Flux Balance Analysis Simulations Research Articles

Combining denoising of RNA-seq data and flux balance analysis for cluster analysis of single cells

BackgroundSophisticated methods to properly pre-process and analyze the increasing collection of single-cell RNA sequencing (scRNA-seq) data are increasingly being developed. On the contrary, the best practices to integrate these data into metabolic networks, aiming at describing metabolic phenotypes within a heterogeneous cell population, have been poorly investigated. In this regard, a critical factor is the presence of false zero values in reactions essential for a fundamental metabolic function, such as biomass or energy production. Here, we investigate the role of denoising strategies in mitigating this problem.MethodsWe applied state-of-the-art denoising strategies - namely MAGIC, ENHANCE, and SAVER - on three public scRNA-seq datasets. We then associated a metabolic flux distribution with every single cell by embedding its noise-free transcriptomics profile in the constraints of the optimization of a core metabolic model. Finally, we used the obtained single-cell optimal metabolic fluxes as features for cluster analysis. We compared the results obtained with different techniques, and with or without the use of denoising. We also investigated the possibility of applying denoising directly on the Reaction Activity Scores, which are metabolic features extracted from the read counts, rather than on the read counts.ResultsWe show that denoising of transcriptomics data improves the clustering of single cells. We also illustrate that denoising restores important metabolic properties, such as the correlation between cell cycle phase and biomass accumulation, and between the RAS scores of reactions belonging to the same metabolic pathway. We show that MAGIC performs better than ENHANCE and SAVER, and that, denoising applied directly on the RAS matrix could be an effective alternative in removing false zero values from essential metabolic reactions.ConclusionsOur results indicate that including denoising as a pre-processing operation represents a milestone to integrate scRNA-seq data into Flux Balance Analysis simulations and to perform single-cell cluster analysis with a focus on metabolic phenotypes.

Patient-Specific Genome-Scale Metabolic Models for Individualized Predictions of Liver Disease

The prevalence of liver disease is steadily increasing, coupled with the limited availability of therapeutic treatments. Recent literature points to metabolic reprogramming as a key feature of liver failure. Hence, we sought to uncover the metabolic pathways and mechanisms associated with liver disease and acute liver failure. We generated patient-specific genome scale metabolic models by integrating RNA-seq data from patient liver samples with a generalized human metabolic model. Flux balance analysis simulations showed a distinct separation of non-alcohol associated and alcohol-associated disease states. Our analysis suggests that the alcohol associated liver has an increased flux through nucleotide and glycerophospholipid metabolic pathways. By contrast, non-alcohol associated liver has an increased flux through fatty acid oxidation, the carnitine shuttle, and bile acid recycling pathways. Importantly, there was significant variation of metabolic fluxes between patients within the same clinical category of disease stage, pointing to the necessity and opportunity for personalized medicine in treating liver disease. We conclude that the metabolic reprogramming occurring in alcohol-associated liver disease is likely distinct from the adaptations in non-alcohol associated liver disease, potentially requiring alternative therapeutic approaches.

Accelerated global sensitivity analysis of genome-wide constraint-based metabolic models

BackgroundGenome-wide reconstructions of metabolism opened the way to thorough investigations of cell metabolism for health care and industrial purposes. However, the predictions offered by Flux Balance Analysis (FBA) can be strongly affected by the choice of flux boundaries, with particular regard to the flux of reactions that sink nutrients into the system. To mitigate possible errors introduced by a poor selection of such boundaries, a rational approach suggests to focus the modeling efforts on the pivotal ones.MethodsIn this work, we present a methodology for the automatic identification of the key fluxes in genome-wide constraint-based models, by means of variance-based sensitivity analysis. The goal is to identify the parameters for which a small perturbation entails a large variation of the model outcomes, also referred to as sensitive parameters. Due to the high number of FBA simulations that are necessary to assess sensitivity coefficients on genome-wide models, our method exploits a master-slave methodology that distributes the computation on massively multi-core architectures. We performed the following steps: (1) we determined the putative parameterizations of the genome-wide metabolic constraint-based model, using Saltelli’s method; (2) we applied FBA to each parameterized model, distributing the massive amount of calculations over multiple nodes by means of MPI; (3) we then recollected and exploited the results of all FBA runs to assess a global sensitivity analysis.ResultsWe show a proof-of-concept of our approach on latest genome-wide reconstructions of human metabolism Recon2.2 and Recon3D. We report that most sensitive parameters are mainly associated with the intake of essential amino acids in Recon2.2, whereas in Recon 3D they are associated largely with phospholipids. We also illustrate that in most cases there is a significant contribution of higher order effects.ConclusionOur results indicate that interaction effects between different model parameters exist, which should be taken into account especially at the stage of calibration of genome-wide models, supporting the importance of a global strategy of sensitivity analysis.

Role of type I NADH dehydrogenase in Synechocystis sp. PCC 6803 under phycobilisome excited red light

Cyanobacterial type I NADH dehydrogenase (NDH-1) is involved in various bioenergetic reactions including respiration, cyclic electron transport (CET), and CO2 uptake. The role of NDH-1 is usually minor under normal growth conditions and becomes important under stress conditions. However, in our previous study, flux balance analysis (FBA) simulation predicted that the drive of NDH-1 as CET pathway with a photosystem (PS) I/PSII excitation ratio around 1.0 contributes to achieving an optimal specific growth rate. In this study, to experimentally elucidate the predicted functions of NDH-1, first, we measured the PSI/PSII excitation ratios of Synechocystis sp. PCC 6803 grown under four types of spectral light conditions. The specific growth rate was the highest and PSI/PSII excitation ratio was with 0.88 under the single-peak light at 630 nm (Red1). Considering this measured excitation ratios, FBA simulation predicted that NDH-1-dependent electron transport was the major pathway under Red1. Moreover, in actual culture, an NDH-1 deletion strain had slower growth rate than that of wild type only under Red1 light condition. Therefore, we experimentally demonstrated that NDH-1 plays an important role under optimal light conditions such as Red1 light, where Synechocystis exhibits the highest specific growth rate and PSI/PSII excitation ratio of around 1.0.

Genome-scale metabolic modeling of Acetobacter pasteurianus 386B reveals its metabolic adaptation to cocoa fermentation conditions

Acetobacter pasteurianus 386B has been selected as a candidate functional starter culture to better control the cocoa fermentation process. Previously, its genome has been sequenced and a genome-scale metabolic model (GEM) has been reconstructed. To understand its metabolic adaptation to cocoa fermentation conditions, different flux balance analysis (FBA) simulations were performed and compared with experimental data. In particular, metabolic flux distributions were simulated for two phases that characterize the growth of A. pasteurianus 386B under cocoa fermentation conditions, predicting a switch in respiratory chain usage in between these phases. The possible influence on the resulting energy production was shown using a reduced version of the GEM. FBA simulations revealed the importance of the compartmentalization of the ethanol oxidation reactions, namely in the periplasm or in the cytoplasm, and highlighted the potential role of ethanol as a source of carbon, energy, and NADPH. Regarding the latter, the physiological function of a proton-translocating NAD(P)+ transhydrogenase was further investigated in silico. This study revealed the potential of using a GEM to simulate the metabolism of A. pasteurianus 386B, and may provide a general framework toward a better physiological understanding of functional starter cultures in food fermentation processes.

Distributed flux balance analysis simulations of serial biomass fermentation by two organisms.

Intelligent biorefinery design that addresses both the composition of the biomass feedstock as well as fermentation microorganisms could benefit from dedicated tools for computational simulation and computer-assisted optimization. Here we present the BioLego Vn2.0 framework, based on Microsoft Azure Cloud, which supports large-scale simulations of biomass serial fermentation processes by two different organisms. BioLego enables the simultaneous analysis of multiple fermentation scenarios and the comparison of fermentation potential of multiple feedstock compositions. Thanks to the effective use of cloud computing it further allows resource intensive analysis and exploration of media and organism modifications. We use BioLego to obtain biological and validation results, including (1) exploratory search for the optimal utilization of corn biomasses—corn cobs, corn fiber and corn stover—in fermentation biorefineries; (2) analysis of the possible effects of changes in the composition of K. alvarezi biomass on the ethanol production yield in an anaerobic two-step process (S. cerevisiae followed by E. coli); (3) analysis of the impact, on the estimated ethanol production yield, of knocking out single organism reactions either in one or in both organisms in an anaerobic two-step fermentation process of Ulva sp. into ethanol (S. cerevisiae followed by E. coli); and (4) comparison of several experimentally measured ethanol fermentation rates with the predictions of BioLego.

Provitamin D3 modulation through prebiotics supplementation: simulation based assessment

Vitamin D is important in multiple health conditions. Vitamin D deficiency is prevalent globally even with exposure to adequate sunlight. Reduction in provitamin D3 (7-dehydrocholesterol, 7-DHC) is an important cause of vitamin D3 deficiency. Vitamin supplementation, food fortification, and use of probiotics are some approaches to reduce vitamin D3 deficiency. This study investigates plausibility of 7-DHC biosynthesis through dietary prebiotics supplementation. Furthermore, it reports mechanistic details and constraints for the biosynthesis using flux balance analysis (FBA) simulations. The FBA simulations using co-metabolism models comprising human host and a resident bacterium (Faecalibacterium prausnitzii or Bacteroides thetaiotamicron) indicated increased flux of 7-DHC with short-chain fructooligosaccharide (scFOS) or inulin supplementation. We observed around 2-fold increase in flux compared to the baseline. Biosynthesis of 7-DHC was primarily modulated through acetate, pyruvate and lactate secreted by the bacterium. We observed diverse mechanisms and dose dependent responses. We extended this assessment to 119 resident bacteria and investigated the metabolites profiles with prebiotics supplementation. In summary, the current study suggests the potential use of applying prebiotics in enhancing 7-DHC biosynthesis. Furthermore, performance of the different gut bacteria with prebiotic supplementation for secreted metabolites profile is reported. These results may be useful to design future clinical studies.

FlySilico: Flux balance modeling of Drosophila larval growth and resource allocation

Organisms depend on a highly connected and regulated network of biochemical reactions fueling life sustaining and growth promoting functions. While details of this metabolic network are well established, knowledge of the superordinate regulatory design principles is limited. Here, we investigated by iterative wet lab and modeling experiments the resource allocation process during the larval development of Drosophila melanogaster. We chose this system, as survival of the animals depends on the successful allocation of their available resources to the conflicting processes of growth and storage metabolite deposition. First, we generated “FlySilico”, a curated metabolic network of Drosophila, and performed time-resolved growth and metabolite measurements with larvae raised on a holidic diet. Subsequently, we performed flux balance analysis simulations and tested the predictive power of our model by simulating the impact of diet alterations on growth and metabolism. Our predictions correctly identified the essential amino acids as growth limiting factor, and metabolic flux differences in agreement with our experimental data. Thus, we present a framework to study important questions of resource allocation in a multicellular organism including process priorization and optimality principles.

Flux balance analysis of cyanobacteria reveals selective use of photosynthetic electron transport components under different spectral lightconditions.

Cyanobacteria acclimate and adapt to changing light conditions by controlling the energy transfer between photosystem I (PSI) and II (PSII) and pigment composition. Photosynthesis is driven by balancing the excitation between PSI and PSII. To predict the detailed electron transfer flux of cyanobacteria, we refined the photosynthesis-related reactions in our previously reconstructed genome-scale model. Two photosynthetic bacteria, Arthrospira and Synechocystis, were used as models. They were grown under various spectral light conditions and flux balance analysis (FBA) was performed using photon uptake fluxes into PSI and PSII, which were converted from each light spectrum by considering the photoacclimation of pigments and the distribution ratio of phycobilisome to PSI and PSII. In Arthrospira, the FBA was verified with experimental data using six types of light-emitting diodes (White, Blue, Green, Yellow, Red1, and Red2). FBA predicted the cell growth of Synechocystis for the LEDs, excepting Red2. In an FBA simulation, cells used respiratory terminal oxidases and two NADH dehydrogenases (NDH-1 and NDH-2) to balance the PSI and PSII excitations depending on the light conditions. FBA simulation with our refined model functionally implicated NDH-1 and NDH-2 as a component of cyclic electron transport in the varied light environments.

Effects of MCHM on yeast metabolism.

On January 2014 approximately 10,000 gallons of crude 4-Methylcyclohexanemethanol (MCHM) and propylene glycol phenol ether (PPH) were accidentally released into the Elk River, West Virginia, contaminating the tap water of around 300,000 residents. Crude MCHM is an industrial chemical used as flotation reagent to clean coal. At the time of the spill, MCHM's toxicological data were limited, an issue that has been addressed by different studies focused on understanding the immediate and long-term effects of MCHM on human health and the environment. Using S. cerevisiae as a model organism we study the effect of acute exposure to crude MCHM on metabolism. Yeasts were treated with MCHM 550 ppm in YPD for 30 minutes. Polar and lipid metabolites were extracted from cells by a chloroform-methanol-water mixture. The extracts were then analyzed by direct injection ESI-MS and by GC-MS. The metabolomics analysis was complemented with flux balance analysis simulations done with genome-scale metabolic network models (GSMNM) of MCHM treated vs non-treated control. We integrated the effect of MCHM on yeast gene expression from RNA-Seq data within these GSMNM. A total of 215 and 73 metabolites were identified by the ESI-MS and GC-MS procedures, respectively. From these 26 and 23 relevant metabolites were selected from ESI-MS and GC-MS respectively, for 49 unique compounds. MCHM induced amino acid accumulation, via its effects on amino acid metabolism, as well as a potential impairment of ribosome biogenesis. MCHM affects phospholipid biosynthesis, with a potential impact on the biophysical properties of yeast cellular membranes. The FBA simulations were able to reproduce the deleterious effect of MCHM on cellular growth and suggest that the effect of MCHM on ubiquinol:ferricytochrome c reductase reaction, caused by the under-expression of CYT1 gene, could be the driven force behind the observed effect on yeast metabolism and growth.

Spatiotemporal modeling of microbial metabolism.

BackgroundMicrobial systems in which the extracellular environment varies both spatially and temporally are very common in nature and in engineering applications. While the use of genome-scale metabolic reconstructions for steady-state flux balance analysis (FBA) and extensions for dynamic FBA are common, the development of spatiotemporal metabolic models has received little attention.ResultsWe present a general methodology for spatiotemporal metabolic modeling based on combining genome-scale reconstructions with fundamental transport equations that govern the relevant convective and/or diffusional processes in time and spatially varying environments. Our solution procedure involves spatial discretization of the partial differential equation model followed by numerical integration of the resulting system of ordinary differential equations with embedded linear programs using DFBAlab, a MATLAB code that performs reliable and efficient dynamic FBA simulations. We demonstrate our methodology by solving spatiotemporal metabolic models for two systems of considerable practical interest: (1) a bubble column reactor with the syngas fermenting bacterium Clostridium ljungdahlii; and (2) a chronic wound biofilm with the human pathogen Pseudomonas aeruginosa. Despite the complexity of the discretized models which consist of 900 ODEs/600 LPs and 250 ODEs/250 LPs, respectively, we show that the proposed computational framework allows efficient and robust model solution.ConclusionsOur study establishes a new paradigm for formulating and solving genome-scale metabolic models with both time and spatial variations and has wide applicability to natural and engineered microbial systems.Electronic supplementary materialThe online version of this article (doi:10.1186/s12918-016-0259-2) contains supplementary material, which is available to authorized users.

Genome-scale metabolic model of Rhodococcus jostii RHA1 (iMT1174) to study the accumulation of storage compounds during nitrogen-limited condition.

BackgroundRhodococcus jostii RHA1 growing on different substrates is capable of accumulating simultaneously three types of carbon storage compounds: glycogen, polyhydroxyalkanoates (PHA), and triacylglycerols (TAG). Under nitrogen-limited (N-limited) condition, the level of storage increases as is commonly observed for other bacteria. The proportion of each storage compound changes with substrate, but it remains unclear what modelling approach should be adopted to predict the relative composition of the mixture of the storage compounds. We analyzed the growth of R. jostii RHA1 under N-limited conditions using a genome-scale metabolic modelling approach to determine which global metabolic objective function could be used for the prediction.ResultsThe R. jostii RHA1 model (iMT1174) produced during this study contains 1,243 balanced metabolites, 1,935 unique reactions, and 1,174 open reading frames (ORFs).Seven objective functions used with flux balance analysis (FBA) were compared for their capacity to predict the mixture of storage compounds accumulated after the sudden onset of N-limitation. Predictive abilities were determined using a Bayesian approach. Experimental data on storage accumulation mixture (glycogen, polyhydroxyalkanoates, and triacylglycerols) were obtained for batch cultures grown on glucose or acetate. The best FBA simulation results were obtained using a novel objective function for the N-limited condition which combined the maximization of the storage fluxes and the minimization of metabolic adjustments (MOMA) with the preceding non-limited conditions (max storage + environmental MOMA). The FBA solutions for the non-limited growth conditions were simply constrained by the objective function of growth rate maximization. Measurement of central metabolic fluxes by 13C-labelling experiments of amino acids further supported the application of the environmental MOMA principle in the context of changing environment. Finally, it was found that the quantitative predictions of the storage mixture during N-limited storage accumulation were fairly sensitive to the biomass composition, as expected.ConclusionsThe genome-scale metabolic model analysis of R. jostii RHA1 cultures suggested that the intracellular reaction flux profile immediately after the onset of N-limited condition are impacted by the values of the same fluxes during the period of non-limited growth. PHA turned out to be the main storage pool of the mixture in R. jostii RHA1.Electronic supplementary materialThe online version of this article (doi:10.1186/s12918-015-0190-y) contains supplementary material, which is available to authorized users.

Software applications for flux balance analysis

Flux balance analysis (FBA) is a widely used computational method for characterizing and engineering intrinsic cellular metabolism. The increasing number of its successful applications and growing popularity are possibly attributable to the availability of specific software tools for FBA. Each tool has its unique features and limitations with respect to operational environment, user-interface and supported analysis algorithms. Presented herein is an in-depth evaluation of currently available FBA applications, focusing mainly on usability, functionality, graphical representation and inter-operability. Overall, most of the applications are able to perform basic features of model creation and FBA simulation. COBRA toolbox, OptFlux and FASIMU are versatile to support advanced in silico algorithms to identify environmental and genetic targets for strain design. SurreyFBA, WEbcoli, Acorn, FAME, GEMSiRV and MetaFluxNet are the distinct tools which provide the user friendly interfaces in model handling. In terms of software architecture, FBA-SimVis and OptFlux have the flexible environments as they enable the plug-in/add-on feature to aid prospective functional extensions. Notably, an increasing trend towards the implementation of more tailored e-services such as central model repository and assistance to collaborative efforts was observed among the web-based applications with the help of advanced web-technologies. Furthermore, most recent applications such as the Model SEED, FAME, MetaFlux and MicrobesFlux have even included several routines to facilitate the reconstruction of genome-scale metabolic models. Finally, a brief discussion on the future directions of FBA applications was made for the benefit of potential tool developers.

The genome-scale metabolic model iIN800 of Saccharomyces cerevisiae and its validation: a scaffold to query lipid metabolism

BackgroundUp to now, there have been three published versions of a yeast genome-scale metabolic model: iFF708, iND750 and iLL672. All three models, however, lack a detailed description of lipid metabolism and thus are unable to be used as integrated scaffolds for gaining insights into lipid metabolism from multilevel omic measurement technologies (e.g. genome-wide mRNA levels). To overcome this limitation, we reconstructed a new version of the Saccharomyces cerevisiae genome-scale model, iIN800 that includes a more rigorous and detailed description of lipid metabolism.ResultsThe reconstructed metabolic model comprises 1446 reactions and 1013 metabolites. Beyond incorporating new reactions involved in lipid metabolism, we also present new biomass equations that improve the predictive power of flux balance analysis simulations. Predictions of both growth capability and large scale in silico single gene deletions by iIN800 were consistent with experimental data. In addition, 13C-labeling experiments validated the new biomass equations and calculated intracellular fluxes. To demonstrate the applicability of iIN800, we show that the model can be used as a scaffold to reveal the regulatory importance of lipid metabolism precursors and intermediates that would have been missed in previous models from transcriptome datasets.ConclusionPerforming integrated analyses using iIN800 as a network scaffold is shown to be a valuable tool for elucidating the behavior of complex metabolic networks, particularly for identifying regulatory targets in lipid metabolism that can be used for industrial applications or for understanding lipid disease states.