On-line estimation of the specific growth rate based on viable biomass measurements: experimental validation

The application of model based control techniques to biotechnological processes is often hampered due to the lack of reliable on-line sensors. This problem can be tackled by the application of software sensors, in which the available hardware measurements are combined with the model equations. The resulting estimates serve as additional measurements useful for process monitoring and control. In this paper, an observer based estimator for the specific growth rate based on on-line viable biomass measurements is studied. Several fed-batch experiments with baker's yeast in a stirred tank bioreactor illustrate the design, tuning, and implementation from a practical point of view. The main contributions of this paper are to illustrate (i) the implementation and validation of the presented algorithm in real-time, (ii) the use of an advanced on-line biomass measurement, and (iii) the design and tuning of the algorithm from a practical point of view. Real-time knowledge of the specific growth rate is important because it yields information on the viability of the cells and it can be used in real-time feedback control algorithms.     

Is there a rational method to purify proteins? From expert systems to proteomics

The purification of recombinant proteins for therapeutic or analytical applications requires the use of several chromatographic steps in order to achieve a high level of purity. A range of techniques is available such as anion and cation exchange chromatography, which can be carried out at different pHs, and hence used at different steps, hydrophobic interaction chromatography, gel filtration and affinity chromatography. Evidently when confronted with a complex mixture of partially unknown proteins or a clarified cell extract there are many different routes one can take in order to choose the minimum and most efficient number of purification steps to achieve a desired level of purity (e.g. 98, 99.5 or 99.9%). In this review we will show how an initial "proteomic" characterization of the complex initial mixture of target protein and protein contaminants can be used to select the most efficient chromatographic separation steps in order to achieve a maximum level of purity with a minimum number of steps. The chosen methodology was implemented in a computer based expert system. The first algorithm developed was used to select the most efficient purification method to separate a protein from its contaminants based on the physicochemical properties of the protein product and the protein contaminants. The second algorithm developed was used to predict the number and concentration of contaminants after each separation as well as protein product purity. The successful application of the expert system approach, based on an initial proteomic characterization, to the practical cases of protein mixtures and clarified fermentation supernatant is presented and discussed. The purification strategy proposed was experimentally tested and validated with a mixture of four proteins and the experimental validation was also carried out with an "unknown" supernatant of Bacillus subtilis producing a recombinant beta-1,3-glucanase. The system was robust to errors <10% which is the range that can be found in the experimental determination of the properties in the database of product and contaminants. On the other hand, the system was sensitive both to larger variations (>20%) in the properties of the contaminant database and the protein product and to variations in one protein property (e.g. hydrophobicity).     

Protein purification using chromatography: selection of type, modelling and optimization of operating conditions

To achieve a high level of purity in the purification of recombinant proteins for therapeutic or analytical application, it is necessary to use several chromatographic steps. There is a range of techniques available including anion and cation exchange, which can be carried out at different pHs, hydrophobic interaction chromatography, gel filtration and affinity chromatography. In the case of a complex mixture of partially unknown proteins or a clarified cell extract, there are many different routes one can take in order to choose the minimum and most efficient number of purification steps to achieve a desired level of purity (e.g. 98%, 99.5% or 99.9%). This review shows how an initial 'proteomic' characterization of the complex mixture of target protein and protein contaminants can be used to select the most efficient chromatographic separation steps in order to achieve a specific level of purity with a minimum number of steps. The chosen methodology was implemented in a computer- based Expert System. Two algorithms were developed, the first algorithm was used to select the most efficient purification method to separate a protein from its contaminants based on the physicochemical properties of the protein product and the protein contaminants and the second algorithm was used to predict the number and concentration of contaminants after each separation as well as protein product purity. The application of the Expert System approach was experimentally tested and validated with a mixture of four proteins and the experimental validation was also carried out with a supernatant of Bacillus subtilis producing a recombinant beta-1,3-glucanase. Once the type of chromatography is chosen, optimization of the operating conditions is essential. Chromatographic elution curves for a three-protein mixture (alpha-lactoalbumin, ovalbumin and beta-lactoglobulin), carried out under different flow rates and ionic strength conditions, were simulated using two different mathematical models. These models were the Plate Model and the more fundamentally based Rate Model. Simulated elution curves were compared with experimental data not used for parameter identification. Deviation between experimental data and the simulated curves using the Plate Model was less than 0.0189 (absorbance units); a slightly higher deviation [0.0252 (absorbance units)] was obtained when the Rate Model was used. In order to optimize operating conditions, a cost function was built that included the effect of the different production stages, namely fermentation, purification and concentration. This cost function was also successfully used for the determination of the fraction of product to be collected (peak cutting) in chromatography. It can be used for protein products with different characteristics and qualities, such as purity and yield, by choosing the appropriate parameters.     

The tuning of a model-based estimator for the specific growth rate of Candida utilis

Control of microbial conversion processes is frequently inhibited by the infeasibility of measuring important process variables. In order to circumvent this lack of measurements, an accurate or valuable and conveniently measurable on-line hardware measurement can be combined with the balance equations describing the process to obtain estimates of less easily measurable variables. In this article the on-line estimation of the specific growth rate of Candida utilis is evaluated. The observer-based estimator requires a hardware measurement of the biomass during fermentations in conjuction with a model of the process; therefore the Biomass Monitor, giving an on-line measurement of viable biomass, is used in the bioreactor experiments described. The optimal tuning of the estimation for the experimental conditions is described and several alternative adaptations of the design of the estimator are presented. The influence of implemented time intervals for discretization of the estimator on the reliability of the estimated growth rate values is discussed. Additionally, the necessary choice of an initial value of the estimated specific growth rate has proven to be of great importance in practice.     

Adaptive steady-state optimization of biomass productivity in continuous fermentors

An adaptive steady-state optimization algorithm is presented and applied to the problem of optimizing the production of biomass in continuous fermentation processes. The algorithm requires no modeling information but is based on an on-line identified linear model, locates the optimum dilution rate, and maintains the chemostat at its optimum operating condition at all times. The behavior of the algorithm is tested against a dynamic model of a chemostat that incorporates metabolic time delay, and it is shown that large disturbances in the subtrate feed concentration and the specific growth rate, causing a shift in the optimum, are handled well. The developed algorithm is also used to drive a methylotroph single-cell production process to its optimum.     

Analysis and purification of plasmid DNA by reversed-phase high-performance liquid chromatography

Large nucleic acids can be separated by reversed-phase high-performance liquid chromatography. Under our experimental conditions, the retention time depends not on the chain length but rather on the base composition and the secondary structure of the molecule. Because of the torsional strain caused by the supercoiling of the plasmid, more of its bases are accessible for interaction with the hydrophobic stationary phase. This increases the retention time of the supercoiled DNA compared to the relaxed or linear DNA. We have exploited these properties to analyze the quality of plasmid preparations. The method is more sensitive to contaminants than common electrophoretic techniques. Furthermore, we describe a convenient and rapid procedure for purifying plasmid DNA. The highly pure plasmid is biologically more active for most of the enzymatic reactions commonly used in genetic engineering.     

Using interpolation to estimate system uncertainty in gene expression experiments

The widespread use of high-throughput experimental assays designed to measure the entire complement of a cell's genes or gene products has led to vast stores of data that are extremely plentiful in terms of the number of items they can measure in a single sample, yet often sparse in the number of samples per experiment due to their high cost. This often leads to datasets where the number of treatment levels or time points sampled is limited, or where there are very small numbers of technical and/or biological replicates. Here we introduce a novel algorithm to quantify the uncertainty in the unmeasured intervals between biological measurements taken across a set of quantitative treatments. The algorithm provides a probabilistic distribution of possible gene expression values within unmeasured intervals, based on a plausible biological constraint. We show how quantification of this uncertainty can be used to guide researchers in further data collection by identifying which samples would likely add the most information to the system under study. Although the context for developing the algorithm was gene expression measurements taken over a time series, the approach can be readily applied to any set of quantitative systems biology measurements taken following quantitative (i.e. non-categorical) treatments. In principle, the method could also be applied to combinations of treatments, in which case it could greatly simplify the task of exploring the large combinatorial space of future possible measurements.     

Analysis of the circular dichroism spectrum of proteins using the convex constraint algorithm: a practical guide.

Due to the time scale of circular dichroism (CD) measurements, it is theoretically possible to deconvolute such a spectrum if the pure CD spectra differ significantly from one another. In the last decade several methods have been published aiming at obtaining the conformational weights, or percentages (which are the coefficients for a linear combination) of the so-called typical secondary structural elements making up the three-dimensional structure of proteins. Two methods that can be used to determine the secondary structures of proteins are described here. The first method, called LINCOMB, is a simple algorithm based on a least-squares fit with a set of reference spectra representing the known secondary structures and yielding an estimation of weights attributed to alpha-helix, beta-pleated sheet (mainly antiparallel), beta-turns, unordered form, and aromatic/disulfide (or nonpeptide) contributions of the protein being analyzed. This method requires a "template" or reference curve set, which was obtained from the second method. The second method, "convex constraint analysis," is a general deconvolution method for a CD spectra set of any variety of conformational type. The algorithm, based on a set of three constraints, is able to deconvolute a set of CD curves to its common "pure"-component curves and conformational weights. To analyze a single CD spectrum with this method, the spectrum is appended to the data set used as a reference data set. As a way to determine the reliability of the algorithm and provide a guideline to its usage, some applications are presented.     

Adaptive on-line simulation of bioreactors: Fermentation monitoring and modeling system

Summary In order to study and control fermentation processes, indirect on-line measurements and mathematical models can be used. Here an on-line model for fermentation processes is presented. The model is based on atom and partial mass balances as well as on stability equations for the protolytes. The model is given an adaptive form by including transport equations for mass transfer and expressions for the fermentation kinetics. The state of the process can be estimated on-line using the balance component of the model completed with measurement equations for the input and the output flows of the process. Adaptivity is realized by means of on-line estimation of the parameters in the transport and kinetic expressions using recursive regression analysis. On-line estimation of the kinetic and mass transfer parameters makes model-based predictions possible and enables intelligent process control while facilitating testing of the validity of the measurement variables. A practical MS-Windows 3.1 model implementation called FMMS—Fermentation Monitoring and Modeling System is shown. The system makes it easy to configure the operating conditions for a run. It uses Windows dialogs for all set-ups, model configuration parameters, elemental compositions, on-line measurement devices and signal conditioning. Advanced on-line data analysis makes it possible to plot variables against each other for easy comparison. FMMS keeps track of over 100 variables per run. These variables are either measured or estimated by the model. Assay results can also be entered and plotted during fermentation. Thus the model can be verified almost instantly. Historical fermentation runs can be re-analyzed in simulation mode. This makes it possible to examine different signal conditining filters as well as the sensitivity of the model. Combined, the data analysis and the simulation mode make it easy to test and develop model theories and new ideas.     

A predictive and feedback control algorithm maintains a constant glucose concentration in fed-batch fermentations.

A combined predictive and feedback control algorithm based on measurements of the concentration of glucose on-line has been developed to control fed-batch fermentations of Escherichia coli. The predictive control algorithm was based on the on-line calculation of glucose demand by the culture and plotting a linear regression to the next datum point to obtain a predicted glucose demand. This provided a predictive "coarse" control for the glucose-based nutrient feed. A direct feedback control using a proportional controller, based on glucose measurements every 2 min, fine-tuned the feed rate. These combined control schemes were used to maintain glucose concentrations in fed-batch fermentations as tight as 0.49 +/- 0.04 g/liter during growth of E. coli to high cell densities.     

A predictive and feedback control algorithm maintains a constant glucose concentration in fed-batch fermentations

A combined predictive and feedback control algorithm based on measurements of the concentration of glucose on-line has been developed to control fed-batch fermentations of Escherichia coli. The predictive control algorithm was based on the on-line calculation of glucose demand by the culture and plotting a linear regression to the next datum point to obtain a predicted glucose demand. This provided a predictive "coarse" control for the glucose-based nutrient feed. A direct feedback control using a proportional controller, based on glucose measurements every 2 min, fine-tuned the feed rate. These combined control schemes were used to maintain glucose concentrations in fed-batch fermentations as tight as 0.49 +/- 0.04 g/liter during growth of E. coli to high cell densities.     

Data processing for solid substrate cultivation bioreactors

Successful scaling up of Solid Substrate Cultivation (SSC) bioreactors has been hampered by the lack of reliable models that describe such processes satisfactorily. Even though experimental data may be available for model development, data analysis is hindered by system heterogeneity and noisy measurements. This work presents a data processing procedure for periodically agitated SSC fixed bed reactors. The procedure considers several steps. First, all measurements were pre-processed on-line during the cultivation using a low pass fourth order Butterworth digital filter. Then, using this pre-processed data, the average bed temperature, evaporation rate, removed heat, and CO₂ production rate were computed off-line. The variables used to compute the evaporation rate and the removed heat were smoothed off-line with a peak shaving algorithm and a non-delay inducing forward/backward moving average scheme. Variables associated with biomass growth (CO₂ and metabolic heat) are known to evolve slowly. Hence, these were reprocessed with a smoothing procedure in order to diminish the effects of bioreactor heterogeneity. Here, moving average smoothing was applied using a larger window than for other variables, and determined empirically in order to smooth the pre-processed data and extract its real trend. The whole procedure was assessed with data from a 200 kg capacity SSC bioreactor in the cultivation of a filamentous fungus (Gibberella fujikuroi) on wheat bran.     

Structured modeling and state estimation in a fermentation process: Lipase production by Candida rugosa

A simple structured mathematical model coupled with a methodology of state and parameter estimation is developed for lipase production by Candida rugosa in batch fermentation. The model describes the system according to the following qualitative observations and hypothesis: Lipase production is induced by extracellular oleic acid present in the medium. The acid is transported into the cell where it is consumed, transformed, and stored. Lipase is excreted to the medium where it is distributed between the available oil-water interphase and aqueous phase. Cell growth is modulated by the intracellular substrate concentration. Model parameters have been determined and the whole model validated against experiments not used in their determination. The estimation problem consists in the estimation of three state variables (biomass, intra- and extracellular substrate) and two kinetic parameters by using only the on-line measurement provided by exhaust gas analysis. The presented estimation strategy divides the complex problem into three subproblems that can be solved by stable algorithms. The estimation of biomass (X) and the specific growth rate (mu), is achieved by a recursive prediction error algorithm using the on-line measurement of the carbon dioxide evolution rate. mu is then used to perform an estimation of intracellular substrate and the other kinetic parameter related to substrate transport (A) by an adaptive observer. Extracellular substrate is then evaluated by means of the estimated values of intracellular substrate and biomass through the material balance of the reactor. Simulation and experimental tests showed good performance of the developed estimator, which appears suitable to be used for process control and monitoring. (c) 1995 John Wiley & Sons, Inc.     

Automatic on-line growth estimation method for outdoor algal biomass production

An on-line measuring procedure for estimating productivity in outdoor algal cultures was developed and tested experimentally. The procedure is based on a previously described method for on-line measuring net O(2) production rate (OPR). The data obtained by this method was found to correlate well with the conventional procedures for estimation productivity by measuring the changes in biomass concentration in the culture. The new procedure seems to be superior to the latter since it can be carried out in an almost continuous way and can give immediate indication on the productivity. OPR could be used to monitor on-line the photosynthetic and/or respiration activity in small research fermentors or in large-scale open systems outdoors.     

An on-line approach to monitor ethanol fermentation using FTIR spectroscopy

Fermentation process control is currently limited by its inability to measure parameters such as substrate, product, and biomass concentrations rapidly for consistent on-line feedback. Physical and chemical parameters, such as temperature and pH, currently can be obtained on-line using appropriate sensors. However, to obtain information on the concentration of the substrate, product, and biomass, samples must be taken off-line for measurement. With the use of spectroscopic techniques, real-time monitoring of process constituents such as product and substrate is possible. Spectroscopic techniques are rapid and nondestructive, require minimal or no sample preparation, and can be used to simultaneously assess several constituents in complex matrices. The production of ethanol is the largest fermentation process in terms of production volume and economic value as a result of its prominence in the food, agricultural, and fuel industries. This study attempts to develop an on-line ethanol fermentation monitoring technique using Fourier transform infrared (FTIR) spectroscopy with a flow-through ATR capability. Models developed using multivariate statistics, employed to obtain on-line FTIR measurements, were successfully validated by off-line HPLC analysis and spectrophotometry data. Standard errors of prediction (SEP) values of 0.985 g/L (R2 = 0.996), 1.386 g/L (R2 = 0.998), and 0.546 (R2 = 0.972) were obtained for ethanol, glucose, and OD, respectively. This work demonstrates that FTIR spectroscopy could be used for rapid on-line monitoring of fermentation.     

Anomeric specificity of glucose uptake systems in Lactococcus cremoris, Escherichia coli, and Saccharomyces cerevisiae: mechanism, kinetics, and implications

The mechanism and kinetics of the glucose uptake systems of three representative microorganisms are studied during cultivation in a chemostat. The three microorganisms are Lactococcus cremoris, Escherichia coli, and Saccharomyces cervisiae. Two models describing respectively competitive and independent uptake of the two glucose anomers are tested on experimental data where alpha- and beta-glucose are determined by flow injection analysis after pulse addition of the pure anomers to a chemostat. The very accurate experimental results are used to give a convincingly clear model discrimination for all three microorganisms. The uptake of glucose by S. cervisiae occurs by a competitive mechanism with preference for alpha-glucose (K(alpha) = 32 mg/L and K(beta) = 48 mg/L). Surprisingly, the glucose uptake by the two bacteria is shown to be mediated by anomer specific transport systems with no competitive inhibition from the other glucose anomer. This novel finding has not been described in the literature on the phosphotransferase system. In L. cremoris the relative uptake rates of the glucose anomers match the equilibrium composition exactly (36% alpha-glucose). In E. coli the relative uptake rate of alpha-glucose at glucose unlimited growth is 26%, which means preference for beta-glucose. However, the saturation constants of the two sites in E. coli are K(alpha) = 2 mg/L and K(alpha) = 15 mg/L, and a preference for alpha-glucose is exhibited at very low glucose concentrations. The results are of considerable improtance in relation to enzyme based on-line measurements during fermentations as well as to the modeling of glucose limited growth and product formation.     

On the applicability of adaptive bioprocess state estimators

This article presents an industrial case study, examining the application of a novel adaptive biomass estimator to an industrial microfungi production process. It is our intention that this contribution should focus upon the implementation issues of the algorithm, in preference to a rigorous theoretical development. The novel algorithm adopted is developed from Adaptive Inferential Estimation studies of Guilandoust and co-workers. The technique utilizes input-output process measurements obtained at different frequencies, thereby providing more frequent estimates of biomass concentration than are otherwise available from off-line laboratory analyses. The algorithm is particularly suited to the biotechnology industry, as it is capable of utilizing irregular assay measurements with varying delays.Although this article demonstrates the encouraging industrial implications of the adaptive algorithm, like all adaptive techniques currently developed, it is restricted by the inability to perform robust on-line system identification. The ultimate selection of a "suboptimal" "fixed parameter" algorithm for on-line implementation, is therefore directly attributable to these inadequacies. Aspects of data acquisition, data pretreatment, and data quality are critical for real process applications, and while some practical approaches are adopted here, many important implementation problems remain unresolved. (c) 1993 John Wiley & Sons, Inc.     

Adaptive on-line model for aerobic Saccharomyces cerevisiae fermentation

In order to study and control fermentation processes, indirect on-tine measurements and mathematical models can be used. In this article we present a mathematical on-line model for fermentation processes. The model is based on atom and partial mass balances as well as on equations describing the acid-base system. The model is brought into an adaptive form by including transport equations for mass transfer and unstructured expressions for the fermentation kinetics. The state of the process, i.e., the concentrations of biomass, substrate, and products, can be estimated on-line using the balance part of the model completed with measurement equations for the input and output flows of the process. Adaptivity is realized by means of on-line estimation of parameters in the transport and kinetic expressions using recursive regression analysis. These expressions can thus be used in the model as valid equations enabling prediction of the process. This makes model-based automation of the process and testing of the validity of the measurement variables possible. The model and the on-line principles are applied to a 3.5-L laboratory tormentor in which Saccharomyces cerevisiae is cultivated. The experimental results show that the model-based estimation of the state and the predictions of the process correlate closely with high-performance liquid chromatography (HPLC) analyses. (c) 1995 John Wiley & Sons, Inc.     

A Computer-aided noninterfering on-line technique for monitoring oxygen-transfer characteristics during fermentation processes

The role of computers in the monitoring and control of fermentation processes has increased steadfastly. The ultimate utility of the machines will not depend on the availability of online sensors but also on the availability of techniques that combine direct measurements, leading towards estimates of variable closely related to the microbial process or its control. In this article, a methodology for on-line and noninterfering evaluation of the volumetric mass-transfer coefficient K(l)a is developed. A detailed presentation of the procedure, called "the static method," is given. Its feasibility is proved through implementation of the method on an antibiotic fermentation process. These experiments indicate that operator actions meant to modify the oxygen-transfer conditions can be checked on-line. The quantitative value of the static method is ascertained by comparing the experimental results with K(l)a estimates obtained with the "gassing-out" method. A sensitivity analysis was carried out, revealing the need for temperature and pressure corrections and showing that the precision of the oxygen analyzer determines the precision of the static method.