsi qu'un calcul de plus faible précondition qui, à partir d'observations expérimentales qualitatives et chronométriques, construit les contraintes minimales sur les paramètres du modèle pour qu’il soit compatible avec les observations. Ce calcul mène à un problème de satisfaction de contraintes sur les réels et nous montrons que celui-ci peut être résolu par le solveur AbSolute.Le prototype Holmes BioNet développé au cours de cette thèse peut non seulement automatiser le processus d'identification des valeurs des paramètres à partir des observations expérimentales, mais aussi simuler l'évolution du modèle obtenu afin de le comparer avec les traces expérimentales. Nous utilisons ce prototype pour modéliser le couplage des cycles cellulaire et circadien. / Modelling biological systems has become instrumental to understand complex and emerging phenomena resulting from partially known influences, and to consider controlling an altered system in order to restore a physiological behaviour. Any model, independent of the underlying paradigm, involves parameters governing its dynamics. However, experimental measurements generally do not allow their identification and this remains one of the major problems of modelling. This PhD proposes an automatic method for identifying the dynamic parameters of biological systems in a hybrid modelling framework. The chosen hybrid framework splits the phase space according to the activity of the biological entities, and associates to each of these subspaces a celerity for each of the components. We introduce a continuous time Hoare logic as well as its weakest precondition calculus which, from qualitative and chronometrical experimental observations, constructs the minimum constraints on the model parameters making it compatible with the observations. This calculus leads to a Constraint Satisfaction Problem on real numbers and we show that it can be solved by the AbSolute solver.The Holmes BioNet prototype developed during this PhD can not only automate the parameter identification process from experimental data, but also simulate the evolution of the obtained model in order to compare it with experimental traces. We use this prototype to model the coupling of the cellular and circadian cycles.

Maintaining cellular homeostasis is a complex task, which involves monitoring energy states and essential nutrients, regulating metabolic fluxes to accommodate energy and biomass needs, and preventing buildup of potentially toxic metabolic intermediates and byproducts. Measures aimed at maintaining a healthy cellular economy inherently depend on the composition of nutrients available to the organism through its diet. We sought to delineate links between dietary composition, metabolic gene regulation, and physiological responses in the model organism C. elegans. As a soil-dwelling bacterivore, C. elegans encounters diverse bacterial diets. Compared to a diet of E. coli OP50, a diet of Comamonas aquatica accelerates C. elegans developmental rate, alters egg-laying dynamics and shortens lifespan. These physiological responses are accompanied by gene expression changes. Taking advantage of this natural, genetically tractable predator-prey system, we performed genetic screens i) in C. elegans to identify regulators of diet-responsive genes, and ii) in E. coli and Comamonas to determine dietary factors driving transcriptional responses in C. elegans. We identified a C. elegans transcriptional program that regulates metabolic genes in response to vitamin B12 content in the bacterial diet. Interestingly, several B12- repressed metabolic genes of unknown function are highly activated when B12- dependent propionyl-CoA breakdown is impaired, and inactivation of these genes renders animals sensitive to propionate-induced toxicity. We provide genetic and metabolomic evidence in support of the hypothesis that these genes form a parallel, B12-independent, β-oxidation-like propionate breakdown shunt in C. elegans, similar to the pathway utilized by organisms like yeast and plants that do not use vitamin B12.