Data analytics and optimization theory are well-developed techniques to describe, predict and optimize real-world systems, and they have been widely used in engineering and science. This dissertation focuses on applications in biomedical systems, ranging from the scale of microbial communities to problems relating to human disease and health care. Starting from the microbial level, the first problem considered is to design metabolic division of labor in microbial communities. Given a number of microbial species living in a community, the starting point of the analysis is a list of all metabolic reactions present in the community, expressed in terms of the metabolite proportions involved in each reaction. Leveraging tools from Flux Balance Analysis (FBA), the problem is formulated as a Mixed Integer Program (MIP) and new methods are developed to solve large scale instances. The strategies found reveal a large space of nuanced and non-intuitive metabolic division of labor opportunities, including, for example, splitting the Tricarboxylic Acid Cycle (TCA) cycle into two separate halves. More broadly, the landscape of possible 1-, 2-, and 3-strain solutions is systematically mapped at increasingly tight constraints on the number of allowed reactions. The second problem addressed involves the prediction and prevention of short-term (30-day) hospital re-admissions. To develop predictive models, a variety of classification algorithms are adapted and coupled with robust (regularized) learning and heuristic feature selection approaches. Using real, large datasets, these methods are shown to reliably predict re-admissions of patients undergoing general surgery, within 30-days of discharge. Beyond predictions, a novel prescriptive method is developed that computes specific control actions with the effect of altering the outcome. This method, termed Prescriptive Support Vector Machines (PSVM), is based on an underlying SVM classifier. Applied to the hospital re-admission data, it is shown to reduce 30-day re-admissions after surgery through better control of the patient’s pre-operative condition. Specifically, using the new method the patient’s pre-operative hematocrit is regulated through limited blood transfusion. In the last problem in this dissertation, a framework for parameter estimation in Regularized Mixed Linear Regression (MLR) problems is developed. In the specific MLR setting considered, training data are generated from a mixture of distinct linear models (or clusters) and the task is to identify the corresponding coefficient vectors. The problem is formulated as a Mixed Integer Program (MIP) subject to regularization constraints on the coefficient vectors. A number of results on the convergence of parameter estimates for MLR are established. In addition, experimental prediction results are presented comparing the prediction algorithm with mean absolute error regression and random forest regression, in terms of both accuracy and interpretability.

Engineering microorganisms at the systems level is recognized to be the future of metabolic engineering. Thanks to the development of genome annotation, mcroorganisms can be understood, as never before, and be reconstructed in the form of computational models. Flux balance analysis provides a deep insight intocellular metabolism and can guide metabolic engineering strategies. In particular, algorithms can assess the cellular complexity of the metabolism and hint at genetic interventions to improve product productivity. In this work, Synechosystis PCC6803 metabolism was invesetigated in silico. Genetic interventions could besuggested to couple butanol synthesis to growth as a way to improve currentproductivities. Cofactor recycling and, in particular, buffering mechanisms were shown to be important targets. Creating a cofacor imbalance and removing thesebuffering mechanisms is an important driving force. This forces a carbon flux through butanol synthesis to maintain cofactor balance and sustain growth.

Microbes are ubiquitous in virtually all habitats on Earth and affect human life in multiple ways, from the health-balancing role of the human microbiome, to the involvement of microbial communities in the global nitrogen and carbon cycles. The capacity of microbes to survive and grow in diverse environments relates directly to their ability to utilize available resources, be they from other microbes or from the environment itself. Hence, understanding how the environment shapes the metabolic functionality of individual microbes and complex communities constitutes an important area of research. In the first part of my thesis work, I explored how environmental nutrient composition and intracellular transcriptional regulation data can be integrated to provide insight into the temporal metabolic behavior of a bacterium through the use of genome-scale stoichiometric modeling approaches (Flux Balance Analysis). Thus I developed the method of Temporal Expression-based Analysis of Metabolism (TEAM), and applied it to Shewanella oneidensis, a bacterium studied for its important bioenergy and bioremediation applications. I found that TEAM improves on previous models' predictions of S. oneidensis metabolic fluxes, and recovers the overflow metabolism that has been seen experimentally. This study demonstrated the value of incorporating environmental context and transcriptional data for the prediction of time-dependent metabolic behavior. In the second part of my work, I extended the exploration of microbial metabolism from single species to complex communities in order to understand the robustness of metabolic functions. Specifically, I implemented novel mathematical analyses of metagenomic sequencing data to ask how functional stability of microbial communities could ensue despite large taxonomic variability. Upon representing in matrix form the metabolic capabilities of all genera found in 202 available metabolic ecosystem datasets, I compared the different communities with each other and with various randomized analogues. My results reveal new connections between the abundance of an organism in the community and the functions that it encodes. Furthermore, I found that genus abundances govern the metabolic robustness of a community more than the distribution of genetically encoded functions among the community members, suggesting that communities rely largely on ecological interactions to regulate their overall functionality.

We are interested in numerical methods for the optimization constrained second order ordinary differential equations arising in biofilm modelling. This class of problems is challenging for several reasons. One of the reasons is that the underlying solution has a steep slope, making it difficult to resolve. We propose a new numerical method with techniques such as domain decomposition and asynchronous iterations for solving certain types of ordinary differential equations more efficiently. In fact, for our class of problems after applying the techniques of domain decomposition with overlap we are able to solve the ordinary differential equations with a steep slope on a larger domain than previously possible. After applying asynchronous iteration techniques, we are able to solve the problem with less time.~We provide theoretical conditions for the convergence of each of the techniques. The other reason is that the second order ordinary differential equations are coupled with an optimization problem, which can be viewed as the constraints. We propose a numerical method for solving the coupled problem and show that it converges under certain conditions. An application of the proposed methods on biofilm modeling is discussed. The numerical method proposed is adopted to solve the biofilm problem, and we are able to solve the problem with larger thickness of the biofilm than possible before as is shown in the numerical experiments. / Mathematics

Computational biology enabled us to manage vast amount of experimental data and make inferences on observations that we had not made. Among the many methods, predicting metabolic functions with genome-scale models had shown promising results in the recent years. Using sophisticated algorithms, such as flux balance analysis, OptKnock, and OptForce, we can predict flux distributions and design metabolic engineering strategies at a greater efficiency. The caveat of these current methods is the accuracy of the predictions. We proposed using flux balance analysis with flux ratios as a possible solution to improving the accuracy of the conventional methods. To examine the accuracy of our approach, we implemented flux balance analyses with flux ratios in five publicly available genome-scale models of five different organisms, including Arabidopsis thaliana, yeast, cyanobacteria, Escherichia coli, and Clostridium acetobutylicum, using published metabolic engineering strategies for improving product yields in these organisms. We examined the limitations of the published strategies, searched for possible improvements, and evaluated the impact of these strategies on growth and product yields. The flux balance analysis with flux ratio method requires a prior knowledge on the critical regions of the metabolic network where altering flux ratios can have significant impact on flux redistribution. Thus, we further developed the reverse flux balance analysis with flux ratio algorithm as a possible solution to automatically identify these critical regions and suggest metabolic engineering strategies. We examined the accuracy of this algorithm using an Arabidopsis genome-scale model and found consistency in the prediction with our experimental data. / Master of Science

Tesis por compendio / [ES] Saccharomyces cerevisiae, además de ser un importante organismo modelo en biología, es indiscutiblemente la especie de levadura más utilizada en procesos fermentativos industriales, incluyendo el sector enológico. Su capacidad de fermentar en concentraciones elevadas de azúcares, tolerar concentraciones altas de etanol y soportar la adición de sulfitos, son algunos de los factores que explican su éxito en fermentaciones vínicas. El metabolismo fermentativo de S. cerevisiae en condiciones enológicas se conoce bien gracias a una amplia bibliografía científica. En cambio, aún se sabe poco sobre el metabolismo de las especies de Saccharomyces criotolerantes, S. uvarum y S. kudriavzevii, quienes han suscitado recientemente el interés del sector vitivinícola por sus buenas propiedades fermentativas a bajas temperaturas, tales como la producción de vinos con mayor contenido en glicerol y alta complejidad aromática, llegando a veces a reducir su contenido en etanol. En este contexto, esta tesis pretende ampliar nuestros conocimientos sobre el metabolismo fermentativo de S. uvarum y S. kudriavzevii en condiciones enológicas, profundizando en el entendimiento de las diferencias existentes con el de S. cerevisiae, así como entre cepas de S. cerevisiae de distintos orígenes. Para ello, hemos utilizado varias técnicas ómicas para analizar la dinámica de los metabolomas (intra- y extracelulares) y/o transcriptomas de cepas representativas de S. cerevisiae, S. uvarum y S. kudriavzevii a alta (25 °C) y baja (12 °C) temperatura de fermentación. También, hemos desarrollado un modelo metabólico a escala de genoma que, junto a un análisis de balance de flujos, es capaz de cuantificar los flujos a través del metabolismo del carbono y del nitrógeno de levaduras en cultivo de tipo batch. Así, el conjunto de estos trabajos nos ha permitido identificar rasgos metabólicos y/o transcriptómicos relevantes para el sector enológico en estas especies. También se aporta nueva información sobre las especificidades de redistribución de flujos en la red metabólica de levaduras del género Saccharomyces acorde a la especie y las fluctuaciones ambientales que ocurren durante una fermentación vínica. / [CAT] Saccharomyces cerevisiae, a més de ser un important organisme model en biologia, és indiscutiblement l'espècie de llevat més utilitzat en processos fermentatius industrials, incloent el sector enològic. La seua capacitat de fermentar grans concentracions de sucres, tolerar concentracions altes d'etanol i suportar l'addició de sulfits, són alguns dels factors que expliquen el seu èxit en fermentacions víniques. D'aquesta manera, el metabolisme fermentatiu de S. cerevisiae en condicions enològiques està ben descrit i es beneficia d'una àmplia bibliografia científica. En canvi, poc se sap encara sobre el metabolisme de les espècies de Saccharomyces criotolerants, S. uvarum i S. kudriavzevii, els qui han recentment suscitat l'interés del sector vitivinícola per les seues bones propietats fermentatives a baixes temperatures, com ara la producció de vins amb major contingut en glicerol, alta complexitat aromàtica i arribant a vegades a reduir el seu contingut en etanol. En aquest context, aquesta tesi pretén ampliar els nostres coneixements sobre el metabolisme fermentatiu de S. uvarum i S. kudriavzevii en condicions enològiques, aprofundint en l'enteniment de les diferències existents amb el de S. cerevisiae, així també com entre ceps de S. cerevisiae de diferents orígens. Per a això, hem utilitzat diverses tècniques omiques per a analitzar la dinàmica dels metabolomes (intra- i extracelul·lars) i/o transcriptomes de ceps representatius de S. cerevisiae, S. uvarum i S. kudriavzevii a alta (25 °C) i baixa (12 °C) temperatures de fermentació. També, hem desenvolupat un model metabòlic a escala del genoma que, al costat d'una anàlisi de balanç de fluxos, és capaç de quantificar els fluxos a través del metabolisme carbonat i nitrogenat de llevats en cultius de tipus batch. Així, el conjunt d'aquests treballs ens ha permés identificar trets metabòlics i/o transcriptómics rellevants per al sector enològic en aquestes espècies. També aporta nova informació sobre les especificitats de redistribució de fluxos en la xarxa metabòlica de llevats del gènere Saccharomyces concorde a l'espècie i les fluctuacions ambientals ocorrent durant una fermentació vínica. / [EN] Saccharomyces cerevisiae, besides being an important model organism in biology, is undoubtedly the most widely used yeast species in industrial fermentation processes, including the winemaking sector. Its ability to ferment at high levels of sugars, tolerate high ethanol concentrations and withstand the addition of sulfites are some of the factors explaining its success in wine fermentation. Accordingly, the fermentative metabolism of S. cerevisiae under oenological conditions is well described and benefits from a large scientific literature. In contrast, little is known about the metabolism of the cryotolerant Saccharomyces species, S. uvarum and S. kudriavzevii, which have recently attracted the interest of the wine industry for their good fermentative properties at low temperatures, such as the production of wines with higher glycerol content, high aromatic complexity and sometimes even reduced ethanol content. In this context, this thesis aims to expand our knowledge on the fermentative metabolism of S. uvarum and S. kudriavzevii under oenological conditions, deepening our understanding of the existing differences with that of S. cerevisiae, as well as between S. cerevisiae strains of different origins. For this purpose, we have used several omics techniques to analyze the dynamics of the (intra- and extracellular) metabolomes and/or transcriptomes of representative strains of S. cerevisiae, S. uvarum and S. kudriavzevii at high (25 °C) and low (12 °C) fermentation temperatures. Also, we have developed a genome-scale metabolic model that, together with a flux balance analysis, is able to quantify fluxes through carbon and nitrogen metabolism of yeast in batch culture. Taken together, this work has allowed us to identify metabolic and/or transcriptomic traits relevant to the oenological sector in these species. It also provides new information on the specificities of flux redistribution in the metabolic network of Saccharomyces yeasts according to the species and environmental fluctuations occurring during wine fermentation. / The present work has been carried out at the Department of Food Biotechnology of the IATA (CSIC). Romain Minebois was funded by a FPI grant (REF: BES-2016-078202) and supported by projects AGL2015-67504-C3-1R and RTI2018-093744-BC31 of the Ministerio de Ciencia e Inovación awarded to Amparo Querol. / Minebois, RCM. (2021). Study of the differences in the fermentative metabolism of S. cerevisiae, S. uvarum and S. kudriavzevii species [Tesis doctoral]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/176018 / Compendio

Advances in bioinformatics and computational biology have enabled integration of an enormous amount of known biological interactions. This has enabled researchers to use models and data to design experiments and guide new discovery as well as test for consistency. One such computational method is constraint-based metabolic flux modeling. This is performed using genome-scale metabolic models (GEMs) that are a collection of biochemical reactions, derived from a genome's annotation. This type of flux modeling enables prediction of net metabolite conversion rates (metabolic fluxes) to help understand metabolic activities under specific environmental conditions. It can also be used to derive metabolic engineering strategies that involve genetic manipulations. Over the past decade, GEMs have been constructed for several different microbes, plants, and animal species. Researchers have also developed advanced algorithms to use GEMs to predict genetic modifications for the overproduction of biofuel and valuable commodity chemicals. Many of the predictive algorithms for microbes were validated with experimental results and some have been applied industrially. However, there is much room for improvement. For example, many algorithms lack straight-forward predictions that truly help non-computationally oriented researchers understand the predicted necessary metabolic modifications. Other algorithms are limited to simple genetic manipulations due to computational demands. Utilization of GEMs and flux-based modeling to predict in vivo characteristics of multicellular organisms has also proven to be challenging. Many researchers have created unique frameworks to use plant GEMs to hypothesize complex cellular interactions, such as metabolic adjustments in rice under variable light intensity and in developing tomato fruit. However, few quantitative predictions have been validated experimentally in plants. This research demonstrates the utility of GEMs and flux-based modeling in both metabolic engineering and analysis by tackling the challenges addressed previously with alternative approaches. Here, a novel predictive algorithm, Node-Reward Optimization (NR-Opt) toolbox, was developed. It delivers concise and accurate metabolic engineering designs (i.e. genetic modifications) that can truly improve the efficiency of strain development. As a proof-of-concept, the algorithm was deployed on GEMs of E. coli and Arabidopsis thaliana, and the predicted metabolic engineering strategies were compared with results of well-accepted algorithms and validated with published experimental data. To demonstrate the utility of GEMs and flux-based modeling in analyzing plant metabolism, specifically its response to changes in the signaling pathway, a novel modeling framework and analytical pipeline were developed to simulate changes of growth and starch metabolism in Arabidopsis over multiple stages of development. This novel framework was validated through simulation of growth and starch metabolism of Arabidopsis plants overexpressing sucrose non-fermenting related kinase 1.1 (SnRK1.1). Previous studies suggest that SnRK1.1 may play a critical signaling role in plant development and starch level (a critical carbon source for plant night growth). It has been shown that overexpressing of SnRK1.1 in Arabidopsis can delay vegetative-to-reproductive transition. Many studies on plant development have correlated the delay in developmental transition to reduction in starch turnover at night. To determine whether starch played a role in the delayed developmental transition in SnRK1.1 overexpressor plants, starch turnover was simulated at multiple developmental stages. Simulations predicted no reduction in starch turnover prior to developmental transition. Predicted results were experimentally validated, and the predictions were in close agreement with experimental data. This result further supports previous data that SnRK1.1 may regulate developmental transition in Arabidopsis. This study further validates the utility of GEMs and flux-based modeling in guiding future metabolic research. / Ph. D. / Recent advances in genetic and biochemical studies revealed the incredible complexity of cells, which generated interests in using computers to aid whole cell analyses and design cell engineering strategies to overproduce valuable commodity chemicals, such as biofuel, medicines, polymers, and many industrial materials. In order to use computers to study cells, current knowledge of cellular machinery is converted into mathematical models, such as genome-scale metabolic models. Genome-scale metabolic models are used to simulate the rates of chemical events in cells, which helps researchers predict cellular outputs of interest, such as growth rate and chemical synthesis rates. Combining genome-scale metabolic models with sophisticated computer algorithms, researchers can simulate numerous cell engineering experiments and select a few candidates to test physically, which can reduce cost and research time significantly. This computational technique has been well validated in microorganisms, such as E. coli and yeast; however, the ability to simulate cellular chemistry accurately in plants remains a challenge, which was a goal in my research. In addition, my research also aimed to reduce the inefficiencies in previous cell engineering design algorithms. I was able to develop a novel genome-scale model framework that enabled accurate simulation of plant growth and changes of starch content over time. I also developed a new computer algorithm that could significantly improve the efficiency in designing cell engineering strategies.

Tomatoes are the fourth most valuable commodity in agriculture after rice, wheat and soybeans globally with 151 million tonnes of fruit being produced in 2012. The tomato fruit is also a model system for fleshy fruit development. During ethylene-regulated fruit ripening there are complex changes in fruit chemical composition due to degradation and synthesis of a number of soluble and volatile metabolites. Ultimately, these changes control the composition of the ripe fruit and dictate its flavour and texture. It is known that ripening can proceed when mature green fruit are removed from the plant (and indeed this is standard commercial practice) but the extent to which metabolic changes are sustained when fruit are ripened in this way has yet to be established. A modelling approach such as constraints-based modelling can provide system-level insights into the workings of the complex tomato metabolic network during ripening. The first aim of this thesis was therefore to construct a genome-scale metabolic network model for tomato and to use this model to explore metabolic network flux distributions during the transitions between the stages of fruit ripening. The flux distributions predicted provided insight into the production and usage of energy and reductants, into routes for climacteric CO₂ release, and the metabolic routes underlying metabolite conversions during ripening. The second aim of this thesis was to use the model to explore metabolic engineering strategies for increased production of lycopene in tomato fruit. The model predictions showed that rearrangement of dominant metabolic fluxes were required to cope with the increased demand for reductants at high lycopene accumulation, which came at a cost of a lower accumulation of other secondary metabolites. Overall the thesis provides an approach to connect underlying metabolic mechanisms to the known metabolic processes that happen during ripening.

The decreasing costs of next generation sequencing technologies and the increasing speeds at which they work have lead to an abundance of 'omic datasets. The need for tools and methods to analyze, annotate, and model these datasets to better understand biological systems is growing. Here we present a novel software pipeline to reconstruct the metabolic model of an organism in silico starting from its genome sequence and a novel compilation of biological databases to better serve the generation of metabolic models. We validate these methods using five Gardnerella vaginalis strains and compare the gene annotation results to NCBI and the FBA results to Model SEED models. We found that our gene annotations were larger and highly similar in terms of function and gene types to the gene annotations downloaded from NCBI. Further, we found that our FBA models required a minimal addition of transport reactions, sources, and escapes indicating that our draft pathway models were very complete. We also found that on average our solutions contained more reactions than the models obtained from Model SEED due to a large amount of baseline reactions and gene products found in ASGARD.

In the first part of this research, we constructed a Genome scale Metabolic Model (GEM) of Taxus cuspidata, a medicinal plant used to produce paclitaxel (Taxol®). The construction of the T. cuspidata GEM was predicated on recent acquisition of a transcriptome of T. cuspidata metabolism under methyl jasmonate (MJ) elicited conditions (when paclitaxel is produced) and unelicited conditions (when paclitaxel is not produced). Construction of the draft model, in which transcriptomic data from elicited and unelicited conditions were included, utilized tools including the ModelSEED developed by Argonne National Laboratory. Although a model was successfully created and gapfilled by ModelSEED using their software, we were not able to reproduce their results using COBRA, a widely accepted FBA software package. Further work needs to be done to figure out how to run ModelSEED models on commonly available software. In the second part of this research, we modeled the MJ elicited/defense response phenotype in Arabidopsis thaliana. Previously published models of A. thaliana were tested for suitability in modeling the MJ elicited phenotype using publicly available computation tools. MJ elicited and unelicited datasets were compared to ascertain differences in metabolism between these two phenotypes. The MJ elicited and unelicited datasets were significantly different in many respects, including the expression levels of many genes associated with secondary metabolism. However, it was found that the expression of genes related to growth and central metabolism were not generally significantly different for the MJ+ and MJ- datasets, the pathways associated with secondary metabolism were incomplete and could not be modeled, and FBA methods did not show the difference in growth that was expected. These results suggest that behavior associated with the MJ+ phenotype such as slow growth and secondary metabolite production may be controlled by factors not easily modeled with transcriptome data alone. Additional research was performed in the area of cryosectioning and immunostaining of fixed Taxus aggregates. Protocols developed for this work can be found in Appendix B.