MLAGO: machine learning-aided global optimization for Michaelis constant estimation of kinetic modeling.

BACKGROUND: Kinetic modeling is a powerful tool for understanding the dynamic behavior of biochemical systems. For kinetic modeling, determination of a number of kinetic parameters, such as the Michaelis constant (Km), is necessary, and global optimization algorithms have long been used for parameter estimation. However, the conventional global optimization approach has three problems: (i) It is computationally demanding. (ii) It often yields unrealistic parameter values because it simply seeks a better model fitting to experimentally observed behaviors. (iii) It has difficulty in identifying a unique solution because multiple parameter sets can allow a kinetic model to fit experimental data equally well (the non-identifiability problem). RESULTS: To solve these problems, we propose the Machine Learning-Aided Global Optimization (MLAGO) method for Km estimation of kinetic modeling. First, we use a machine learning-based Km predictor based only on three factors: EC number, KEGG Compound ID, and Organism ID, then conduct a constrained global optimization-based parameter estimation by using the machine learning-predicted Km values as the reference values. The machine learning model achieved relatively good prediction scores: RMSE = 0.795 and R2 = 0.536, making the subsequent global optimization easy and practical. The MLAGO approach reduced the error between simulation and experimental data while keeping Km values close to the machine learning-predicted values. As a result, the MLAGO approach successfully estimated Km values with less computational cost than the conventional method. Moreover, the MLAGO approach uniquely estimated Km values, which were close to the measured values. CONCLUSIONS: MLAGO overcomes the major problems in parameter estimation, accelerates kinetic modeling, and thus ultimately leads to better understanding of complex cellular systems. The web application for our machine learning-based Km predictor is accessible at https://sites.google.com/view/kazuhiro-maeda/software-tools-web-apps , which helps modelers perform MLAGO on their own parameter estimation tasks.

Ranking network mechanisms by how they fit diverse experiments and deciding on E. coli's ammonium transport and assimilation network.

The complex ammonium transport and assimilation network of E. coli involves the ammonium transporter AmtB, the regulatory proteins GlnK and GlnB, and the central N-assimilating enzymes together with their highly complex interactions. The engineering and modelling of such a complex network seem impossible because functioning depends critically on a gamut of data known at patchy accuracy. We developed a way out of this predicament, which employs: (i) a constrained optimization-based technology for the simultaneous fitting of models to heterogeneous experimental data sets gathered through diverse experimental set-ups, (ii) a 'rubber band method' to deal with different degrees of uncertainty, both in experimentally determined or estimated parameter values and in measured transient or steady-state variables (training data sets), (iii) integration of human expertise to decide on accuracies of both parameters and variables, (iv) massive computation employing a fast algorithm and a supercomputer, (v) an objective way of quantifying the plausibility of models, which makes it possible to decide which model is the best and how much better that model is than the others. We applied the new technology to the ammonium transport and assimilation network, integrating recent and older data of various accuracies, from different expert laboratories. The kinetic model objectively ranked best, has E. coli's AmtB as an active transporter of ammonia to be assimilated with GlnK minimizing the futile cycling that is an inevitable consequence of intracellular ammonium accumulation. It is 130 times better than a model with facilitated passive transport of ammonia.

Generation of the membrane potential and its impact on the motility, ATP production and growth in Campylobacter jejuni.

The generation of a membrane potential (Δψ), the major constituent of the proton motive force (pmf), is crucial for ATP synthesis, transport of nutrients and flagellar rotation. Campylobacter jejuni harbors a branched electron transport chain, enabling respiration with different electron donors and acceptors. Here, we demonstrate that a relatively high Δψ is only generated in the presence of either formate as electron donor or oxygen as electron acceptor, in combination with an acceptor/donor respectively. We show the necessity of the pmf for motility and growth of C. jejuni. ATP generation is not only accomplished by oxidative phosphorylation via the pmf, but also by substrate-level phosphorylation via the enzyme AckA. In response to a low oxygen tension, C. jejuni increases the transcription and activity of the donor complexes formate dehydrogenase (FdhABC) and hydrogenase (HydABCD) as well as the transcription of the alternative respiratory acceptor complexes. Our findings suggest that in the gut of warm-blooded animals, C. jejuni depends on at least formate or hydrogen as donor (in the anaerobic lumen) or oxygen as acceptor (near the epithelial cells) to generate a pmf that sustains efficient motility and growth for colonization and pathogenesis.

Macromolecular networks and intelligence in microorganisms.

Living organisms persist by virtue of complex interactions among many components organized into dynamic, environment-responsive networks that span multiple scales and dimensions. Biological networks constitute a type of information and communication technology (ICT): they receive information from the outside and inside of cells, integrate and interpret this information, and then activate a response. Biological networks enable molecules within cells, and even cells themselves, to communicate with each other and their environment. We have become accustomed to associating brain activity - particularly activity of the human brain - with a phenomenon we call "intelligence." Yet, four billion years of evolution could have selected networks with topologies and dynamics that confer traits analogous to this intelligence, even though they were outside the intercellular networks of the brain. Here, we explore how macromolecular networks in microbes confer intelligent characteristics, such as memory, anticipation, adaptation and reflection and we review current understanding of how network organization reflects the type of intelligence required for the environments in which they were selected. We propose that, if we were to leave terms such as "human" and "brain" out of the defining features of "intelligence," all forms of life - from microbes to humans - exhibit some or all characteristics consistent with "intelligence." We then review advances in genome-wide data production and analysis, especially in microbes, that provide a lens into microbial intelligence and propose how the insights derived from quantitatively characterizing biomolecular networks may enable synthetic biologists to create intelligent molecular networks for biotechnology, possibly generating new forms of intelligence, first in silico and then in vivo.

Nitrogen assimilation in Escherichia coli: putting molecular data into a systems perspective.

We present a comprehensive overview of the hierarchical network of intracellular processes revolving around central nitrogen metabolism in Escherichia coli. The hierarchy intertwines transport, metabolism, signaling leading to posttranslational modification, and transcription. The protein components of the network include an ammonium transporter (AmtB), a glutamine transporter (GlnHPQ), two ammonium assimilation pathways (glutamine synthetase [GS]-glutamate synthase [glutamine 2-oxoglutarate amidotransferase {GOGAT}] and glutamate dehydrogenase [GDH]), the two bifunctional enzymes adenylyl transferase/adenylyl-removing enzyme (ATase) and uridylyl transferase/uridylyl-removing enzyme (UTase), the two trimeric signal transduction proteins (GlnB and GlnK), the two-component regulatory system composed of the histidine protein kinase nitrogen regulator II (NRII) and the response nitrogen regulator I (NRI), three global transcriptional regulators called nitrogen assimilation control (Nac) protein, leucine-responsive regulatory protein (Lrp), and cyclic AMP (cAMP) receptor protein (Crp), the glutaminases, and the nitrogen-phosphotransferase system. First, the structural and molecular knowledge on these proteins is reviewed. Thereafter, the activities of the components as they engage together in transport, metabolism, signal transduction, and transcription and their regulation are discussed. Next, old and new molecular data and physiological data are put into a common perspective on integral cellular functioning, especially with the aim of resolving counterintuitive or paradoxical processes featured in nitrogen assimilation. Finally, we articulate what still remains to be discovered and what general lessons can be learned from the vast amounts of data that are available now.

Isolation and characterization of gallium resistant Pseudomonas aeruginosa mutants.

Pseudomonas aeruginosa PA14 cells resistant to the novel antimicrobial gallium nitrate (Ga) were developed using transposon mutagenesis and by selecting spontaneous mutants. The mutants showing the highest growth in the presence of Ga were selected for further characterization. These mutants showed 4- to 12-fold higher Ga minimal inhibitory growth concentrations and a greater than 8-fold increase in the minimum biofilm eliminating Ga concentration. Both types of mutants produced Ga resistant biofilms whereas the formation of wild-type biofilms was strongly inhibited by Ga. The gene interrupted in the transposon mutant was hitA, which encodes a periplasmic iron binding protein that delivers Fe³âº to the HitB iron permease; complementation of the mutant with the hitA gene restored the Ga sensitivity. This hitA mutant showed a 14-fold decrease in Ga internalization versus the wild-type strain, indicating that the HitAB system is also involved in the Ga uptake. Ga uptake in the spontaneous mutant was also lower, although no mutations were found in the hitAB genes. Instead, this mutant harbored 64 non-silent mutations in several genes including those of the phenazine pyocyanin biosynthesis. The spontaneous mutant produced 2-fold higher pyocyanin basal levels than the wild-type; the addition of this phenazine to wild-type cultures protected them from the Ga bacteriostatic effect. The present data indicate that mutations affecting Ga transport and probably pyocyanin biosynthesis enable cells to develop resistance to Ga.

A new regulatory principle for in vivo biochemistry: pleiotropic low affinity regulation by the adenine nucleotides--illustrated for the glycolytic enzymes of Saccharomyces cerevisiae.

Enzymology tends to focus on highly specific effects of substrates, allosteric modifiers, and products occurring at low concentrations, because these are most informative about the enzyme's catalytic mechanism. We hypothesized that at relatively high in vivo concentrations, important molecular monitors of the state of living cells, such as ATP, affect multiple enzymes of the former and that these interactions have gone unnoticed in enzymology. We test this hypothesis in terms of the effect that ATP, ADP, and AMP might have on the major free-energy delivering pathway of the yeast Saccharomyces cerevisiae. Assaying cell-free extracts, we collected a comprehensive set of quantitative kinetic data concerning the enzymes of the glycolytic and the ethanol fermentation pathways. We determined systematically the extent to which the enzyme activities depend on the concentrations of the adenine nucleotides. We found that the effects of the adenine nucleotides on enzymes catalysing reactions in which they are not directly involved as substrate or product, are substantial. This includes effects on the Michaelis-Menten constants, adding new perspective on these, 100 years after their introduction.

Why in vivo may not equal in vitro - new effectors revealed by measurement of enzymatic activities under the same in vivo-like assay conditions.

Does the understanding of the dynamics of biochemical networks in vivo, in terms of the properties of their components determined in vitro, require the latter to be determined all under the same conditions? An in vivo-like assay medium for enzyme activity determination was designed based on the concentrations of the major ionic constituents of the Escherichia coli cytosol: K(+), Na(+), Mg(2+), phosphate, glutamate, sulfate and Cl(-). The maximum capacities (V(max)) of the extracted enzymes of two pathways were determined using both this in vivo-like assay medium and the assay medium specific for each enzyme. The enzyme activities differed between the two assay conditions. Most of the differences could be attributed to unsuspected, pleiotropic effects of K(+) and phosphate. K(+) activated some enzymes (aldolase, enolase and glutamate dehydrogenase) and inhibited others (phosphoglucose isomerase, phosphofructokinase, triosephosphate isomerase, glyceraldehyde 3-phosphate dehydrogenase, phosphoglycerate kinase, phosphoglycerate mutase), whereas phosphate inhibited all glycolytic enzymes and glutamine synthetase but only activated glutamine 2-oxoglutarate amidotransferase. Neither a high glutamate concentration, nor macromolecular crowding affected the glycolytic or nitrogen assimilation enzymes, other than through the product inhibition of glutamate dehydrogenase by glutamate. This strategy of assessing all pathway enzymes kinetically under the same conditions may be necessary to avoid inadvertent differences between in vivo and in vitro biochemistry. It may also serve to reveal otherwise unnoticed pleiotropic regulation, such as that demonstrated in the present study by K(+) and phosphate.

Emergence of the silicon human and network targeting drugs.

The development of disease may be characterized as a pathological shift of homeostasis; the main goal of contemporary drug treatment is, therefore, to return the pathological homeostasis back to the normal physiological range. From the view point of systems biology, homeostasis emerges from the interactions within the network of biomolecules (e.g. DNA, mRNA, proteins), and, hence, understanding how drugs impact upon the entire network should improve their efficacy at returning the network (body) to physiological homeostasis. Large, mechanism-based computer models, such as the anticipated human whole body models (silicon or virtual human), may help in the development of such network-targeting drugs. Using the philosophical concept of weak and strong emergence, we shall here take a more general look at the paradigm of network-targeting drugs, and propose our approaches to scale the strength of strong emergence. We apply these approaches to several biological examples and demonstrate their utility to reveal principles of bio-modeling. We discuss this in the perspective of building the silicon human.

AmtB-mediated NH3 transport in prokaryotes must be active and as a consequence regulation of transport by GlnK is mandatory to limit futile cycling of NH4(+)/NH3.

The nature of the ammonium import into prokaryotes has been controversial. A systems biological approach makes us hypothesize that AmtB-mediated import must be active for intracellular NH(4)(+) concentrations to sustain growth. Revisiting experimental evidence, we find the permeability assays reporting passive NH(3) import inconclusive. As an inevitable consequence of the proposed NH(4)(+) transport, outward permeation of NH(3) constitutes a futile cycle. We hypothesize that the regulatory protein GlnK is required to fine-tune the active transport of ammonium in order to limit futile cycling whilst enabling an intracellular ammonium level sufficient for the cell's nitrogen requirements.

Modelling nitrogen assimilation of Escherichia coli at low ammonium concentration.

Modelling is an important methodology in systems biology research. In this paper, we presented a kinetic model for the complex ammonium assimilation regulation system of Escherichia coli. Based on a previously published model, the new model included AmtB mediated ammonium transport and AmtB regulation by GlnK. Protein concentrations and several parameter values were determined or refined based on new experimental data. Steady state analysis of the model showed that the expression of AmtB increased the ammonium assimilation rate 4-5-fold at external ammonium concentrations as low as 5muM. Model analysis also suggested that AmtB and GS levels were coupled to maximize the assimilation flux and to avoid a possible negative ammonia diffusion flux. In addition, model simulation of the short term dynamic response to increased external ammonium concentrations implied that the maximal rate for GlnB/GlnK uridylylation/deuridylylation might be higher for a quick response to environmental changes.

Introduction to systems biology.

The developments in the molecular biosciences have made possible a shift to combined molecular and system-level approaches to biological research under the name of Systems Biology. It integrates many types of molecular knowledge, which can best be achieved by the synergistic use of models and experimental data. Many different types of modeling approaches are useful depending on the amount and quality of the molecular data available and the purpose of the model. Analysis of such models and the structure of molecular networks have led to the discovery of principles of cell functioning overarching single species. Two main approaches of systems biology can be distinguished. Top-down systems biology is a method to characterize cells using system-wide data originating from the Omics in combination with modeling. Those models are often phenomenological but serve to discover new insights into the molecular network under study. Bottom-up systems biology does not start with data but with a detailed model of a molecular network on the basis of its molecular properties. In this approach, molecular networks can be quantitatively studied leading to predictive models that can be applied in drug design and optimization of product formation in bioengineering. In this chapter we introduce analysis of molecular network by use of models, the two approaches to systems biology, and we shall discuss a number of examples of recent successes in systems biology.

The multifarious short-term regulation of ammonium assimilation of Escherichia coli: dissection using an in silico replica.

Ammonium assimilation in Escherichia coli is regulated through multiple mechanisms (metabolic, signal transduction leading to covalent modification, transcription, and translation), which (in-)directly affect the activities of its two ammonium-assimilating enzymes, i.e. glutamine synthetase (GS) and glutamate dehydrogenase (GDH). Much is known about the kinetic properties of the components of the regulatory network that these enzymes are part of, but the ways in which, and the extents to which the network leads to subtle and quasi-intelligent regulation are unappreciated. To determine whether our present knowledge of the interactions between and the kinetic properties of the components of this network is complete - to the extent that when integrated in a kinetic model it suffices to calculate observed physiological behaviour - we now construct a kinetic model of this network, based on all of the kinetic data on the components that is available in the literature. We use this model to analyse regulation of ammonium assimilation at various carbon statuses for cells that have adapted to low and high ammonium concentrations. We show how a sudden increase in ammonium availability brings about a rapid redirection of the ammonium assimilation flux from GS/glutamate synthase (GOGAT) to GDH. The extent of redistribution depends on the nitrogen and carbon status of the cell. We develop a method to quantify the relative importance of the various regulators in the network. We find the importance is shared among regulators. We confirm that the adenylylation state of GS is the major regulator but that a total of 40% of the regulation is mediated by ADP (22%), glutamate (10%), glutamine (7%) and ATP (1%). The total steady-state ammonium assimilation flux is remarkably robust against changes in the ammonium concentration, but the fluxes through GS and GDH are completely nonrobust. Gene expression of GOGAT above a threshold value makes expression of GS under ammonium-limited conditions, and of GDH under glucose-limited conditions, sufficient for ammonium assimilation.

Oxygen protection of nitrogen fixation in free-living Azorhizobium caulinodans: the role of cytochrome aa3.

The growth properties of Azorhizobium caulinodans wild-type and a cytochrome aa3 mutant strain, both growing with N2 as N source at fixed dissolved partial oxygen pressures in the range 0.5--4.0 kPa, were studied by making use of continuous cultures (chemostats and pH-auxostats) and transient cultures. In succinate-limited chemostats, the wild-type exhibited a higher growth yield than the aa3 mutant at every dissolved oxygen tension tested, indicating activity of cytochrome aa3 in this entire oxygen regime. The growth yield of both the wild-type and the aa3 mutant declined when the dissolved oxygen tension was raised. In contrast, for growth on ammonia at the same dilution rate, the wild-type showed an increase in growth yield with increasing dissolved oxygen tension, whereas the growth yield of the aa3 mutant remained constant. The transient changes in growth properties observed in chemostat cultures after pulsing with succinate pointed to a negative effect of oxygen on the maximum specific growth rate. This was studied further in steady-state pH-auxostat cultures. The specific growth rate of both strains decreased with increasing dissolved oxygen tension. The less steep decline in growth rate of the wild-type compared to the aa3 mutant confirmed that cytochrome aa3 is active in the wild-type. Again, the growth yield of both strains decreased with the dissolved oxygen tension, but in contrast to the results obtained with chemostats, no difference in growth yield was observed between wild-type and mutant at any oxygen tension. In either type of continuous culture a decrease in the overall P/O ratio with increasing dissolved oxygen tension is improbable for the wild-type, and even more so for the aa3 mutant. Therefore, the adverse effects of oxygen on the growth of A. caulinodans are not readily explained by respiratory protection; alternatively, it is proposed that the catalytic oxidation of nitrogen-fixation-specific redox enzymes by oxygen (auto-protection) enables the bacterium to deal with intracellular oxygen at the expense of reducing equivalents and free energy. To compensate for the loss of free energy, respiration increases and an active cytochrome aa3 contributes to this by keeping the P/O ratio high.