<p>In every living cell, millions of different types of molecules constantly interact and react chemically in a complex system that can adapt to fluctuating environments and extreme conditions, living to survive and reproduce itself. The information required to produce these components is stored in the genome, which is copied in each cell division and transferred and mixed with another genome from parent to child. The regulatory mechanisms that control biological systems, for instance the regulation of expression levels for each gene, has evolved so that global robustness and ability to survive under harsh conditions is a strength, at the same time as biological tasks on a detailed molecular level must be carried out with good precision and without failures. This has resulted in systems that can be described as a hierarchy of levels of complexity: from the lowest level, where molecular mechanisms control other components at the same level, to pathways of coordinated interactions between components, formed to carry out particular biological tasks, and up to large-scale systems consisting of all components, connected in a network with a topology that makes the system robust and flexible. This thesis reports on work that model and analyze complex biological systems, and the signals and noise that regulate them, at all different levels of complexity. Also, it shows how signals are transduced vertically from one level to another, as when a single mutation can cause errors in low level mechanisms, disrupting pathways and create systemwide imbalances, such as in type 2 diabetes. The advancement of our knowledge of biological systems requires both that we go deeper and towards more detail, of single molecules in single cells, as well as taking a step back to understand the organisation and dynamics in the large networks of all components, and unite the different levels of complexity.</p>

In every living cell, millions of different types of molecules constantly interact and react chemically in a complex system that can adapt to fluctuating environments and extreme conditions, living to survive and reproduce itself. The information required to produce these components is stored in the genome, which is copied in each cell division and transferred and mixed with another genome from parent to child. The regulatory mechanisms that control biological systems, for instance the regulation of expression levels for each gene, has evolved so that global robustness and ability to survive under harsh conditions is a strength, at the same time as biological tasks on a detailed molecular level must be carried out with good precision and without failures. This has resulted in systems that can be described as a hierarchy of levels of complexity: from the lowest level, where molecular mechanisms control other components at the same level, to pathways of coordinated interactions between components, formed to carry out particular biological tasks, and up to large-scale systems consisting of all components, connected in a network with a topology that makes the system robust and flexible. This thesis reports on work that model and analyze complex biological systems, and the signals and noise that regulate them, at all different levels of complexity. Also, it shows how signals are transduced vertically from one level to another, as when a single mutation can cause errors in low level mechanisms, disrupting pathways and create systemwide imbalances, such as in type 2 diabetes. The advancement of our knowledge of biological systems requires both that we go deeper and towards more detail, of single molecules in single cells, as well as taking a step back to understand the organisation and dynamics in the large networks of all components, and unite the different levels of complexity.

Lignin polymers provide natural rigidity to plant cell walls by forming complex molecular networks with polysaccharides such as cellulose and hemicellulose. This evolved strategy equips plants with recalcitrance to biological and chemical degradation. While naturally beneficial, recalcitrance complicates the use of inedible plant materials as feedstocks for biofuel production. Genetically modifying lignin biosynthesis is an effective way to generate varieties of bioenergy crops with reduced recalcitrance, but certain lignin-modified plants display undesirable phenotypes and/or unexplained effects on lignin composition, suggesting that the process and regulation of lignin biosynthesis is not fully understood. Given the intrinsic complexities of metabolic pathways in plants and the technical hurdles in understanding them purely with experimental methods, the objective of this dissertation is to develop novel computational tools combining static, constraint-based, and dynamic, kinetics-based modeling approaches for a systematic analysis of lignin biosynthesis in wild-type and genetically engineered plants. Pathway models are constructed and analyzed, yielding insights that are difficult to obtain with traditional molecular and biochemical approaches and allowing the formulation of new, testable hypotheses with respect to pathway regulation. These model-based insights, once they are verified experimentally, will form a solid foundation for the rational design of genetic modification strategies towards the generation of lignin-modified crops with reduced recalcitrance. More generically, the methods developed in this dissertation are likely to have wide applicability in similar studies of complex, ill-characterized pathways where regulation occurring at the metabolic level is not entirely known.

L'estudi de les diferències fenotípiques entre espècies, i entre individus, ha estat una de les grans qüestions fonamentals en els camps de la biologia evolutiva i la genètica. Ben aviat, es va fer palès que la regulació de l’expressió gènica tindria un paper clau en establir aquestes diferències de complexitat. L’adveniment de les tècniques massives de seqüenciació no van sinó confirmar aquesta visió primerenca. Avui dia coneixem un grapat de mecanismes que determinen aquestes diferències entre organismes, com són la divergència de seqüència proteica, la duplicació gènica o la divergència de la regió cis-reguladora, entre d’altres. En la darrera dècada, l’splicing alternatiu ha anat afermant-se com a mecanisme post-transcripcional freqüent i ha anat prenent protagonisme com a font de variabilitat de transcrits i isoformes proteiques, a més a més de jugar un paper regulador de l’expressió gènica. Per tant, l’splicing alternatiu és un ferm candidat a introduir diferències substancials al proteoma que expliquin la diversitat fenotípica entre organismes. Així doncs, aquest treball es va marcar com a objectiu aclarir fins a quin punt la variabilitat que introduïa l’splicing alternatiu tenia implicacions en el fenotip, quina era la seva conservació i si actuava de manera coordinada o independent amb d’altres mecanismes. En primer lloc, vam estudiar la relació que hi havia entre l’splicing alternatiu i les altres fonts moleculars de diversitat fenotípica i si era possible que l’splicing alternatiu pogués introduir variabilitat amb implicacions fenotípiques per si sola. A continuació, ens vam centrar en els mecanismes reguladors de l’expressió gènica basats en splicing alternatiu, analitzant les seves propietats i la seva conservació entre espècies. Finalment, vam examinar la implicació de l’splicing alternatiu en el fenomen de la domin{ncia gènica, ja que és un procés conegut que determina diferències fenotípiques intraespecífiques. El primer pas fou, doncs, comparar l’splicing alternatiu amb d’altres fonts moleculars de diferències fenotípiques: les divergències de la seqüència proteica, de la regió cis-reguladora del gen i de l’expressió gènica entre hum{ i ratolí. En un estudi massiu de les propietats de tots aquests fenòmens entre 13970 parelles d’ortòlegs, vam observar que l’splicing alternatiu podia introduir diferències abans que les altres variables poguessin fer-ho. Quan les identitats de seqüència proteica o de la regió cis-reguladora eren massa elevades com per introduir diferències, l’splicing alternatiu ja presentava patrons prou diferents en la concurrència d’splicing entre hum{ i ratolí. A més a més, la relació entre l’equivalència d’isoformes amb aquestes divergències també va resultar ser molt lleu, fet que ens va fer pensar que l’splicing alternatiu pot introduir isoformes específiques que contribueixin a les diferències entre espècies abans que les altres divergències puguin fer-ho. Pel que fa al segon bloc, vam investigar la conservació i propietats dels mecanismes reguladors de l’expressió gènica basats en AS. Primer de tot, vam confirmar la independència entre les divergències d’expressió gènica i l’splicing alternatiu, fet que ens indica que actuen a diferents nivells. A continuació, vam definir i classificar aquests mecanismes reguladors depenent com l’splicing alternatiu alterava l’arquitectura de dominis de les isoformes. La conservació d’aquests efectes, dels mecanismes reguladors basats en AS, va resultar ser baixa per tots els casos. Pel que fa als esdeveniments on es perdien un o més dominis a les isoformes alternatives, a més a més de ser baixa la conservació del mecanisme, també ho va ser l’equivalència dels esdeveniments d’splicing alternatiu. Així, tot i tenir efectes a nivell de seqüència no homòlegs, la funció es conservava, fet que ens porta a suggerir que aquests esdeveniments d’AS són un exemple de convergència funcional. Per últim, ens vam fixar en el procés de la dominància, abastament conegut, que introdueix diferències fenotípiques clares entre individus de la mateixa espècie, sobretot en el cas de malalties. Donat el fet que es coneixia una relació inversa entre paralogia i haploinsuficiència, per una banda, i paralogia i splicing per l’altra, sumat a la capacitat d’introduir variabilitat per part de l’splicing alternatiu, vam endegar aquest estudi amb la idea de descriure la relació entre dominància i splicing. El resultat final ens va mostrar una independència dels dos processos, fet que ens va fer qüestionar la relació entre paralogia i splicing alternatiu. Per la resta de variables estudiades, la caracterització de la dominància va concordar amb els resultats de treballs anteriors. / RESUMEN El estudio de las diferencias fenotípicas entre especies ha sido una de les cuestiones fundamentales de la biología evolutiva y la genética. Muy pronto fue evidente que la regulación de la expresión génica seria clava en el establecimiento de estas diferencias, tesis confirmada con las técnicas masivas de secuenciación actuales. Hoy en día, se conocen una serie de mecanismos que determinan estas diferencias, como son la divergencia de la secuencia proteica, la duplicación génica o la divergencia de la región cis-reguladora. En la última década, el splicing alternativo (AS) ha ido afianzándose como mecanismo post-transcripcional y ha ido tomando protagonismo como fuente de variabilidad de transcritos y isoformas, además de jugar un papel regulador de la expresión génica. Por lo tanto, el AS es un firme candidato a introducir diferencias sustanciales en el proteoma que expliquen la diversidad fenotípica entre organismos. Así pues, este trabajo se marcó como objetivo aclarar hasta qué punto la variabilidad que introducía el AS tenía implicaciones en el fenotipo, cuál era su conservación y si actuaba de manera coordinada o independiente con otros mecanismos. En primer lugar, estudiamos la relación que había entre el AS y las otras fuentes moleculares de diversidad fenotípica y si era posible que el AS pudiera introducir variabilidad con implicaciones fenotípicas por sí sola. A continuación, nos centramos en los mecanismos reguladores de la expresión génica basados en AS, analizando sus propiedades y su conservación entre especies. Finalmente, examinamos la implicación del AS en la dominancia génica. En el primer bloque comparamos el AS con otras fuentes moleculares de diferencias fenotípicas: las divergencias de la secuencia proteica, de la región cis-reguladora del gen y de la expresión génica entre humano y ratón. En un estudio masivo de las propiedades de todos estos fenómenos entre 13.970 ortólogos, observamos que el AS podía introducir diferencias antes que las otras variables pudieran hacerlo. Cuando las identidades de secuencia proteica o de la región cis-reguladora eran demasiado elevadas como para introducir diferencias, el AS ya presentaba patrones bastante diferentes en la concurrencia de AS entre humano y ratón. Además, la relación entre la equivalencia de isoformas con estas divergencias también resultó ser muy leve, lo que nos hizo pensar que el AS puede introducir isoformas específicas que contribuyan a las diferencias entre especies antes que las demás divergencias puedan hacerlo. En el segundo bloque investigamos la conservación y propiedades de los mecanismos reguladores de la expresión génica basados en AS. En primer lugar, confirmamos la independencia entre las divergencias de expresión génica y del AS, lo que nos indica que actúan a diferentes niveles. A continuación, definimos estos mecanismos reguladores dependiendo como el AS alteraba la arquitectura de dominios de las isoformas. La conservación de los mecanismos reguladores basados en AS resultó ser baja en todos los casos. En cuanto a los eventos donde se perdían uno o más dominios en las isoformas alternativas, también fue baja la equivalencia de los eventos de AS. Así, pese a tener efectos a nivel de secuencia no homólogos, la función se conservaba, lo que nos permite sugerir que éste es un escenario de convergencia funcional. Por último, nos fijamos en el proceso de la dominancia, largamente conocido, que introduce diferencias fenotípicas intraespecíficas. Dado que se conocía una relación inversa entre paralogía y haploinsuficiencia, por un lado, y paralogía y AS por la otra, sumado a la capacidad de introducir variabilidad por parte del AS, iniciamos este estudio con la idea de describir la relación entre dominancia y AS. El estudio nos mostró una independencia de los dos procesos, cuestionando así la relación entre paralogía y AS. Para el resto de variables estudiadas, la caracterización de la dominancia concordó con resultados de trabajos anteriores. / The study of phenotypic differences between species, and between individuals, has been one of the great fundamental questions in the fields of evolutionary biology and genetics. Soon, it became clear that the regulation of gene expression would have a key role in establishing these differences in complexity. The advent of mass sequencing techniques did confirm this view. Nowadays, we know a handful of mechanisms that determine these differences between organisms, such as protein sequence divergence, gene duplication and divergence of cis-regulatory region, among others. In the last decade, alternative splicing has been asserting itself as a post-transcriptional mechanism and frequently has taken center stage as a source of variability of transcripts and protein isoforms, and also as a key player in the regulation the gene expression. Therefore, alternative splicing is a strong candidate to introduce substantial differences in the proteome that could explain the phenotypic diversity among organisms. Thus, this work was intended to clarify to what extent the variability introduced the alternative splicing (AS) had implications for the phenotype, which was its conservation and if it acted in a coordinated or independent way relative to other mechanisms. First, we studied the relationship that existed between AS and other sources of molecular and phenotypic diversity and elucidate if AS could introduce phenotypic variability with its own implications. Then we focused on the regulatory mechanisms of gene expression based on AS, analyzing their properties and their conservation between species. Finally, we examined the involvement of AS in the phenomenon of genetic dominance, since it is a known process that determines intraspecific phenotypic differences. The first step was therefore to compare the AS with other sources of molecular phenotypic differences: differences in the protein sequence, the cis-regulatory region of the gene and gene expression between human and mouse. In a massive study of the properties of these phenomena among 13,970 pairs of orthologous, we observed that alternative splicing could introduce differences before other variables could do it. When the identities of protein sequence or cis-regulatory region were too high for introducing differences, AS patterns appeared quite different in the occurrence of splicing between human and mouse. Furthermore, we found that the relationship between the equivalence of isoforms with those differences was very mild, which made us think that AS can introduce specific isoforms that contribute to differences between species before other divergences can do it. Regarding the second section, we investigated the properties and the conservation of the regulatory mechanisms of gene expression based on AS. First, we confirmed the independence between the divergence of gene expression and AS, which indicates that they act at different levels. Then we defined and classified these regulatory mechanisms depending on how the AS altered the domain architecture of the isoforms. The conservation of these effects, the regulatory mechanisms based on AS, was found to be low for all cases. With regard to the events where they lost one or more domains in the alternative isoforms, in addition to the low conservation of the mechanism, it was also low the equivalence of alternative splicing events. So, despite having an non-homologue effect on the level of sequence, the function was preserved, which leads us to suggest that these AS events are an example of functional convergence. Finally, we studied the well known process of dominance which introduces clear phenotypic differences between individuals of the same species, especially in the case of diseases. Given the fact that it is known the inverse relationship between paralogy and haploinsufficiency and, in the other hand, the inverse relationship between paralogy and AS, adding to this the ability of introducing variability by AS, we undertook this study with the idea of describe the relationship between dominance and splicing. The final result showed us that they are two independent processes, which made us question the relationship between paralogy and AS. For the remaining variables, the characterization of the dominance results agreed with previous work.

Lipids are the main component of semipermeable cell membranes and linked to several important physiological processes. Shotgun lipidomics relies on the direct infusion of total lipid extracts from cells, tissues or organisms into the mass spectrometer and is a powerful tool to elucidate their molecular composition. Despite the technical advances in modern mass spectrometry the currently available software underperforms in several aspects of the lipidomics pipeline. This thesis addresses these issues by presenting a new concept for lipid identification using a customized query language for mass spectra in combination with efficient spectra alignment algorithms which are implemented in the open source kit “LipidXplorer”.

Despite numerous advances in experimental methodologies capable of addressing the various phenomenon occurring on metal surfaces, atomic scale resolution of the microscopic dynamics remains elusive for most systems. Computational models of the processes may serve as an alternative tool to fill this void. To this end, parallel molecular dynamics simulations of self-diffusion on metal surfaces have been developed and employed to address microscopic details of the system. However these simulations are not without their limitations and prove to be computationally impractical for a variety of chemically relevant systems, particularly for diffusive events occurring in the low temperature regime. To circumvent this difficulty, a corresponding coarse-grained representation of the surface is also developed resulting in a reduction of the required computational effort by several orders of magnitude, and this description becomes all the more advantageous with increasing system size and complexity. This representation provides a convenient framework to address fundamental aspects of diffusion in nonequilibrium environments and an interesting mechanism for directing diffusive motion along the surface is explored. In the ensuing discussion, additional topics including transition state theory in noisy systems and the construction of a checking function for protein structure validation are outlined. For decades the former has served as a cornerstone for estimates of chemical reaction rates. However, in complex environments transition state theory most always provides only an upper bound for the true rate. An alternative approach is described that may alleviate some of the difficulties associated with this problem. Finally, one of the grand challenges facing the computational sciences is to develop methods capable of reconstructing protein structure based solely on readily-available sequence information. Herein a checking function is developed that may prove useful for addressing whether a particular proposed structure is a viable possibility.

A subfield of artificial intelligence, machine learning (ML), is concerned with the development of algorithms that allow computers to &ldquo / learn&rdquo / . ML is the process of training a system with large number of examples, extracting rules and finding patterns in order to make predictions on new data points (examples). The most common machine learning schemes are supervised, semi-supervised, unsupervised and reinforcement learning. These schemes apply to natural language processing, search engines, medical diagnosis, bioinformatics, detecting credit fraud, stock market analysis, classification of DNA sequences, speech and hand writing recognition in computer vision, to encounter just a few. In this thesis, we focus on Support Vector Machines (SVMs) which is one of the most powerful methods currently in machine learning. As a first motivation, we develop a model selection tool induced into SVM in order to solve a particular problem of computational biology which is prediction of eukaryotic pro-peptide cleavage site applied on the real data collected from NCBI data bank. Based on our biological example, a generalized model selection method is employed as a generalization for all kinds of learning problems. In ML algorithms, one of the crucial issues is the representation of the data. Discrete geometric structures and, especially, linear separability of the data play an important role in ML. If the data is not linearly separable, a kernel function transforms the nonlinear data into a higher-dimensional space in which the nonlinear data are linearly separable. As the data become heterogeneous and large-scale, single kernel methods become insufficient to classify nonlinear data. Convex combinations of kernels were developed to classify this kind of data [8]. Nevertheless, selection of the finite combinations of kernels are limited up to a finite choice. In order to overcome this discrepancy, we propose a novel method of &ldquo / infinite&rdquo / kernel combinations for learning problems with the help of infinite and semi-infinite programming regarding all elements in kernel space. This will provide to study variations of combinations of kernels when considering heterogeneous data in real-world applications. Combination of kernels can be done, e.g., along a homotopy parameter or a more specific parameter. Looking at all infinitesimally fine convex combinations of the kernels from the infinite kernel set, the margin is maximized subject to an infinite number of constraints with a compact index set and an additional (Riemann-Stieltjes) integral constraint due to the combinations. After a parametrization in the space of probability measures, it becomes semi-infinite. We analyze the regularity conditions which satisfy the Reduction Ansatz and discuss the type of distribution functions within the structure of the constraints and our bilevel optimization problem. Finally, we adapted well known numerical methods of semiinfinite programming to our new kernel machine. We improved the discretization method for our specific model and proposed two new algorithms. We proved the convergence of the numerical methods and we analyzed the conditions and assumptions of these convergence theorems such as optimality and convergence.

This thesis consists of two parts. The first part focuses on high-dimensional classification problems in microarray experiments. The second part deals with forecasting problems with a large number of categories in predictors. Classification problems in microarray experiments refer to discriminating subjects with different biologic phenotypes or known tumor subtypes as well as to predicting the clinical outcomes or the prognostic stages of subjects. One important characteristic of microarray data is that the number of genes is much larger than the sample size. The penalized logistic regression method is known for simultaneous variable selection and classification. However, the performance of this method declines as the number of variables increases. With this concern, in the first study, we propose a new classification approach that employs the penalized logistic regression method iteratively with a controlled size of gene subsets to maintain variable selection consistency and classification accuracy. The second study is motivated by a modern microarray experiment that includes two layers of replicates. This new experimental setting causes most existing classification methods, including penalized logistic regression, not appropriate to be directly applied because the assumption of independent observations is violated. To solve this problem, we propose a new classification method by incorporating random effects into penalized logistic regression such that the heterogeneity among different experimental subjects and the correlations from repeated measurements can be taken into account. An efficient hybrid algorithm is introduced to tackle computational challenges in estimation and integration. Applications to a breast cancer study show that the proposed classification method obtains smaller models with higher prediction accuracy than the method based on the assumption of independent observations. The second part of this thesis develops a new forecasting approach for large-scale datasets associated with a large number of predictor categories and with predictor structures. The new approach, beyond conventional tree-based methods, incorporates a general linear model and hierarchical splits to make trees more comprehensive, efficient, and interpretable. Through an empirical study in the air cargo industry and a simulation study containing several different settings, the new approach produces higher forecasting accuracy and higher computational efficiency than existing tree-based methods.

How good is our understanding of the way cells treat information and make decisions? To what extend our current understanding enables us to reprogram and control the way cells behave? In this manuscript I describe several approaches developed for the computational analysis of the dynamics of biological networks. In particular I present work done on (i) the analysis of large gene networks with partial information on parameter values, (ii) the use of specification languages to express observations or desired properties in an abstract manner and efficiently search for parameters satisfying these properties, and (iii) recent efforts to use models to drive gene expression in real-time at the cellular level.

Loline alkaloids, found in many grass-Epichloë symbiota, are toxic or feeding deterrent to invertebrates. The loline alkaloids all share a saturated pyrrolizidine ring with a 1-amine group and an ether bridge linking C2 and C7. The steps in biosynthesis of loline alkaloids are catalyzed by enzymes encoded by a gene cluster, designated LOL, in the Epichloë genome. This dissertation addresses the enzymatic, genetic and evolutionary basis for diversification of these alkaloids, focusing on ether bridge formation and the subsequent modifications of the 1-amine to form different loline alkaloids. Through gene complementation of a natural lolO mutant and comparison of LOL clusters in strains with different loline alkaloid profiles, I found that lolO, predicted to encode a 2-oxoglutarate-dependent nonheme iron (2OG/Fe) dioxygenase, is required in formation of the ether bridge. Through application of isotopically labeled compound to Epichloë uncinata culture, I established that exo-1-acetamidopyrrolizidine (AcAP) and N-acetylnorloline (NANL) are true pathway intermediates. Application of AcAP to yeast expressing lolO resulted in production of NANL, establishing that LolO is sufficient to catalyze this unusual oxygenation reaction. After ether formation, modifications on the 1-amino group give loline, N-methylloline (NML), N-formylloline (NFL) and N-acetylloline (NAL). A double knockout of lolN, predicted to encode an acetamidase, and lolM, predicted to encode a methyltransferase, produced only NANL. Complementation of the double knockout with wild-type lolN and lolM restored the loline alkaloid profile. These results indicate that LolN is involved in deacetylating NANL to produce norloline, which is then modified to form the other lolines. Crude protein extract of a yeast transformant expressing LolM converted norloline to loline and NML, and loline to NML, supporting the hypothesis that LolM functions as a methyltransferase in the loline-alkaloid biosynthesis pathway. The alkaloid NAL was observed in some but not all plants symbiotic with Epichloë siegelii, and when provided with exogenous loline, asymbiotic meadow fescue (Lolium pratense) plants produced N-acetylloline (NAL), indicating that a plant acetyltransferase converts loline to NAL. I further analyzed the basis for loline alkaloid diversity by comparing the LOL clusters in the Epichloë and Atkinsonella species with different loline alkaloid profiles, and found that LOL clusters changed position, orientation and gene content over their evolutionary history. Frequent, independent losses of some or all late pathway genes, lolO, lolN, lolM and lolP, resulted in diverse loline alkaloid profiles. In addition, phylogenetic analyses demonstrated transspecies polymorphism of the LOL clusters. Based on my findings, I established that in Epichloë and Atkinsonella species the ether bridge is formed on acetamidopyrrolizidine. My study of the loline alkaloid profile of Adenocarpus decorticans (Fabaceae) suggests that these plants probably use a similar strategy at least with respect to ether-bridge formation. Further diversification steps of loline alkaloids in grass-Clavicipitaceae symbiota are carried out by enzymes of both Epichloë species and the host plant. Finally, I present evidence that LOL clusters have evolved by balancing selection for chemical diversity.