BACKGROUND:Shewanella is a genus of facultatively anaerobic, Gram-negative bacteria that have highly adaptable metabolism which allows them to thrive in diverse environments. This quality makes them an attractive bacterial target for research in bioremediation and microbial fuel cell applications. Constraint-based modeling is a useful tool for helping researchers gain insights into the metabolic capabilities of these bacteria. However, Shewanella oneidensis MR-1 is the only strain with a genome-scale metabolic model constructed out of 21 sequenced Shewanella strains.RESULTS:In this work, we updated the model for Shewanella oneidensis MR-1 and constructed metabolic models for three other strains, namely Shewanella sp. MR-4, Shewanella sp. W3-18-1, and Shewanella denitrificans OS217 which span the genus based on the number of genes lost in comparison to MR-1. We also constructed a Shewanella core model that contains the genes shared by all 21 sequenced strains and a few non-conserved genes associated with essential reactions. Model comparisons between the five constructed models were done at two levels - for wildtype strains under different growth conditions and for knockout mutants under the same growth condition. In the first level, growth/no-growth phenotypes were predicted by the models on various carbon sources and electron acceptors. Cluster analysis of these results revealed that the MR-1 model is most similar to the W3-18-1 model, followed by the MR-4 and OS217 models when considering predicted growth phenotypes. However, a cluster analysis done based on metabolic gene content revealed that the MR-4 and W3-18-1 models are the most similar, with the MR-1 and OS217 models being more distinct from these latter two strains. As a second level of comparison, we identified differences in reaction and gene content which give rise to different functional predictions of single and double gene knockout mutants using Comparison of Networks by Gene Alignment (CONGA). Here, we showed how CONGA can be used to find biomass, metabolic, and genetic differences between models.CONCLUSIONS:We developed four strain-specific models and a general core model that can be used to do various in silico studies of Shewanella metabolism. The developed models provide a platform for a systematic investigation of Shewanella metabolism to aid researchers using Shewanella in various biotechnology applications.

No description available.

Macroecological patterns, such as the highly uneven distribution of individuals among species and the monotonic increase of species richness with area, exist across ecological systems despite major differences in the biology of different species and locations. These patterns capture the general structure of ecological communities, and allow relatively accurate predictions to be made with limited information for under-studied systems. This is particularly important given ongoing climate change and loss of biodiversity. Understanding the mechanisms behind these patterns has both scientific and practical merits. I explore two conceptually different approaches that have been proposed as explanations for ecological patterns – the process-based approaches, which directly model key ecological processes such as birth, death, competition, and dispersal; and the constraint-based approaches, which view the patterns as the most likely state when the system is constrained in certain ways (e.g., the system has a fixed number of 100 individuals among five species, but the distribution may vary). While the process-based approaches directly link patterns to processes, the constraint-based approaches do not rely on the operation of specific processes and thus can be more broadly applied. I develop a new constraint-based approach to one of the most well established patterns in ecology, the power-law relationship between the mean and variance of a population. This pattern has been widely observed and adopted as characterization of population stability. I find that the shape of the pattern can be well explained with two numerical constraints on the system, lending support to the idea that some macroecological patterns may not arise from specific processes but be statistical in nature instead. I further examine the performance of the process- and constraint-based approaches for patterns of biodiversity and energy use, which are among the most essential as well as most well-studied aspects of community structure. Candidate models from both categories are able to partially capture the patterns across 60 globally distributed forest communities, however the process-based model is shown to provide a better general characterization of community structure than the constraint-base model in all communities. Thus the constraint-based approaches in their current forms do not fully encapsulate the effect of processes, which also contribute to the shape of the macroecological patterns of biodiversity and body size in addition to the constraints.

Constraint-based models of metabolism seldom incorporate capacity constraints on intracellular fluxes due to the lack of experimental data. This can sometimes lead to inaccurate growth phenotype predictions. Meanwhile, other forms of data such as fitness profiling data from growth competition experiments have been demonstrated to contain valuable information for elucidating key aspects of the underlying metabolic network. Hence, the optimal capacity constraint identification (OCCI) algorithm is developed to reconcile constraint-based models of metabolism with fitness profiling data by identifying a set of flux capacity constraints that optimally fits a wide array of strains. OCCI is able to identify capacity constraints with considerable accuracy by matching 1,155 in silico-generated growth rates using a simplified model of Escherichia coli central carbon metabolism. Capacity constraints identified using experimental fitness profiles with OCCI generated novel hypotheses, while integrating thermodynamics-based metabolic flux analysis allowed prediction of metabolite concentrations.

Constraint-based models of metabolism seldom incorporate capacity constraints on intracellular fluxes due to the lack of experimental data. This can sometimes lead to inaccurate growth phenotype predictions. Meanwhile, other forms of data such as fitness profiling data from growth competition experiments have been demonstrated to contain valuable information for elucidating key aspects of the underlying metabolic network. Hence, the optimal capacity constraint identification (OCCI) algorithm is developed to reconcile constraint-based models of metabolism with fitness profiling data by identifying a set of flux capacity constraints that optimally fits a wide array of strains. OCCI is able to identify capacity constraints with considerable accuracy by matching 1,155 in silico-generated growth rates using a simplified model of Escherichia coli central carbon metabolism. Capacity constraints identified using experimental fitness profiles with OCCI generated novel hypotheses, while integrating thermodynamics-based metabolic flux analysis allowed prediction of metabolite concentrations.

Empirical studies have shown that Intelligent Tutoring Systems (ITS) are effective tools for education. However, developing an ITS is a labour-intensive and time-consuming process. A major share of the development effort is devoted to acquiring the domain knowledge that accounts for the intelligence of the system. The goal of this research is to reduce the knowledge acquisition bottleneck and enable domain experts to build the domain model required for an ITS. In pursuit of this goal an authoring system capable of producing a domain model with the assistance of a domain expert was developed. Unlike previous authoring systems, this system (named CAS) has the ability to acquire knowledge for non-procedural as well as procedural tasks. CAS was developed to generate the knowledge required for constraint-based tutoring systems, reducing the effort as well as the amount of expertise in knowledge engineering and programming required. Constraint-based modelling is a student modelling technique that assists in somewhat easing the knowledge acquisition bottleneck due to the abstract representation. CAS expects the domain expert to provide an ontology of the domain, example problems and their solutions. It uses machine learning techniques to reason with the information provided by the domain expert for generating a domain model. A series of evaluation studies of this research produced promising results. The initial evaluation revealed that the task of composing an ontology of the domain assisted with the manual composition of a domain model. The second study showed that CAS was effective in generating constraints for the three vastly different domains of database modelling, data normalisation and fraction addition. The final study demonstrated that CAS was also effective in generating constraints when assisted by novice ITS authors, producing constraint sets that were over 90% complete.

A Bayesian network is graphical representation of the probabilistic relationships among set of variables and can be used to encode expert knowledge about uncertain domains. The structure of this model represents the set of conditional independencies among the variables in the data. Bayesian networks are widely applicable, having been used to model domains ranging from monitoring patients in an emergency room to predicting the severity of hailstorms. In this thesis, I focus on the problem of learning the structure of Bayesian networks from data. Under certain assumptions, the learned structure of a Bayesian network can represent causal relationships in the data. Constraint-based algorithms for structure learning are designed to accurately identify the structure of the distribution underlying the data and, therefore, the causal relationships. These algorithms use a series of conditional hypothesis tests to learn independence constraints on the structure of the model. When sample size is limited, these hypothesis tests are prone to errors. I present a comprehensive empirical evaluation of constraint-based algorithms and show that existing constraint-based algorithms are prone to many false negative errors in the constraints due to run- ning hypothesis tests with low statistical power. Furthermore, this analysis shows that many statistical solutions fail to reduce the overall errors of constraint-based algorithms. I show that new algorithms inspired by constraint satisfaction are able to produce significant improvements in structural accuracy. These constraint satisfaction algo- rithms exploit the interaction among the constraints to reduce error. First, I introduce an algorithm based on constraint optimization that is sound in the sample limit, like existing algorithms, but is guaranteed to produce a DAG. This new algorithm learns models with structural accuracy equivalent or better to existing algorithms. Second, I introduce an algorithm based constraint relaxation. Constraint relaxation combines different statistical techniques to identify constraints that are likely to be incorrect, and remove those constraints from consideration. I show that an algorithm combining constraint relaxation with constraint optimization produces Bayesian networks with significantly better structural accuracy when compared to existing structure learning algorithms, demonstrating the effectiveness of constraint satisfaction approaches for learning accurate structure of Bayesian networks.

ITC/USA 2008 Conference Proceedings / The Forty-Fourth Annual International Telemetering Conference and Technical Exhibition / October 27-30, 2008 / Town and Country Resort & Convention Center, San Diego, California / Predictive maintenance is the combination of inspection and data analysis to perform maintenance when the need is indicated by unit performance. Significant cost savings are possible while preserving a high level of system performance and readiness. Identifying predictors of maintenance conditions requires expert knowledge and the ability to process large data sets. This paper describes a novel use of constraint-based data-mining to model exceedence conditions. The approach extends the extract, transformation, and load process with domain aggregate approximation to encode expert knowledge. A data-mining workbench enables an expert to pose hypotheses that constrain a multivariate data-mining process.

Metabolic engineering of microorganisms for chemical production involves the coordination of regulatory, kinetic, and thermodynamic parameters within the context of the entire network, as well as the careful allocation of energetic and structural resources such as ATP, redox potential, and amino acids. The exponential progression of “omics” technologies over the past few decades has transformed our ability to understand these network interactions by generating enormous amounts of data about cell behavior. The great challenge of the new biological era is in processing, integrating, and rationally interpreting all of this information, leading to testable hypotheses. In silico metabolic reconstructions are versatile computational tools for integrating multiple levels of bioinformatics data, facilitating interpretation of that data, and making functional predictions related to the metabolic behavior of the cell. To explore the use of this modeling paradigm as a tool for enabling metabolic engineering in a poorly understood microorganism, an in silico constraint-based metabolic reconstruction for the anaerobic, cellulolytic bacterium Clostridium thermocellum was constructed based on available genome annotations, published phenotypic information, and specific biochemical assays. This dissertation describes the analysis and experimental validation of this model, the integration of transcriptomic data from an RNAseq experiment, and the use of the resulting model for generating novel strain designs for significantly improved production of ethanol from cellulosic biomass. The genome-scale metabolic reconstruction is shown to be a powerful framework for understanding and predicting various metabolic phenotypes, and contributions described here enhance the utility of these models for interpretation of experimental datasets for successful metabolic engineering.

The goal of this thesis is to develop a constraint based viscoplastic fluid model suitable for time-efficient dynamics simulation in 3D of granular matter. The model should be applicable to both the static and dense flow regime and at large pressures. The thesis is performed for UMIT Research Lab at Umeå University. It is a part of the research at UMIT connected to LKAB and Volvo CE and its applications can be in simulating industrial processes or training simulators. My work is based on previous work done by Claude Lacoursière, Martin Servin and Kenneth Bodin. They have created a constraint fluid model based on {\sph} and Claude's PhD. thesis. This model is extended with additional constraints to handle shear forces, which is necessary to model granular material. Some test cases are specified and compared visually to each other and to the results of other work. The model seems to work visually but more analysis and larger systems are needed to be certain. The model should scale well and is well suited for parallellization.