Enzymatic synthesis of amoxicillin: avoiding limitations of the mechanistic approach for reaction kinetics

A recurrent doubt that occurs to the enzyme‐kinetics modeler is, When should I stop adding parameters to my mechanistic model in order to fit a non‐conventional behavior? This problem becomes more and more involving when the complexity of the reaction network increases. This work intends to show how the use of artificial neural networks may circumvent the need of including an overwhelming number of parameters in the rate equations obtained through the classical, mechanistic approach. We focus on the synthesis of amoxicillin by the reaction of p‐OH‐phenylglycine methyl ester and 6‐aminopenicillanic acid, catalyzed by penicillin G acylase immobilized on glyoxyl‐agarose, at 25°C and pH 6.5. The reaction was carried on a batch reactor. Three kinetic models of this system were compared: a mechanistic, a semi‐empiric, and a hybrid–neural network (NN). A semi‐empiric, simplified model with a reasonable number of parameters was initially built‐up. It was able to portray many typical process conditions. However, it either underestimated or overestimated the rate of synthesis of amoxicillin when substrates' concentrations were low. A more complex, full‐scale mechanistic model that could span all operational conditions was intractable for all practical purposes. Finally, a hybrid model, that coupled artificial neural networks (NN) to mass‐balance equations was established, that succeeded in representing all situations of interest. Particularly, the NN could predict with accuracy reaction rates for conditions where the semi‐empiric model failed, namely, at low substrate concentrations, a situation that would occur, for instance, at the end of a fed‐batch industrial process. © 2002 Wiley Periodicals, Inc. Biotechnol Bioeng 80: 622–631, 2002.     

One-pot, two-step enzymatic synthesis of amoxicillin by complexing with Zn2+

A one-pot, two-step enzymatic synthesis of amoxicillin from penicillin G, using penicillin acylase, is presented. Immobilized penicillin acylase from Kluyvera citrophila was selected as the biocatalyst for its good pH stability and selectivity. Hydrolysis of penicillin G and synthesis of amoxicillin from the 6-aminopenicillanic acid formed and d-p-hydroxyphenylglycine methyl ester were catalyzed in situ by a single enzyme. Zinc ions can react with amoxicillin to form complexes, and the yield of 76.5% was obtained after optimization. In the combined one-pot synthesis process, zinc sulfate was added to remove produced amoxicillin as complex for shifting the equilibrium to the product in the second step. By controlling the conditions in two separated steps, the conversion of the first and second step was 93.8% and 76.2%, respectively. With one-pot continuous procedure, a 71.5% amoxicillin yield using penicillin G was obtained.     

Ionization constants and solubility of compounds involved in enzymatic synthesis of aminopenicillins and aminocephalosporins

The article deals with experimental determination of ionization constants and solubility for the compounds (target products, initial β-lactams, acylating agents and by-products) involved in enzymatic synthesis of some therapeutically used aminopenicillins and aminocephalosporins, namely ampicillin, amoxicillin, cephalexin, cephadroxil, cephaloglycin, cefaclor, cefprozil, cefatrizine. Methodology of investigations and the evaluation of experimental data for the determination of ionization constants and solubility of the different type electrolytes are presented. Applications of the methods based on acid–base potentiometric titration and on determination of solubility–pH dependence of assayed substances are discussed. The original data on ionization constants and solubility of amoxicillin, cefprozil, cefatrizine, cephadroxil and initial β-lactams for production of cefaclor, cefprozil and cefatrizine, as well as solubility of by-product d-(−)-p-hydroxyphenylglycine are presented. Experimentally determined parameters and constants available in the literature for all abovementioned aminopenicillins and aminocephalosporins are collected. These data might be used for choice of the conditions of both processes: the enzymatic synthesis and the isolation of the product from reaction mixture.     

Estimation of the number of extreme pathways for metabolic networks

Background  

The set of extreme pathways (ExPa), {p _i }, defines the convex basis vectors used for the mathematical characterization of the null space of the stoichiometric matrix for biochemical reaction networks. ExPa analysis has been used for a number of studies to determine properties of metabolic networks as well as to obtain insight into their physiological and functional states in silico. However, the number of ExPas, p = |{p _i }|, grows with the size and complexity of the network being studied, and this poses a computational challenge. For this study, we investigated the relationship between the number of extreme pathways and simple network properties.     

Designing fermentation media: A comparison of neural networks to factorial design

Summary Neural networks were compared to factorial experiments as techniques for designing fermentation media. The production of intracellular oil byRhodotorula gracilis (Rhodosporidium toruloides) was used as a model system. Investigating three factors, the molasses, ammonium nitrate and yeast extract concentrations, each at three concentrations, 27 experiments were required for a complete factorial expriment. In contrast, neural networks could be trained on 10 experiments and predict the test experiments with reasonable accuracy. This represents a 63% saving in the number of experiments that need to be conducted. Thus neural networks are a useful tool in developing fermentation medium.     

Construction and co-expression of polycistronic plasmid encoding d-hydantoinase and d-carbamoylase for the production of d-amino acids

d-Hydantoinase and d-carbamoylase genes from Agrobacterium radiobacter TH572 were cloned by polymerase chain reaction (PCR). The plasmid pUCCH3 with a polycistronic structure that is controlled by the native hydantoinase promoter was constructed to co-express the two genes and transformed into Escherichia coli strain JM105. To obtain the highest level of expression of the d-carbamoylase and avoid intermediate accumulation, the d-carbamoylase gene was cloned closer to the promoter and the RBS region in the upstream of it was optimized. This resulted in high active expression of soluble d-hydantoinase and d-carbamoylase that is obtained without any inducer. Thus, by the constitutive recombinant JM105/pUCCH3, d-p-hydroxyphenylglycine (d-HPG) was obtained directly with 95.2% production yield and 96.3% conversion yield.     

Robust artificial intelligence tool for automatic start-up of the supplementary medium feeding in recombinant E. coli cultivations

One of the most important events in fed-batch fermentations is the definition of the moment to start the feeding. This paper presents a methodology for a rational selection of the architecture of an artificial intelligence (AI) system, based on a neural network committee (NNC), which identifies the end of the batch phase. The AI system was successfully used during high cell density cultivations of recombinant Escherichia coli. The AI algorithm was validated for different systems, expressing three antigens to be used in human and animal vaccines: fragments of surface proteins of Streptococcus pneumoniae (PspA), clades 1 and 3, and of Erysipelothrix rhusiopathiae (SpaA). Standard feed-forward neural networks (NNs), with a single hidden layer, were the basis for the NNC. The NN architecture with best performance had the following inputs: stirrer speed, inlet air, and oxygen flow rates, carbon dioxide evolution rate, and CO₂ molar fraction in the exhaust gas.     

Analysis of an optimal hidden Markov model for secondary structure prediction

Background  

Secondary structure prediction is a useful first step toward 3D structure prediction. A number of successful secondary structure prediction methods use neural networks, but unfortunately, neural networks are not intuitively interpretable. On the contrary, hidden Markov models are graphical interpretable models. Moreover, they have been successfully used in many bioinformatic applications. Because they offer a strong statistical background and allow model interpretation, we propose a method based on hidden Markov models.     

Some structural determinants of Pavlovian conditioning in artificial neural networks

This paper investigates the possible role of neuroanatomical features in Pavlovian conditioning, via computer simulations with layered, feedforward artificial neural networks. The networks’ structure and functioning are described by a strongly bottom-up model that takes into account the roles of hippocampal and dopaminergic systems in conditioning. Neuroanatomical features were simulated as generic structural or architectural features of neural networks. We focused on the number of units per hidden layer and connectivity. The effect of the number of units per hidden layer was investigated through simulations of resistance to extinction in fully connected networks. Large networks were more resistant to extinction than small networks, a stochastic effect of the asynchronous random procedure used in the simulator to update activations and weights. These networks did not simulate second-order conditioning because weight competition prevented conditioning to a stimulus after conditioning to another. Partially connected networks simulated second-order conditioning and devaluation of the second-order stimulus after extinction of a similar first-order stimulus. Similar stimuli were simulated as nonorthogonal input-vectors.     

Optimum ratio of d-carbamoylase to d-hydantoinase for maximizing d-p-hydroxyphenylglycine productivity

By use of PCR, the genes encoding d-carbamoylase and d-hydantoinase, responsible for the synthesis of d-p-hydroxyphenylglycine (d-HPG), were cloned separately from Agrobacterium radiobacter NRRL B11291 into Escherichia coli. d-HPG production was carried out with a mixed population consisting of various proportions of two recombinants producing individual enzymes. As a result, with a total amount of 192 mg dry cell, the optimal productivity of d-HPG (ca. 4.5 g h^–1 l^–1) was obtained when the ratio of d-carbamoylase to d-hydantoinase ranged between 1 and 2.     

General description of Lake Peipsi-Pihkva

Neural networks and multiple linear regression models of the abundance of brown trout (Salmo trutta L.) on the mesohabitat scale were developed from combinations of physical habitat variables in 220 channel morphodynamic units (pools, riffles, runs, etc.) of 11 different streams in the central Pyrenean mountains. For all the 220 morphodynamic units, the determination coefficients obtained between the estimated and observed values of density or biomass were significantly higher for the neural network (r ² adjusted= 0.93 and r ² adjusted=0.92 (p<0.01) for biomass and density respectively with the neural network, against r ² adjusted=0.69 (p<0.01) and r ² adjusted = 0.54 (p<0.01) with multiple linear regression). Validation of the multivariate models and learning of the neural network developed from 165 randomly chosen channel morphodynamic units, was tested on the 55 other channel morphodynamic units. This showed that the biomass and density estimated by both methods were significantly related to the observed biomass and density. Determination coefficients were significantly higher for the neural network (r ² adjusted =0.72 (p<0.01) and 0.81 (p<0.01) for biomass and density respectively) than for the multiple regression model (r ² adjusted=0.59 and r ² adjusted=0.37 for biomass and density respectively). The present study shows the advantages of the backpropagation procedure with neural networks over multiple linear regression analysis, at least in the field of stochastic salmonid ecology.     

Modeling of the pyruvate production with Escherichia coli: comparison of mechanistic and neural networks-based models

Three different models: the unstructured mechanistic black-box model, the input–output neural network-based model and the externally recurrent neural network model were used to describe the pyruvate production process from glucose and acetate using the genetically modified Escherichia coli YYC202 ldhA::Kan strain. The experimental data were used from the recently described batch and fed-batch experiments [ Zelić B, Study of the process development for Escherichia coli-based pyruvate production. PhD Thesis, University of Zagreb, Faculty of Chemical Engineering and Technology, Zagreb, Croatia, July 2003. (In English); Zelić et al. Bioproc Biosyst Eng 26:249–258 (2004); Zelić et al. Eng Life Sci 3:299–305 (2003); Zelić et al Biotechnol Bioeng 85:638–646 (2004)]. The neural networks were built out of the experimental data obtained in the fed-batch pyruvate production experiments with the constant glucose feed rate. The model validation was performed using the experimental results obtained from the batch and fed-batch pyruvate production experiments with the constant acetate feed rate. Dynamics of the substrate and product concentration changes was estimated using two neural network-based models for biomass and pyruvate. It was shown that neural networks could be used for the modeling of complex microbial fermentation processes, even in conditions in which mechanistic unstructured models cannot be applied.     

DL-4-hydroxyphenylglycine catabolism in Pseudomonas putida MW27

Thirteen bacteria were isolated on D-4-hydroxyphenylglycine as sole carbon and energy source. Seven strains transaminated only the D-enantiomer while the other six isolates transaminated both enantiomers of 4-hydroxyphenylglycine. One of the six strains utilizing both enantiomers was characterized as a Pseudomonas putida. This strain, MW27, employed two enantioselective transaminases, to catalyze the initial step in the metabolism of DL-4-hydroxyphenylglycine. The product of the transamination, 4-hydroxyphenylglyoxylate, was further metabolized via 4-hydroxybenzaldehyde and 4-hydroxybenzoate to protocatechuate. Preliminary results indicate that both transaminases are co-ordinately synthesized together with the 4-hydroxyphenylglyoxylate decarboxylase and the NADP⁺-dependent 4-hydroxybenzaldehyde dehydrogenase.     

Enhancement of operational stability of immobilized whole cell D-hydantoinase

Summary A bacteriumPseudomonas sp. KBEL 101 isolated from soil was immobilized within polyacrylamide gel and used for the synthesis of D-p-hydroxyphenylglycine from DL-5-substituted hydantoin. The half-life of immobilized whole cell D-hydantoinase was found to be about 50 hrs. In order to increase the operational stability of immobilized whole cell D-hydantoinase, a carbon or nitrogen source was supplied with the reaction mixture in the continuous stirred tank reactor. As a sole source of carbon, glycerol was found to be most effective, and the activity of immobilized whole cell enzyme was maintained stably during 7 days when 0.1%(W/V) glycerol solution was provided. In the case of nitrogen source, supplying of 0.1%(W/V) yeast extract prolonged the half-life of immobilized whole cell D-hydantoinase to about 25 days.     

Annual Change of Primary Resistance to Clarithromycin among Helicobacter pylori Isolates from 1996 through 2008 in Japan

Background: Recent studies have shown that the combination of proton pump inhibitor, amoxicillin and clarithromycin is one of the best choices for Helicobacter pylori eradication therapy. However, increasing number of cases of H. pylori infection showing resistance to clarithromycin therapy has been reported and this is currently the main cause of eradication failure. We investigated the annual changes of the antimicrobial susceptibility to clarithromycin, amoxicillin and minocycline during a period of 12 years in Japan. Methods: This study comprised 3521 patients (mean age (SD), 55.4 (13.7) years‐old, 2467 males and 1054 females) positive for H. pylori as assessed by microaerobic bacterial culture from 1996 through 2008. All patients were previously untreated for H. pylori and were enrolled in the study to assess primary resistance to the three antibiotics. Results: The overall primary resistance to clarithromycin, amoxicillin and minocycline were 16.4%, (577/3521), 0.03% (1/3521) and 0.06% (2/3521), respectively. From1996 through 2004, the resistance rate to clarithromycin increased gradually to approximately 30% and then it remained without marked fluctuation since 2004. Analysis by gender showed a significant increase (p < .0001) in resistance rate to clarithromycin among females (217/1057, 20.6%) compared to males (360/2467, 14.6%). Analysis by age, disclosed significantly (p < .0001) higher resistance rate to clarithromycin in patients of more than 65‐years‐old compared to the younger population. Conclusions: The resistance rate of H. pylori infection to clarithromycin in Japan has increased gradually to approximately 30% from 1996 through 2004, and remained unchanged since 2004. Elderly and females were at high risk of having resistance to clarithromycin. Our results suggested that the level of clarithromycin resistance in Japan has now risen to the point where it should no longer be used as empiric therapy.     

The use of artificial neural networks to reduce data collection demands in determining spine loading: a laboratory based analysis

The extensive data requirements of three-dimensional inverse dynamics and joint modelling to estimate spinal loading prevent the implementation of these models in industry and may hinder development of advanced injury prevention standards. This work examines the potential of feed forward artificial neural networks (ANNs) as a data reduction approach and compared predictions to rigid link and EMG-assisted models. Ten males and ten females performed dynamic lifts, all approaches were applied and comparisons of predicted joint moments and joint forces were evaluated. While the ANN under- predicted peak extension moments (p = 0.0261) and joint compression (p < 0.0001), predictions of cumulative extension moments (p = 0.8293) and cumulative joint compression (p = 0.9557) were not different. Therefore, the ANNs proposed may be used to obtain estimates of cumulative exposure variables with reduced input demands; however they should not be applied to determine peak demands of a worker's exposure.     

Neural state estimation and prediction in amino acid fermentation

Summary On-line control of amino acid fermentations is complicated by uncertainties typical of biological processes and by difficulties in real-time monitoring of key process variables. Lysine is an essential amino acid in human nutrition, and also widely used in animal feed formulations. The paper discusses the construction and application of feed-forward, back-propagation neural networks as software sensors in state estimation, and multi-step ahead prediction of produced lysine and consumed sugar. Neural networks were programmed in MS Visual C++ for Windows for implementation in a PC, with a userfriendly interface for convenience and ease of operation. It is demonstrated that a well trained neural network of optimal architecture can be succesfsully used in control of amino acid production     

A neural network approach for the prediction of in vitro culture parameters for maximum biomass yields in hairy root cultures

The present study deals with ANN based prediction of culture parameters in terms of inoculum density, pH and volume of growth medium per culture vessel and sucrose content of the growth medium for Glycyrrhiza hairy root cultures. This kind of study could be a model system in exploitation of hairy root cultures for commercial production of pharmaceutical compounds using large bioreactors. The study is aimed to evaluate the efficiency of regression neural network and back propagation neural network for the prediction of optimal culture conditions for maximum hairy root biomass yield. The training data for regression and back propagation networks were primed on the basis of function approximation, where final biomass fresh weight (f_wt) was considered as a function of culture parameters. On this basis the variables in culture conditions were described in the form of equations which are for inoculum density: y=0.02x+0.04, for pH of growth medium: y=x+2.8, for sucrose content in medium: y=9.9464x+(−9.7143) and for culture medium per culture vessel: y=10x. The fresh weight values obtained from training data were considered as target values and further compared with predicted fresh weight values. The empirical data were used as testing data and further compared with values predicted from trained networks. Standard MATLAB inbuilt generalized regression network with radial basis function radbas as transfer function in layer one and purelin in layer two and back propagation having purelin as transfer function in output layer and logsig in hidden layer were used. Although in comparative assessment both the networks were found efficient for prediction of optimal culture conditions for high biomass production, more accuracy in results was seen with regression network.     

Large-scale production of the carbohydrate portion of the sialyl–Tn epitope, α-Neup5Ac-(2→6)-d-GalpNAc, through bacterial coupling

α-Neup5Ac-(2→6)-d-GalpNAc, the carbohydrate portion of sialyl–Tn epitope of the tumor-associated carbohydrate antigen, was prepared by a whole-cell reaction through the combination of recombinant Escherichia coli strains and Corynebacterium ammoniagenes. Two recombinant E. coli strains overexpressed the CMP-Neup5Ac biosynthetic genes and the α-(2→6)-sialyltransferase gene of Photobacterium damsela. C. ammoniagenes contributed to the production of UTP from orotic acid. α-Neup5Ac-(2→6)-d-GalpNAc was accumulated at 87 mM (45 g/L) after a 25-h reaction starting from orotic acid, N-acetylneuraminic acid, and 2-acetamide-2-deoxy-d-galactose.