Hyun-Seob Song

Reduction of a set of elementary modes using yield analysis

This article proposes a new concept termed “yield analysis” (YA) as a method of extracting a subset of elementary modes (EMs) essential for describing metabolic behaviors. YA can be defined as the analysis of metabolic pathways in yield space where the solution space is a bounded convex hull. Two important issues arising in the analysis and modeling of a metabolic network are handled. First, from a practical sense, the minimal generating set spanning the yield space is recalculated. This refined generating set excludes all the trivial modes with negligible contribution to convex hull in yield space. Second, we revisit the problem of decomposing the measured fluxes among the EMs. A consistent way of choosing the unique, minimal active modes among a number of possible candidates is discussed and compared with two other existing methods, that is, those of Schwartz and Kanehisa (Schwartz and Kanehisa, 2005. Bioinformatics 21: 204–205) and of Provost et al. (Provost et al., 2007. Proceedings of the 10th IFAC Symposium on Computer Application in Biotechnology, 321–326). The proposed idea is tested in a case study of a metabolic network of recombinant yeasts fermenting both glucose and xylose. Due to the nature of the network with multiple substrates, the flux space is split into three independent yield spaces to each of which the two‐staged reduction procedure is applied. Through a priori reduction without any experimental input, the 369 EMs in total was reduced to 35 modes, which correspond to about 91% reduction. Then, three and four modes were finally chosen among the reduced set as the smallest active sets for the cases with a single substrate of glucose and xylose, respectively. It should be noted that the refined minimal generating set obtained from a priori reduction still provides a practically complete description of all possible states in the subspace of yields, while the active set covers only a specific set of experimental data. Biotechnol. Bioeng. 2009;102: 554–568.

Operating strategies for Fischer-Tropsch reactors: A model-directed study

A comprehensive parametric study for a Fischer-Tropsch (FT) synthesis process has been conducted to investigate the relation between process parameters and reactor characteristics such as conversion, selectivity, multiplicity, and stability. A flexible model was employed for this purpose, featuring the dependence of Anderson-Shultz-Flory (ASF) factor on composition and temperature. All variable process parameters in industrial FT reactors were subject to variation, including reaction temperature, reactor pressure, feed ratio, inlet mass flux, feed temperature, heat transfer coefficient, catalyst concentration, catalyst activity, etc. While typical trade-off was encountered in most cases, i.e., the change of a parameter in one direction enhances one aspect but deteriorating another, the change of feed conditions gave some promising results. It has been found that decreasing the feed rate (or increasing the residence time) and/or lowering the feed concentration can successfully enhance the conversion up to more than 90% for our specific case, without hurting the product selectivity as well as effectively condense the region of multiple steady states. The benefits and limitations accompanied with the variation of the parameters were discussed in detail and a rational start-up strategy was proposed based on the preceding results. It is shown that the decrease of inlet mass flux (say, 85% decrease of the feed rate or 60% decrease of the feed concentration from the nominal condition chosen here) or the decrease of H2/CO ratio (specifically, below about 0.25), or their combination can eliminate multiple steady states. The resulting unique relation between temperature and manipulated variable (i.e., coolant flow rate) appears to assure a safe arrival at the target condition at the start-up stage.

Systematic development of hybrid cybernetic models: Application to recombinant yeast co-consuming glucose and xylose

The hybrid cybernetic modeling approach of Kim et al. (Kim et al. [2008] Biotechnol. Prog., in press) views the substrate uptake flux in microorganisms as being distributed in a regulated way among different elementary modes (EMs) of a metabolic network, which intracellular fluxes related to the uptake rates by the pseudo‐steady‐state approximation on intracellular metabolites. While the conceptual development has been demonstrated by Kim et al. (Kim et al. [2008] Biotechnol. Prog., in press) using a rather simple example (i.e., Escherichia coli metabolizing a single substrate), its extension to a larger scale network involving multiple substrates results in serious overparameterization (which implies an excessive number of parameters relative to the measurements available to determine them). Through the case study of recombinant Saccharomyces yeast co‐consuming glucose and xylose, we present a systematic way of formulating a minimal order hybrid cybernetic model (HCM) for a general metabolic network. The overparameterization problem mostly arising from a large number of EMs is avoided using a model reduction technique developed by Song and Ramkrishna (Song and Ramkrishna [2009a] Biotechnol. Bioeng. 102(2):554–568) where an original set of EMs is condensed to a much smaller subset. Detailed discussions follow on the issue of determining the minimal set of active modes needed for the description of the simultaneous consumption of multiple substrates. The developed HCM is compared with other metabolic models: macroscopic bioreaction models (Provost et al. [2006] Bioprocess Biosyt. Eng. 29(5–6):349–366), and dynamic flux balance analysis. It is shown that the HCM outperforms the other two as validated using various sets of fermentation data. The difference among the models is more dramatic in a situation such as the sequential utilization of glucose and xylose, which is observed under realistic fermentation conditions. Biotechnol. Bioeng. 2009;103: 984–1002.

Prediction of metabolic function from limited data: Lumped hybrid cybernetic modeling (L-HCM)

Motivated by the need for a quick quantitative assessment of metabolic function without extensive data, we present an adaptation of the cybernetic framework, denoted as the lumped hybrid cybernetic model (L‐HCM), which combines the attributes of the classical lumped cybernetic model (LCM) and the recently developed HCM. The basic tenet of L‐HCM and HCM is the same, that is, they both view the uptake flux as being split among diverse pathways in an optimal way as a result of cellular regulation such that some chosen metabolic objective is realized. The L‐HCM, however, portrays this flux distribution to occur in a hierarchical way, that is, first among lumped pathways, and next among individual elementary modes (EM) in each lumped pathway. Both splits are described by the cybernetic control laws using operational and structural return‐on‐investments, respectively. That is, the distribution of uptake flux at the first split is dynamically regulated according to environmental conditions, while the subsequent split is based purely on the stoichiometry of EMs. The resulting model is conveniently represented in terms of lumped pathways which are fully identified with respect to yield coefficients of all products unlike classical LCMs based on instinctive lumping. These characteristics enable the model to account for the complete set of EMs for arbitrarily large metabolic networks despite containing only a small number of parameters which can be identified using minimal data. However, the inherent conflict of questing for quantification of larger networks with smaller number of parameters cannot be resolved without a mechanism for parameter tuning of an empirical nature. In this work, this is accomplished by manipulating the relative importance of EMs by tuning the cybernetic control of mode‐averaged enzyme activity with an empirical parameter. In a case study involving aerobic batch growth of Saccharomyces cerevisiae, L‐HCM is compared with LCM. The former provides a much more satisfactory prediction than the latter when parameters are identified from a few primary metabolites. On the other hand, the classical model is more accurate than L‐HCM when sufficient datasets are involved in parameter identification. In applying the two models to a chemostat scenario, L‐HCM shows a reasonable prediction on metabolic shift from respiration to fermentation due to the Crabtree effect, which LCM predicts unsatisfactorily. While L‐HCM appears amenable to expeditious estimates of metabolic function with minimal data, the more detailed dynamic models [such as HCM or those of Young et al. (Young et al., Biotechnol Bioeng, 2008; 100: 542–559)] are best suited for accurate treatment of metabolism when the potential of modern omic technology is fully realized. However, in view of the monumental effort surrounding the development of detailed models from extensive omic measurements, the preliminary insight into the behavior of a genotype and metabolic engineering directives that can come from L‐HCM is indeed valuable. Biotechnol. Bioeng. 2010;106: 271–284.

Cybernetic Models Based on Lumped Elementary Modes Accurately Predict Strain-Specific Metabolic Function

In a recent article, Song and Ramkrishna (Song and Ramkrishna [2010]. Biotechnol Bioeng 106(2):271–284) proposed a lumped hybrid cybernetic model (L‐HCM) towards extracting maximum information about metabolic function from a minimum of data. This approach views the total uptake flux as distributed among lumped elementary modes (L‐EMs) so as to maximize a prescribed metabolic objective such as growth or uptake rate. L‐EM is computed as a weighted average of EMs where the weights are related to the yields of vital products (i.e., biomass and ATP). In this article, we further enhance the predictive power of L‐HCMs through modifications in lumping weights with additional parameters that can be tuned with data viewed to be critical. The resulting model is able to make predictions of diverse metabolic behaviors varying greatly with strain types as evidenced from case studies of anaerobic growth of various Escherichia coli strains. Incorporation of the new lumping formula into L‐HCM remarkably improves model predictions with a few critical data, thus presenting L‐HCM as a dynamic tool as being not only qualitatively correct but also quantitatively accurate. Biotechnol. Bioeng. 2011; 108:127–140.

Exacting predictions by cybernetic model confirmed experimentally: Steady state multiplicity in the chemostat

We demonstrate strong experimental support for the cybernetic model based on maximizing carbon uptake rate in describing the microorganisms regulatory behavior by verifying exacting predictions of steady state multiplicity in a chemostat. Experiments with a feed mixture of glucose and pyruvate show multiple steady state behavior as predicted by the cybernetic model. When multiplicity occurs at a dilution (growth) rate, it results in hysteretic behavior following switches in dilution rate from above and below. This phenomenon is caused by transient paths leading to different steady states through dynamic maximization of the carbon uptake rate. Thus steady state multiplicity is a manifestation of the nonlinearity arising from cybernetic mechanisms rather than of the nonlinear kinetics. The predicted metabolic multiplicity would extend to intracellular states such as enzyme levels and fluxes to be verified in future experiments.

Microbial Community Metabolic Modeling: A Community Data-Driven Network Reconstruction

Metabolic network modeling of microbial communities provides an in‐depth understanding of community‐wide metabolic and regulatory processes. Compared to single organism analyses, community metabolic network modeling is more complex because it needs to account for interspecies interactions. To date, most approaches focus on reconstruction of high‐quality individual networks so that, when combined, they can predict community behaviors as a result of interspecies interactions. However, this conventional method becomes ineffective for communities whose members are not well characterized and cannot be experimentally interrogated in isolation. Here, we tested a new approach that uses community‐level data as a critical input for the network reconstruction process. This method focuses on directly predicting interspecies metabolic interactions in a community, when axenic information is insufficient. We validated our method through the case study of a bacterial photoautotroph–heterotroph consortium that was used to provide data needed for a community‐level metabolic network reconstruction. Resulting simulations provided experimentally validated predictions of how a photoautotrophic cyanobacterium supports the growth of an obligate heterotrophic species by providing organic carbon and nitrogen sources. J. Cell. Physiol. 231: 2339–2345, 2016.

Dynamic modeling of aerobic growth of Shewanella oneidensis. Predicting triauxic growth, flux distributions, and energy requirement for growth

A model-based analysis is conducted to investigate metabolism of Shewanella oneidensis MR-1 strain in aerobic batch culture, which exhibits an intriguing growth pattern by sequentially consuming substrate (i.e., lactate) and by-products (i.e., pyruvate and acetate). A general protocol is presented for developing a detailed network-based dynamic model for S. oneidensis based on the Lumped Hybrid Cybernetic Model (L-HCM) framework. The L-HCM, although developed from only limited data, is shown to accurately reproduce exacting dynamic metabolic shifts, and provide reasonable estimates of energy requirement for growth. Flux distributions in S. oneidensis predicted by the L-HCM compare very favorably with (13)C-metabolic flux analysis results reported in the literature. Predictive accuracy is enhanced by incorporating measurements of only a few intracellular fluxes, in addition to extracellular metabolites. The L-HCM developed here for S. oneidensis is consequently a promising tool for the analysis of intracellular flux distribution and metabolic engineering.

On enhancing productivity of bioethanol with multiple species

The present work is initiated to investigate whether a defined culture comprising a mixture of three yeast species, Kluyveromyces marxianus, Saccharomyces cerevisiae, and Pichia stipitis can ferment a mixture of sugars to produce bioethanol at rates higher than those achieved by pure cultures of the same. For this purpose, we develop models of single species based on the hybrid cybernetic model framework, and simulate fermentations in the mixed culture by combining individual models. An underlying assumption is that the behavior of each species is determined only by the common environment independently of the presence and metabolism of other species. Model performance is thoroughly assessed using the experimental data available in the literature. The dynamic behavior of mixed cultures in mixed culture experiments are accurately predicted by the model reflecting faithfully the simultaneous/sequential uptake patterns of mixed substrates. This model is then used to investigate performance of various possible reactor configurations. With the foregoing species of organisms, mixed culture itself does not lead to a significant increase of bioethanol productivity. Rather, the model shows that substantial improvement is acquired by sequential use of different, properly chosen organisms during fermentation. Thus, the successive use of K. marxianus and P. stipitis is shown to increase bioethanol productivity up to about 58% in comparison to fermentation by single species alone. Biotechnol. Bioeng. 2012; 109:1508–1517.

Multiplicity and sensitivity analysis of Fischer–Tropsch bubble column slurry reactors: plug-flow gas and well-mixed slurry model

Existence of multiple steady states of Fischer–Tropsch (FT) synthesis has been reported experimentally as well as theoretically even in simple stirred tank slurry reactors. Bhattacharjee, Tierney, and Shah (1986) have found steady-state multiplicity in product distribution and heat generation rate in their experiments on supported ruthenium catalyst. Shah, Dassori, and Tierney (1990) have subsequently provided an explanation of this observation based on the analysis of heat generation and dissipation curves. They maintained that ignition should be avoided since, once it occurs, the normal FT synthesis reaction would be switched to the methane-forming mode seemingly never to return. This is because the catalyst could su=er serious deactivation on exposure to the high temperature resulting from the ignition. Recently, Song, Ramkrishna, Trinh, andWright (2003) have reported more complex experimental observations for an FT stirred tank slurry reactor with cobalt catalyst. The operating state of the FT process suddenly jumps from the normal wax-producing mode to the undesirable methane-forming mode accompanying the abrupt temperature lift. This multiplicity behavior is peculiar in the following two senses: (i) this happens sporadically, not always, and (ii) the normal wax-producing mode was, contrary to the interpretation by Shah et al. (1990), eventually recovered after a couple of hours. A plausible scenario explaining this peculiar multiplicity behavior was proposed by Song et al. (2003) using rigorous nonlinear analysis. It was found that decrease of the Stanton number for heat transfer (StH ) could be responsible for the @rst jump, while a decrease of the DamkA ohler