A guide to gene regulatory network inference for obtaining predictive solutions: Underlying assumptions and fundamental biological and data constraints.

The study of biological systems at a system level has become a reality due to the increasing powerful computational approaches able to handle increasingly larger datasets. Uncovering the dynamic nature of gene regulatory networks in order to attain a system level understanding and improve the predictive power of biological models is an important research field in systems biology. The task itself presents several challenges, since the problem is of combinatorial nature and highly depends on several biological constraints and also the intended application. Given the intrinsic interdisciplinary nature of gene regulatory network inference, we present a review on the currently available approaches, their challenges and limitations. We propose guidelines to select the most appropriate method considering the underlying assumptions and fundamental biological and data constraints.

Multiscale modelling approaches for assessing cosmetic ingredients safety.

The European Union's ban on animal testing for cosmetic ingredients and products has generated a strong momentum for the development of in silico and in vitro alternative methods. One of the focus of the COSMOS project was ab initio prediction of kinetics and toxic effects through multiscale pharmacokinetic modeling and in vitro data integration. In our experience, mathematical or computer modeling and in vitro experiments are complementary. We present here a summary of the main models and results obtained within the framework of the project on these topics. A first section presents our work at the organelle and cellular level. We then go toward modeling cell levels effects (monitored continuously), multiscale physiologically based pharmacokinetic and effect models, and route to route extrapolation. We follow with a short presentation of the automated KNIME workflows developed for dissemination and easy use of the models. We end with a discussion of two challenges to the field: our limited ability to deal with massive data and complex computations.

A hybrid approach identifies metabolic signatures of high-producers for chinese hamster ovary clone selection and process optimization.

In-depth characterization of high-producer cell lines and bioprocesses is vital to ensure robust and consistent production of recombinant therapeutic proteins in high quantity and quality for clinical applications. This requires applying appropriate methods during bioprocess development to enable meaningful characterization of CHO clones and processes. Here, we present a novel hybrid approach for supporting comprehensive characterization of metabolic clone performance. The approach combines metabolite profiling with multivariate data analysis and fluxomics to enable a data-driven mechanistic analysis of key metabolic traits associated with desired cell phenotypes. We applied the methodology to quantify and compare metabolic performance in a set of 10 recombinant CHO-K1 producer clones and a host cell line. The comprehensive characterization enabled us to derive an extended set of clone performance criteria that not only captured growth and product formation, but also incorporated information on intracellular clone physiology and on metabolic changes during the process. These criteria served to establish a quantitative clone ranking and allowed us to identify metabolic differences between high-producing CHO-K1 clones yielding comparably high product titers. Through multivariate data analysis of the combined metabolite and flux data we uncovered common metabolic traits characteristic of high-producer clones in the screening setup. This included high intracellular rates of glutamine synthesis, low cysteine uptake, reduced excretion of aspartate and glutamate, and low intracellular degradation rates of branched-chain amino acids and of histidine. Finally, the above approach was integrated into a workflow that enables standardized high-content selection of CHO producer clones in a high-throughput fashion. In conclusion, the combination of quantitative metabolite profiling, multivariate data analysis, and mechanistic network model simulations can identify metabolic traits characteristic of high-performance clones and enables informed decisions on which clones provide a good match for a particular process platform. The proposed approach also provides a mechanistic link between observed clone phenotype, process setup, and feeding regimes, and thereby offers concrete starting points for subsequent process optimization. Biotechnol. Bioeng. 2016;113: 2005-2019. © 2016 Wiley Periodicals, Inc.

Metabolic engineering with multi-objective optimization of kinetic models.

Kinetic models have a great potential for metabolic engineering applications. They can be used for testing which genetic and regulatory modifications can increase the production of metabolites of interest, while simultaneously monitoring other key functions of the host organism. This work presents a methodology for increasing productivity in biotechnological processes exploiting dynamic models. It uses multi-objective dynamic optimization to identify the combination of targets (enzymatic modifications) and the degree of up- or down-regulation that must be performed in order to optimize a set of pre-defined performance metrics subject to process constraints. The capabilities of the approach are demonstrated on a realistic and computationally challenging application: a large-scale metabolic model of Chinese Hamster Ovary cells (CHO), which are used for antibody production in a fed-batch process. The proposed methodology manages to provide a sustained and robust growth in CHO cells, increasing productivity while simultaneously increasing biomass production, product titer, and keeping the concentrations of lactate and ammonia at low values. The approach presented here can be used for optimizing metabolic models by finding the best combination of targets and their optimal level of up/down-regulation. Furthermore, it can accommodate additional trade-offs and constraints with great flexibility.

In Silico Modeling for the Prediction of Dose and Pathway-Related Adverse Effects in Humans From In Vitro Repeated-Dose Studies.

Long-term repeated-dose toxicity is mainly assessed in animals despite poor concordance of animal data with human toxicity. Nowadays advanced human in vitro systems, eg, metabolically competent HepaRG cells, are used for toxicity screening. Extrapolation of in vitro toxicity to in vivo effects is possible by reverse dosimetry using pharmacokinetic modeling. We assessed long-term repeated-dose toxicity of bosentan and valproic acid (VPA) in HepaRG cells under serum-free conditions. Upon 28-day exposure, the EC50 values for bosentan and VPA decreased by 21- and 33-fold, respectively. Using EC(10) as lowest threshold of toxicity in vitro, we estimated the oral equivalent doses for both test compounds using a simplified pharmacokinetic model for the extrapolation of in vitro toxicity to in vivo effect. The model predicts that bosentan is safe at the considered dose under the assumed conditions upon 4 weeks exposure. For VPA, hepatotoxicity is predicted for 4% and 47% of the virtual population at the maximum recommended daily dose after 3 and 4 weeks of exposure, respectively. We also investigated the changes in the central carbon metabolism of HepaRG cells exposed to orally bioavailable concentrations of both drugs. These concentrations are below the 28-day EC(10) and induce significant changes especially in glucose metabolism and urea production. These metabolic changes may have a pronounced impact in susceptible patients such as those with compromised liver function and urea cycle deficiency leading to idiosyncratic toxicity. We show that the combination of modeling based on in vitro repeated-dose data and metabolic changes allows the prediction of human relevant in vivo toxicity with mechanistic insights.

BioPreDyn-bench: a suite of benchmark problems for dynamic modelling in systems biology.

BACKGROUND: Dynamic modelling is one of the cornerstones of systems biology. Many research efforts are currently being invested in the development and exploitation of large-scale kinetic models. The associated problems of parameter estimation (model calibration) and optimal experimental design are particularly challenging. The community has already developed many methods and software packages which aim to facilitate these tasks. However, there is a lack of suitable benchmark problems which allow a fair and systematic evaluation and comparison of these contributions. RESULTS: Here we present BioPreDyn-bench, a set of challenging parameter estimation problems which aspire to serve as reference test cases in this area. This set comprises six problems including medium and large-scale kinetic models of the bacterium E. coli, baker's yeast S. cerevisiae, the vinegar fly D. melanogaster, Chinese Hamster Ovary cells, and a generic signal transduction network. The level of description includes metabolism, transcription, signal transduction, and development. For each problem we provide (i) a basic description and formulation, (ii) implementations ready-to-run in several formats, (iii) computational results obtained with specific solvers, (iv) a basic analysis and interpretation. CONCLUSIONS: This suite of benchmark problems can be readily used to evaluate and compare parameter estimation methods. Further, it can also be used to build test problems for sensitivity and identifiability analysis, model reduction and optimal experimental design methods. The suite, including codes and documentation, can be freely downloaded from the BioPreDyn-bench website, https://sites.google.com/site/biopredynbenchmarks/ .

A consensus approach for estimating the predictive accuracy of dynamic models in biology.

Mathematical models that predict the complex dynamic behaviour of cellular networks are fundamental in systems biology, and provide an important basis for biomedical and biotechnological applications. However, obtaining reliable predictions from large-scale dynamic models is commonly a challenging task due to lack of identifiability. The present work addresses this challenge by presenting a methodology for obtaining high-confidence predictions from dynamic models using time-series data. First, to preserve the complex behaviour of the network while reducing the number of estimated parameters, model parameters are combined in sets of meta-parameters, which are obtained from correlations between biochemical reaction rates and between concentrations of the chemical species. Next, an ensemble of models with different parameterizations is constructed and calibrated. Finally, the ensemble is used for assessing the reliability of model predictions by defining a measure of convergence of model outputs (consensus) that is used as an indicator of confidence. We report results of computational tests carried out on a metabolic model of Chinese Hamster Ovary (CHO) cells, which are used for recombinant protein production. Using noisy simulated data, we find that the aggregated ensemble predictions are on average more accurate than the predictions of individual ensemble models. Furthermore, ensemble predictions with high consensus are statistically more accurate than ensemble predictions with large variance. The procedure provides quantitative estimates of the confidence in model predictions and enables the analysis of sufficiently complex networks as required for practical applications.

Bridging the layers: towards integration of signal transduction, regulation and metabolism into mathematical models.

Mathematical modelling is increasingly becoming an indispensable tool for the study of cellular processes, allowing their analysis in a systematic and comprehensive manner. In the vast majority of the cases, models focus on specific subsystems, and in particular describe either metabolism, gene expression or signal transduction. Integrated models that are able to span and interconnect these layers are, by contrast, rare as their construction and analysis face multiple challenges. Such methods, however, would represent extremely useful tools to understand cell behaviour, with application in distinct fields of biological and medical research. In particular, they could be useful tools to study genotype-phenotype mappings, and the way they are affected by specific conditions or perturbations. Here, we review existing computational approaches that integrate signalling, gene regulation and/or metabolism. We describe existing challenges, available methods and point at potentially useful strategies.

Quantitative Evaluation and Prediction of Drug Effects and Toxicological Risk Using Mechanistic Multiscale Models.

Integrating in vitro and in silico approaches has great potential for reducing experimental effort and delivering know-how and intellectual property in drug development. Here, we focus on a possible framework for multiscale modeling in pharmaceutical drug development. Looking at the modeling frameworks at different scales, it is obvious that choosing the proper level of complexity and abstraction is not a trivial task. At cellular level, we consider that the application of validated kinetic models of cellular toxicity mechanisms of drugs is particularly important for deriving valid predictions. These kinetic models can be applied for integrating inter-individual differences, e.g. obtained from data measured in surgical liver samples, into predictions of drug effects. Challenges identified include (i) the development of sufficiently detailed, structured organ models, (ii) definition of multiscale models that can be efficiently handled by available super-computing facilities, and (iii) availability of validated cell-type and organ-specific kinetic metabolic models. Multiscale models can streamline drug development by facilitating the design of experiments and trials, by providing and testing hypotheses, and by reducing time and costs due to less experiments and improved decision-making. In this review, we discuss the required pieces, possibilities, and challenges in multiscale modeling for the prediction of drug effects.

Central energy metabolism remains robust in acute steatotic hepatocytes challenged by a high free fatty acid load.

Overnutrition is one of the major causes of non-alcoholic fatty liver disease (NAFLD). NAFLD is characterized by an accumulation of lipids (triglycerides) in hepatocytes and is often accompanied by high plasma levels of free fatty acids (FFA). In this study, we compared the energy metabolism in acute steatotic and non-steatotic primary mouse hepatocytes. Acute steatosis was induced by pre-incubation with high concentrations of oleate and palmitate. Labeling experiments were conducted using [U-(13)C(5),U-(15)N(2)] glutamine. Metabolite concentrations and mass isotopomer distributions of intracellular metabolites were measured and applied for metabolic flux estimation using transient 13C metabolic flux analysis. FFAs were efficiently taken up and almost completely incorporated into triglycerides (TAGs). In spite of high FFA uptake rates and the high synthesis rate of TAGs, central energy metabolism was not significantly changed in acute steatotic cells. Fatty acid ß-oxidation does not significantly contribute to the detoxification of FFAs under the applied conditions.

A multi-scale modeling framework for individualized, spatiotemporal prediction of drug effects and toxicological risk.

In this study, we focus on a novel multi-scale modeling approach for spatiotemporal prediction of the distribution of substances and resulting hepatotoxicity by combining cellular models, a 2D liver model, and whole body model. As a case study, we focused on predicting human hepatotoxicity upon treatment with acetaminophen based on in vitro toxicity data and potential inter-individual variability in gene expression and enzyme activities. By aggregating mechanistic, genome-based in silico cells to a novel 2D liver model and eventually to a whole body model, we predicted pharmacokinetic properties, metabolism, and the onset of hepatotoxicity in an in silico patient. Depending on the concentration of acetaminophen in the liver and the accumulation of toxic metabolites, cell integrity in the liver as a function of space and time as well as changes in the elimination rate of substances were estimated. We show that the variations in elimination rates also influence the distribution of acetaminophen and its metabolites in the whole body. Our results are in agreement with experimental results. What is more, the integrated model also predicted variations in drug toxicity depending on alterations of metabolic enzyme activities. Variations in enzyme activity, in turn, reflect genetic characteristics or diseases of individuals. In conclusion, this framework presents an important basis for efficiently integrating inter-individual variability data into models, paving the way for personalized or stratified predictions of drug toxicity and efficacy.

Dynamics and control of the central carbon metabolism in hepatoma cells.

BACKGROUND: The liver plays a major role in metabolism and performs a number of vital functions in the body. Therefore, the determination of hepatic metabolite dynamics and the analysis of the control of the respective biochemical pathways are of great pharmacological and medical importance. Extra- and intracellular time-series data from stimulus-response experiments are gaining in importance in the identification of in vivo metabolite dynamics, while dynamic network models are excellent tools for analyzing complex metabolic control patterns. This is the first study that has been undertaken on the data-driven identification of a dynamic liver central carbon metabolism model and its application in the analysis of the distribution of metabolic control in hepatoma cells. RESULTS: Dynamic metabolite data were collected from HepG2 cells after they had been deprived of extracellular glucose. The concentration of 25 extra- and intracellular intermediates was quantified using HPLC, LC-MS-MS, and GC-MS. The in silico metabolite dynamics were in accordance with the experimental data. The central carbon metabolism of hepatomas was further analyzed with a particular focus on the control of metabolite concentrations and metabolic fluxes. It was observed that the enzyme glucose-6-phosphate dehydrogenase exerted substantial negative control over the glycolytic flux, whereas oxidative phosphorylation had a significant positive control. The control over the rate of NADPH consumption was found to be shared between the NADPH-demand itself (0.65) and the NADPH supply (0.38). CONCLUSIONS: Based on time-series data, a dynamic central carbon metabolism model was developed for the investigation of new and complex metabolic control patterns in hepatoma cells. The control patterns found support the hypotheses that the glucose-6-phosphate dehydrogenase and the Warburg effect are promising targets for tumor treatment. The systems-oriented identification of metabolite dynamics is a first step towards the genome-based assessment of potential risks posed by nutrients and drugs.

Quantification of statin effects on hepatic cholesterol synthesis by transient (13)C-flux analysis.

The present work is the first to deal with the determination of cholesterol synthesis rates in primary rat hepatocytes using transient (13)C-flux analysis. The effects of statins on cholesterol biosynthesis and central carbon fluxes were quantified at a therapeutic concentration of 50 nM atorvastatin using carbon-labeled glutamine. The flux through the cholesterol pathway decreased from 0.27 to 0.08 mmol/l(cv)h in response to the administration of the hypolipidemic drug. Isotopic steady state was reached within 4h in the central carbon metabolism but not in the cholesterol pathway, regardless of whether atorvastatin was administered or not. Marked channeling was observed for the symmetrical tricarboxylic acid cycle intermediates, succinate and fumarate. Non-stationary (13)C-based flux identification delivers both intracellular fluxes and intermediate levels, which was for the first time utilized for investigating systems-level effects of the administered drug by quantifying the flux control of the 3-hydroxy-3-methylglutaryl-coenzyme A reductase.

Identification of metabolic fluxes in hepatic cells from transient 13C-labeling experiments: Part II. Flux estimation.

This contribution addresses the identification of metabolic fluxes and metabolite concentrations in mammalian cells from transient (13)C-labeling experiments. Whilst part I describes experimental set-up and acquisition of required metabolite and (13)C-labeling data, part II focuses on setting up network models and the estimation of intracellular fluxes. Metabolic fluxes were determined in glycolysis, pentose-phosphate pathway (PPP), and citric acid cycle (TCA) in a hepatoma cell line grown in aerobic batch cultures. In glycolytic and PPP metabolite pools isotopic stationarity was observed within 30 min, whereas in the TCA cycle the labeling redistribution did not reach isotopic steady state even within 180 min. In silico labeling dynamics were in accordance with in vivo (13)C-labeling data. Split ratio between glycolysis and PPP was 57%:43%; intracellular glucose concentration was estimated at 101.6 nmol per 10(6) cells. In contrast to isotopic stationary (13)C-flux analysis, transient (13)C-flux analysis can also be applied to industrially relevant mammalian cell fed-batch and batch cultures.

Identification of metabolic fluxes in hepatic cells from transient 13C-labeling experiments: Part I. Experimental observations.

An experimental set-up for acquiring metabolite and transient (13)C-labeling data in mammalian cells is presented. An efficient sampling procedure was established for hepatic cells cultured in six-well plates as a monolayer attached to collagen, which allowed simultaneous quenching of metabolism and extraction of the intracellular intermediates of interest. Extracellular concentrations of glucose, amino acids, lactate, pyruvate, and urea were determined by GC-MS procedures and were used for estimation of metabolic uptake and excretion rates. Sensitive LC-MS and GC-MS methods were used to quantify the intracellular intermediates of tricarboxylic acid cycle, glycolysis, and pentose phosphate pathway and for the determination of isotopomer fractions of the respective metabolites. Mass isotopomer fractions were determined in a transient (13)C-labeling experiment using (13)C-labeled glucose as substrate. The absolute amounts of intracellular metabolites were obtained from a non-labeled experiment carried out in exactly the same way as the (13)C-labeling experiment, except that the media contained naturally labeled glucose only. Estimation of intracellular metabolic fluxes from the presented data is addressed in part II of this contribution.

Metabolic flux analysis in Escherichia coli by integrating isotopic dynamic and isotopic stationary 13C labeling data.

The novel concept of isotopic dynamic 13C metabolic flux analysis (ID-13C MFA) enables integrated analysis of isotopomer data from isotopic transient and/or isotopic stationary phase of a 13C labeling experiment, short-time experiments, and an extended range of applications of 13C MFA. In the presented work, an experimental and computational framework consisting of short-time 13C labeling, an integrated rapid sampling procedure, a LC-MS analytical method, numerical integration of the system of isotopomer differential equations, and estimation of metabolic fluxes was developed and applied to determine intracellular fluxes in glycolysis, pentose phosphate pathway (PPP), and citric acid cycle (TCA) in Escherichia coli grown in aerobic, glucose-limited chemostat culture at a dilution rate of D = 0.10 h(-1). Intracellular steady state concentrations were quantified for 12 metabolic intermediates. A total of 90 LC-MS mass isotopomers were quantified at sampling times t = 0, 91, 226, 346, 589 s and at isotopic stationary conditions. Isotopic stationarity was reached within 10 min in glycolytic and PPP metabolites. Consistent flux solutions were obtained by ID-13C MFA using isotopic dynamic and isotopic stationary 13C labeling data and by isotopic stationary 13C MFA (IS-13C MFA) using solely isotopic stationary data. It is demonstrated that integration of dynamic 13C labeling data increases the sensitivity of flux estimation, particularly at the glucose-6-phosphate branch point. The identified split ratio between glycolysis and PPP was 55%:44%. These results were confirmed by IS-13C MFA additionally using labeling data in proteinogenic amino acids (GC-MS) obtained after 5 h from sampled biomass.

Model of central and trimethylammonium metabolism for optimizing L-carnitine production by E. coli.

The application of metabolic engineering principles to the rational design of microbial production processes crucially depends on the ability to make quantitative descriptions of the systemic ability of the central carbon metabolism to redirect fluxes to the product-forming pathways. The aim of this work was to further our understanding of the steps controlling the biotransformation of trimethylammonium compounds into L-carnitine by Escherichia coli. Despite the importance of L-carnitine production processes, development of a model of the central carbon metabolism linked to the secondary carnitine metabolism of E. coli has been severely hampered by the lack of stoichiometric information on the metabolic reactions taking place in the carnitine metabolism. Here we present the design and experimental validation of a model which, for the first time, links the carnitine metabolism with the reactions of glycolysis, the tricarboxylic acid cycle and the pentose-phosphate pathway. The results demonstrate a need for a high production rate of ATP to be devoted to the biotransformation process. The results demonstrate that ATP is used up in a futile cycle, since both trimethylammonium compound carriers CaiT and ProU operate simultaneously. To improve the biotransformation process, resting processes as well as CaiT or ProU knock out mutants would yield a more efficient system for producing L-carnitine from crotonobetaine or D-carnitine.

Metabolic design based on a coupled gene expression-metabolic network model of tryptophan production in Escherichia coli.

The presumably high potential of a holistic design approach for complex biochemical reaction networks is exemplified here for the network of tryptophan biosynthesis from glucose, a system whose components have been investigated thoroughly before. A dynamic model that combines the behavior of the trp operon gene expression with the metabolic network of central carbon metabolism and tryptophan biosynthesis is investigated. This model is analyzed in terms of metabolic fluxes, metabolic control, and nonlinear optimization. We compare two models for a wild-type strain and another model for a tryptophan producer. An integrated optimization of the whole network leads to a significant increase in tryptophan production rate for all systems under study. This enhancement is well above the increase that can be achieved by an optimization of subsystems. A constant ratio of control coefficients on tryptophan synthesis rate has been identified for the models regarding or disregarding trp operon expression. Although we found some examples where flux control coefficients even contradict the trends of enzyme activity changes in an optimized profile, flux control can be used as an indication for enzymes that have to be taken into account in optimization.

Optimal re-design of primary metabolism in Escherichia coli using linlog kinetics.

This paper examines the validity of the linlog approach, which was recently developed in our laboratory, by comparison of two different kinetic models for the metabolic network of Escherichia coli. The first model is a complete mechanistic model; the second is an approximative model in which linlog kinetics are applied. The parameters of the linlog model (elasticities) are derived from the mechanistic model. Three different optimization cases are examined. In all cases, the objective is to calculate the enzyme levels that maximize a certain flux while keeping the total amount of enzyme constant and preventing large changes of metabolite concentrations. For an average variation of metabolite levels of 10% and individual changes of a factor 2, the predicted enzyme levels, metabolite concentrations and fluxes of both models are highly similar. This similarity holds for changes in enzyme level of a factor 4-6 and for changes in fluxes up to a factor 6. In all three cases, the predicted optimal enzyme levels could neither have been found by intuition-based approaches, nor on basis of flux control coefficients. This demonstrates that kinetic models are essential tools in Metabolic Engineering. In this respect, the linlog approach is a valuable extension of MCA, since it allows construction of kinetic models, based on MCA parameters, that can be used for constrained optimization problems and are valid for large changes of metabolite and enzyme levels.