KETCHUP: Parameterizing of large-scale kinetic models using multiple datasets with different reference states.

Large-scale kinetic models provide the computational means to dynamically link metabolic reaction fluxes to metabolite concentrations and enzyme levels while also conforming to substrate level regulation. However, the development of broadly applicable frameworks for efficiently and robustly parameterizing models remains a challenge. Challenges arise due to both the heterogeneity, paucity, and difficulty in obtaining flux and/or concentration data but also due to the computational difficulties of the underlying parameter identification problem. Both the computational demands for parameterization, degeneracy of obtained parameter solutions and interpretability of results has so far limited widespread adoption of large-scale kinetic models despite their potential. Herein, we introduce the Kinetic Estimation Tool Capturing Heterogeneous Datasets Using Pyomo (KETCHUP), a flexible parameter estimation tool that leverages a primal-dual interior-point algorithm to solve a nonlinear programming (NLP) problem that identifies a set of parameters capable of recapitulating the (non)steady-state fluxes and concentrations in wild-type and perturbed metabolic networks. KETCHUP is benchmarked against previously parameterized large-scale kinetic models demonstrating an at least an order of magnitude faster convergence than the tool K-FIT while at the same time attaining better data fits. This versatile toolbox accepts different kinetic descriptions, metabolic fluxes, enzyme levels and metabolite concentrations, under either steady-state or instationary conditions to enable robust kinetic model construction and parameterization. KETCHUP supports the SBML format and can be accessed at https://github.com/maranasgroup/KETCHUP.

Comparative study of two Saccharomyces cerevisiae strains with kinetic models at genome-scale.

The parameterization of kinetic models requires measurement of fluxes and/or metabolite levels for a base strain and a few genetic perturbations thereof. Unlike stoichiometric models that are mostly invariant to the specific strain, it remains unclear whether kinetic models constructed for different strains of the same species have similar or significantly different kinetic parameters. This important question underpins the applicability range and prediction limits of kinetic reconstructions. To this end, herein we parameterize two separate large-scale kinetic models using K-FIT with genome-wide coverage corresponding to two distinct strains of Saccharomyces cerevisiae: CEN.PK 113-7D strain (model k-sacce306-CENPK), and growth-deficient BY4741 (isogenic to S288c; model k-sacce306-BY4741). The metabolic network for each model contains 306 reactions, 230 metabolites, and 119 substrate-level regulatory interactions. The two models (for CEN.PK and BY4741) recapitulate, within one standard deviation, 77% and 75% of the fitted dataset fluxes, respectively, determined by 13C metabolic flux analysis for wild-type and eight single-gene knockout mutants of each strain. Strain-specific kinetic parameterization results indicate that key enzymes in the TCA cycle, glycolysis, and arginine and proline metabolism drive the metabolic differences between these two strains of S. cerevisiae. Our results suggest that although kinetic models cannot be readily used across strains as stoichiometric models, they can capture species-specific information through the kinetic parameterization process.

Recent advances in constraint and machine learning-based metabolic modeling by leveraging stoichiometric balances, thermodynamic feasibility and kinetic law formalisms.

Understanding the governing principles behind organisms' metabolism and growth underpins their effective deployment as bioproduction chassis. A central objective of metabolic modeling is predicting how metabolism and growth are affected by both external environmental factors and internal genotypic perturbations. The fundamental concepts of reaction stoichiometry, thermodynamics, and mass action kinetics have emerged as the foundational principles of many modeling frameworks designed to describe how and why organisms allocate resources towards both growth and bioproduction. This review focuses on the latest algorithmic advancements that have integrated these foundational principles into increasingly sophisticated quantitative frameworks.

Orchestrating hi-fi annotations.

By incorporating sequence homology and context associations, global probabilistic approaches to annotate genome-scale metabolic networks can substantially improve the accuracy of biochemical predictions, revealing potential functionality and directing experimental validation.

Metabolic engineering of low-pH-tolerant non-model yeast, Issatchenkia orientalis, for production of citramalate.

Methyl methacrylate (MMA) is an important petrochemical with many applications. However, its manufacture has a large environmental footprint. Combined biological and chemical synthesis (semisynthesis) may be a promising alternative to reduce both cost and environmental impact, but strains that can produce the MMA precursor (citramalate) at low pH are required. A non-conventional yeast, Issatchenkia orientalis, may prove ideal, as it can survive extremely low pH. Here, we demonstrate the engineering of I. orientalis for citramalate production. Using sequence similarity network analysis and subsequent DNA synthesis, we selected a more active citramalate synthase gene (cimA) variant for expression in I. orientalis. We then adapted a piggyBac transposon system for I. orientalis that allowed us to simultaneously explore the effects of different cimA gene copy numbers and integration locations. A batch fermentation showed the genome-integrated-cimA strains produced 2.0 g/L citramalate in 48 h and a yield of up to 7% mol citramalate/mol consumed glucose. These results demonstrate the potential of I. orientalis as a chassis for citramalate production.

MetRxn: a knowledgebase of metabolites and reactions spanning metabolic models and databases.

BACKGROUND: Increasingly, metabolite and reaction information is organized in the form of genome-scale metabolic reconstructions that describe the reaction stoichiometry, directionality, and gene to protein to reaction associations. A key bottleneck in the pace of reconstruction of new, high-quality metabolic models is the inability to directly make use of metabolite/reaction information from biological databases or other models due to incompatibilities in content representation (i.e., metabolites with multiple names across databases and models), stoichiometric errors such as elemental or charge imbalances, and incomplete atomistic detail (e.g., use of generic R-group or non-explicit specification of stereo-specificity). DESCRIPTION: MetRxn is a knowledgebase that includes standardized metabolite and reaction descriptions by integrating information from BRENDA, KEGG, MetaCyc, Reactome.org and 44 metabolic models into a single unified data set. All metabolite entries have matched synonyms, resolved protonation states, and are linked to unique structures. All reaction entries are elementally and charge balanced. This is accomplished through the use of a workflow of lexicographic, phonetic, and structural comparison algorithms. MetRxn allows for the download of standardized versions of existing genome-scale metabolic models and the use of metabolic information for the rapid reconstruction of new ones. CONCLUSIONS: The standardization in description allows for the direct comparison of the metabolite and reaction content between metabolic models and databases and the exhaustive prospecting of pathways for biotechnological production. This ever-growing dataset currently consists of over 76,000 metabolites participating in more than 72,000 reactions (including unresolved entries). MetRxn is hosted on a web-based platform that uses relational database models (MySQL).

Mathematical optimization applications in metabolic networks.

Genome-scale metabolic models are increasingly becoming available for a variety of microorganisms. This has spurred the development of a wide array of computational tools, and in particular, mathematical optimization approaches, to assist in fundamental metabolic network analyses and redesign efforts. This review highlights a number of optimization-based frameworks developed towards addressing challenges in the analysis and engineering of metabolic networks. In particular, three major types of studies are covered here including exploring model predictions, correction and improvement of models of metabolism, and redesign of metabolic networks for the targeted overproduction of a desired compound. Overall, the methods reviewed in this paper highlight the diversity of queries, breadth of questions and complexity of redesigns that are amenable to mathematical optimization strategies.

Examining organic acid production potential and growth-coupled strategies in Issatchenkia orientalis using constraint-based modeling.

Growth-coupling product formation can facilitate strain stability by aligning industrial objectives with biological fitness. Organic acids make up many building block chemicals that can be produced from sugars obtainable from renewable biomass. Issatchenkia orientalis is a yeast strain tolerant to acidic conditions and is thus a promising host for industrial production of organic acids. Here, we use constraint-based methods to assess the potential of computationally designing growth-coupled production strains for I. orientalis that produce 22 different organic acids under aerobic or microaerobic conditions. We explore native and engineered pathways using glucose or xylose as the carbon substrates as proxy constituents of hydrolyzed biomass. We identified growth-coupled production strategies for 37 of the substrate-product pairs, with 15 pairs achieving production for any growth rate. We systematically assess the strain design solutions and categorize the underlying principles involved.

Construction of an E. Coli genome-scale atom mapping model for MFA calculations.

Metabolic flux analysis (MFA) has so far been restricted to lumped networks lacking many important pathways, partly due to the difficulty in automatically generating isotope mapping matrices for genome-scale metabolic networks. Here we introduce a procedure that uses a compound matching algorithm based on the graph theoretical concept of pattern recognition along with relevant reaction information to automatically generate genome-scale atom mappings which trace the path of atoms from reactants to products for every reaction. The procedure is applied to the iAF1260 metabolic reconstruction of Escherichia coli yielding the genome-scale isotope mapping model imPR90068. This model maps 90,068 non-hydrogen atoms that span all 2,077 reactions present in iAF1260 (previous largest mapping model included 238 reactions). The expanded scope of the isotope mapping model allows the complete tracking of labeled atoms through pathways such as cofactor and prosthetic group biosynthesis and histidine metabolism. An EMU representation of imPR90068 is also constructed and made available.

OptForce: an optimization procedure for identifying all genetic manipulations leading to targeted overproductions.

Computational procedures for predicting metabolic interventions leading to the overproduction of biochemicals in microbial strains are widely in use. However, these methods rely on surrogate biological objectives (e.g., maximize growth rate or minimize metabolic adjustments) and do not make use of flux measurements often available for the wild-type strain. In this work, we introduce the OptForce procedure that identifies all possible engineering interventions by classifying reactions in the metabolic model depending upon whether their flux values must increase, decrease or become equal to zero to meet a pre-specified overproduction target. We hierarchically apply this classification rule for pairs, triples, quadruples, etc. of reactions. This leads to the identification of a sufficient and non-redundant set of fluxes that must change (i.e., MUST set) to meet a pre-specified overproduction target. Starting with this set we subsequently extract a minimal set of fluxes that must actively be forced through genetic manipulations (i.e., FORCE set) to ensure that all fluxes in the network are consistent with the overproduction objective. We demonstrate our OptForce framework for succinate production in Escherichia coli using the most recent in silico E. coli model, iAF1260. The method not only recapitulates existing engineering strategies but also reveals non-intuitive ones that boost succinate production by performing coordinated changes on pathways distant from the last steps of succinate synthesis.

Building kinetic models for metabolic engineering.

Kinetic formalisms of metabolism link metabolic fluxes to enzyme levels, metabolite concentrations and their allosteric regulatory interactions. Though they require the identification of physiologically relevant values for numerous parameters, kinetic formalisms uniquely establish a mechanistic link across heterogeneous omics datasets and provide an overarching vantage point to effectively inform metabolic engineering strategies. Advances in computational power, gene annotation coverage, and formalism standardization have led to significant progress over the past few years. However, careful interpretation of model predictions, limited metabolic flux datasets, and assessment of parameter sensitivity remain as challenges. In this review we highlight fundamental considerations which influence model quality and prediction, advances in methodologies, and success stories of deploying kinetic models to guide metabolic engineering.

Genome scale reconstruction of a Salmonella metabolic model: comparison of similarity and differences with a commensal Escherichia coli strain.

Salmonella are closely related to commensal Escherichia coli but have gained virulence factors enabling them to behave as enteric pathogens. Less well studied are the similarities and differences that exist between the metabolic properties of these organisms that may contribute toward niche adaptation of Salmonella pathogens. To address this, we have constructed a genome scale Salmonella metabolic model (iMA945). The model comprises 945 open reading frames or genes, 1964 reactions, and 1036 metabolites. There was significant overlap with genes present in E. coli MG1655 model iAF1260. In silico growth predictions were simulated using the model on different carbon, nitrogen, phosphorous, and sulfur sources. These were compared with substrate utilization data gathered from high throughput phenotyping microarrays revealing good agreement. Of the compounds tested, the majority were utilizable by both Salmonella and E. coli. Nevertheless a number of differences were identified both between Salmonella and E. coli and also within the Salmonella strains included. These differences provide valuable insight into differences between a commensal and a closely related pathogen and within different pathogenic strains opening new avenues for future explorations.

Improved computational performance of MFA using elementary metabolite units and flux coupling.

Extending the scope of isotope mapping models becomes increasingly important in order to analyze strains and drive improved product yields as more complex pathways are engineered into strains and as secondary metabolites are used as starting points for new products. Here we present how the elementary metabolite unit (EMU) framework and flux coupling significantly decrease the computational burden of metabolic flux analysis (MFA) when applied to large-scale metabolic models. We applied these techniques to a previously published isotope mapping model of Escherichia coli accounting for 238 reactions. We find that the combined use of EMU and flux coupling analysis leads to a ten-fold decrease in the number of variables in comparison to the original isotope distribution vector (IDV) version of the model. In addition, using OptMeas the task of identifying additional measurement choices to fully specify the flows in the metabolic network required only 2% of the computation time of the one using IDVs. The observed computational savings reveal the rapid progress in performing MFA with increasingly larger isotope models with the ultimate goal of handling genome-scale models of metabolism.

Genome-scale gene/reaction essentiality and synthetic lethality analysis.

Synthetic lethals are to pairs of non-essential genes whose simultaneous deletion prohibits growth. One can extend the concept of synthetic lethality by considering gene groups of increasing size where only the simultaneous elimination of all genes is lethal, whereas individual gene deletions are not. We developed optimization-based procedures for the exhaustive and targeted enumeration of multi-gene (and by extension multi-reaction) lethals for genome-scale metabolic models. Specifically, these approaches are applied to iAF1260, the latest model of Escherichia coli, leading to the complete identification of all double and triple gene and reaction synthetic lethals as well as the targeted identification of quadruples and some higher-order ones. Graph representations of these synthetic lethals reveal a variety of motifs ranging from hub-like to highly connected subgraphs providing a birds-eye view of the avenues available for redirecting metabolism and uncovering complex patterns of gene utilization and interdependence. The procedure also enables the use of falsely predicted synthetic lethals for metabolic model curation. By analyzing the functional classifications of the genes involved in synthetic lethals, we reveal surprising connections within and across clusters of orthologous group functional classifications.

A genome-scale metabolic reconstruction of Mycoplasma genitalium, iPS189.

With a genome size of approximately 580 kb and approximately 480 protein coding regions, Mycoplasma genitalium is one of the smallest known self-replicating organisms and, additionally, has extremely fastidious nutrient requirements. The reduced genomic content of M. genitalium has led researchers to suggest that the molecular assembly contained in this organism may be a close approximation to the minimal set of genes required for bacterial growth. Here, we introduce a systematic approach for the construction and curation of a genome-scale in silico metabolic model for M. genitalium. Key challenges included estimation of biomass composition, handling of enzymes with broad specificities, and the lack of a defined medium. Computational tools were subsequently employed to identify and resolve connectivity gaps in the model as well as growth prediction inconsistencies with gene essentiality experimental data. The curated model, M. genitalium iPS189 (262 reactions, 274 metabolites), is 87% accurate in recapitulating in vivo gene essentiality results for M. genitalium. Approaches and tools described herein provide a roadmap for the automated construction of in silico metabolic models of other organisms.

Genome-scale metabolic reconstruction of the non-model yeast Issatchenkia orientalis SD108 and its application to organic acids production.

Many platform chemicals can be produced from renewable biomass by microorganisms, with organic acids making up a large fraction. Intolerance to the resulting low pH growth conditions, however, remains a challenge for the industrial production of organic acids by microorganisms. Issatchenkia orientalis SD108 is a promising host for industrial production because it is tolerant to acidic conditions as low as pH 2.0. With the goal to systematically assess the metabolic capabilities of this non-model yeast, we developed a genome-scale metabolic model for I. orientalis SD108 spanning 850 genes, 1826 reactions, and 1702 metabolites. In order to improve the model's quantitative predictions, organism-specific macromolecular composition and ATP maintenance requirements were determined experimentally and implemented. We examined its network topology, including essential genes and flux coupling analysis and drew comparisons with the Yeast 8.3 model for Saccharomyces cerevisiae. We explored the carbon substrate utilization and examined the organism's production potential for the industrially-relevant succinic acid, making use of the OptKnock framework to identify gene knockouts which couple production of the targeted chemical to biomass production. The genome-scale metabolic model iIsor850 is a data-supported curated model which can inform genetic interventions for overproduction.

A comprehensive genome-scale model for Rhodosporidium toruloides IFO0880 accounting for functional genomics and phenotypic data.

Rhodosporidium toruloides is a red, basidiomycetes yeast that can accumulate a large amount of lipids and produce carotenoids. To better assess this non-model yeast's metabolic capabilities, we reconstructed a genome-scale model of R. toruloides IFO0880's metabolic network (iRhto1108) accounting for 2204 reactions, 1985 metabolites and 1108 genes. In this work, we integrated and supplemented the current knowledge with in-house generated biomass composition and experimental measurements pertaining to the organism's metabolic capabilities. Predictions of genotype-phenotype relations were improved through manual curation of gene-protein-reaction rules for 543 reactions leading to correct recapitulations of 84.5% of gene essentiality data (sensitivity of 94.3% and specificity of 53.8%). Organism-specific macromolecular composition and ATP maintenance requirements were experimentally measured for two separate growth conditions: (i) carbon and (ii) nitrogen limitations. Overall, iRhto1108 reproduced R. toruloides's utilization capabilities for 18 alternate substrates, matched measured wild-type growth yield, and recapitulated the viability of 772 out of 819 deletion mutants. As a demonstration to the model's fidelity in guiding engineering interventions, the OptForce procedure was applied on iRhto1108 for triacylglycerol overproduction. Suggested interventions recapitulated many of the previous successful implementations of genetic modifications and put forth a few new ones.

Identification of optimal measurement sets for complete flux elucidation in metabolic flux analysis experiments.

Metabolic flux analysis (MFA) methods use external flux and isotopic measurements to quantify the magnitude of metabolic flows in metabolic networks. A key question in this analysis is choosing a set of measurements that is capable of yielding a unique flux distribution (identifiability). In this article, we introduce an optimization-based framework that uses incidence structure analysis to determine the smallest (or most cost-effective) set of measurements leading to complete flux elucidation. This approach relies on an integer linear programming formulation OptMeas that allows for the measurement of external fluxes and the complete (or partial) enumeration of the isotope forms of metabolites without requiring any of these to be chosen in advance. We subsequently query and refine the measurement sets suggested by OptMeas for identifiability and optimality. OptMeas is first tested on small to medium-size demonstration examples. It is subsequently applied to a large-scale E. coli isotopomer mapping model with more than 17,000 isotopomers. A number of additional measurements are identified leading to maximum flux elucidation in an amorphadiene producing E. coli strain.

Zea mays iRS1563: a comprehensive genome-scale metabolic reconstruction of maize metabolism.

The scope and breadth of genome-scale metabolic reconstructions have continued to expand over the last decade. Herein, we introduce a genome-scale model for a plant with direct applications to food and bioenergy production (i.e., maize). Maize annotation is still underway, which introduces significant challenges in the association of metabolic functions to genes. The developed model is designed to meet rigorous standards on gene-protein-reaction (GPR) associations, elementally and charged balanced reactions and a biomass reaction abstracting the relative contribution of all biomass constituents. The metabolic network contains 1,563 genes and 1,825 metabolites involved in 1,985 reactions from primary and secondary maize metabolism. For approximately 42% of the reactions direct literature evidence for the participation of the reaction in maize was found. As many as 445 reactions and 369 metabolites are unique to the maize model compared to the AraGEM model for A. thaliana. 674 metabolites and 893 reactions are present in Zea mays iRS1563 that are not accounted for in maize C4GEM. All reactions are elementally and charged balanced and localized into six different compartments (i.e., cytoplasm, mitochondrion, plastid, peroxisome, vacuole and extracellular). GPR associations are also established based on the functional annotation information and homology prediction accounting for monofunctional, multifunctional and multimeric proteins, isozymes and protein complexes. We describe results from performing flux balance analysis under different physiological conditions, (i.e., photosynthesis, photorespiration and respiration) of a C4 plant and also explore model predictions against experimental observations for two naturally occurring mutants (i.e., bm1 and bm3). The developed model corresponds to the largest and more complete to-date effort at cataloguing metabolism for a plant species.

Rapid responses of ribosomal RNA synthesis to nutrient shifts.

A major challenge in systems biology is to integrate our mechanistic understanding of gene regulation to predict quantitatively how cells will respond to environmental changes. Living cells respond rapidly to the availability of nutrients in part by altering production of ribosomal RNA (rRNA), a limiting component in the biosynthesis of ribosomes. Studies of rRNA transcription by the RNA polymerase of Escherichia coli have identified regulatory roles for guanosine tetraphosphate (ppGpp), the initiating nucleotide, and the protein DksA. To what extent findings from in vitro studies can be used to quantitatively predict in vivo responses to changing nutrient environments is unknown. We developed a mechanistic mathematical model for rRNA transcriptional responses to such changes. Our model accounts for binding of RNAP to its rRNA promoter to form a closed complex, isomerization from a closed complex to an open complex, reversible incorporation of the initiating NTP (iNTP), transcript elongation, and clearance of the promoter. Further, the model incorporates interactions between ppGpp and DksA with transcription intermediates, and it includes an empirical correction to account for salt effects. The model biophysical parameters were determined using 33 single- and multi-round transcription experiments spanning 487 in vitro measurements. By incorporating in vivo measurements of ppGpp and ATP, the model correctly predicted rRNA production rates for cellular responses to nutrient upshifts, downshifts, and outgrowth into fresh medium. Inclusion of DksA was essential in all three cases. Our work provides a foundation for using data-driven computational models to predict the kinetics of in vivo transcriptional responses.