Modelling and optimization of fed-batch fermentation processes using dynamic neural networks and genetic algorithms

Enzymes from Fishery and Aquaculture Waste: Research Trends in the Era of Artificial Intelligence and Circular Bio-Economy

In the era of the blue bio-economy, which promotes the sustainable utilization and exploitation of marine resources for economic growth and development, the fisheries and aquaculture industries still face huge sustainability issues. One of the major challenges of these industries is associated with the generation and management of wastes, which pose a serious threat to human health and the environment if not properly treated. In the best-case scenario, fishery and aquaculture waste is processed into low-value commodities such as fishmeal and fish oil. However, this renewable organic biomass contains a number of highly valuable bioproducts, including enzymes, bioactive peptides, as well as functional proteins and polysaccharides. Marine-derived enzymes are known to have unique physical, chemical and catalytic characteristics and are reported to be superior to those from plant and animal origins. Moreover, it has been established that enzymes from marine species possess cold-adapted properties, which makes them interesting from technological, economic and sustainability points of view. Therefore, this review centers around enzymes from fishery and aquaculture waste, with a special focus on proteases, lipases, carbohydrases, chitinases and transglutaminases. Additionally, the use of fishery and aquaculture waste as a substrate for the production of industrially relevant microbial enzymes is discussed. The application of emerging technologies (i.e., artificial intelligence and machine learning) in microbial enzyme production is also presented.

NMPC-Based Workflow for Simultaneous Process and Model Development Applied to a Fed-Batch Process for Recombinant C. glutamicum

For the fast and improved development of bioprocesses, new strategies are required where both strain and process development are performed in parallel. Here, a workflow based on a Nonlinear Model Predictive Control (NMPC) algorithm is described for the model-assisted development of biotechnological processes. By using the NMPC algorithm, the process is designed with respect to a target function (product yield, biomass concentration) with a drastically decreased number of experiments. A workflow for the usage of the NMPC algorithm as a process development tool is outlined. The NMPC algorithm is capable of improving various process states, such as product yield and biomass concentration. It uses on-line and at-line data and controls and optimizes the process by model-based process extrapolation. In this study, the algorithm is applied to a Corynebacterium glutamicum process. In conclusion, the potency of the NMPC algorithm as a powerful tool for process development is demonstrated. In particular, the benefits of the system regarding the characterization and optimization of a fed-batch process are outlined. With the NMPC algorithm, process development can be run simultaneously to strain development, resulting in a shortened time to market for novel products.

Optimization of process conditions for mammalian fed-batch cell culture in automated micro-bioreactor system using genetic algorithm

Recombinant proteins produced by mammalian cell culture technology represent an important segment of therapeutic molecules. Development of their manufacturing processes is a time- and resource-consuming task. A wide array of process conditions, e.g. physico-chemical parameters, medium composition, feeding strategy, needs to be optimized to design a commercially feasible process with the desired productivity and product characteristics. Traditionally, statistical experimental designs, i.e. design-of-experiments methodology, have been used for such optimizations. However, statistical design approach has several limitations related to high dimensionality of the explored parameter space originating from the complexity of the mammalian cell culture processes. An alternative is therefore desired to overcome these limitations. In this study, we have successfully used a simple genetic algorithm as a method of experimental design for optimization of mammalian cell culture processes for two recombinant cell lines, one expressing a monoclonal antibody and one an Fc-fusion protein. Harnessing the automation capability of a robotically driven micro-bioreactor system to execute the genetic algorithm-derived experiments, a set of 14 process parameters was optimized within 132 experiments per cell line (six generations of 22 experiments), showing the feasibility of this approach as an alternative to classical statistical experimental designs.

Accurate and cost-effective prediction of HBsAg titer in industrial scale fermentation process of recombinant Pichia pastoris by using neural network based soft sensor

In the current work, the attempt was made to apply best-fitted artificial neural network (ANN) architecture and the respective training process for predicting final titer of hepatitis B surface antigen (HBsAg), produced intracellularly by recombinant Pichia pastoris Mut⁺ in the commercial scale. For this purpose, in large-scale fed-batch fermentation, using methanol for HBsAg induction and cell growth, three parameters of average specific growth rate, biomass yield, and dry biomass concentration-in the definite integral form with respect to fermentation time-were selected as input vectors; the final concentration of HBsAg was selected for the ANN output. Used dataset consists of 38 runs from previous batches; feed-forward ANN 3:5:1 with training algorithm of backpropagation based on a Bayesian regularization was trained and tested with a high degree of accuracy. Implementing the verified ANN for predicting the HBsAg titer of the five new fermentation runs, excluded from the dataset, in the full-scale production, the coefficient of regression and root-mean-square error were found to be 0.969299 and 2.716774, respectively. These results suggest that this verified soft sensor could be an excellent alternative for the current relatively expensive and time-intensive analytical techniques such as enzyme-linked immunosorbent assay in the biopharmaceutical industry.

Application of recurrent neural network for online prediction of cell density of recombinant Pichia pastoris producing HBsAg

Artificial neural networking (ANN) seems to be a promising soft sensor for implementing current approaches of quality by design (QbD) and process analytical technologies (PAT) in the biopharmaceutical industry. In this study, we aimed to implement best-fitted ANN architecture for online prediction of the biomass amount of recombinant Pichia pastoris (P. pastoris) - expressing intracellular hepatitis B surface antigen (HBsAg) - during the fed-batch fermentation process using methanol as a sole carbon source. For this purpose, at the induction phase of methanol fed-batch fermentation, carbon evolution rate (CER), dissolved oxygen (DO), and methanol feed rate were selected as input vectors and total wet cell weight (WCW) was considered as output vector for the ANN. The obtained results indicated that after training recurrent ANN with data sets of four fed-batch runs, this toolbox could predict the WCW of the next fed-batch fermentation process at each specified time point with high accuracy. The R-squared and root-mean-square error between actual and predicted values were found to be 0.9985 and 13.73, respectively. This verified toolbox could have major importance in the biopharmaceutical industry since recombinant P. pastoris is widely used for the large-scale production of HBsAg.

An Adaptive Fuzzy Feedforward-Feedback Control System Applied to a Saccharification Process

In industrial bioprocess control, disturbance sources typically influences process variable regulation. These disturbances may reduce a system control performance or even affect the final bioproduct quality. Therefore, feedforward control is desired because it anticipates the effects caused by these disturbances in an attempt to keep the process variable at the setpoint value. However, designing a feedforward control law requires process modeling, which can be a tough task when dealing with bioprocesses that are intrinsically nonlinear and multivariable systems. Thus, an adaptive feedforward control law or other advanced control system is needed for satisfactory disturbance rejection. For this reason, a general fuzzy feedforward control system is proposed in this paper to replace the classical feedforward control, making it easier to implement the feedforward control action by avoiding nonlinear and multivariable process modeling. The adaptive fuzzy feedforward-feedback (A4FB) system was applied to a product concentration control loop in an enzymatic reactor, to reject disturbances caused by variations in the substrate and enzymatic solutions feed concentration. The results showed that the A4FB controller rejected much more disturbance effects than classical feedforward control law, demonstrating its advantage, supported by not only its simple implementation, but also its improved disturbance rejection.

Experimental and kinetic study for lead removal via photosynthetic consortia using genetic algorithms to parameter estimation

This work presents an experimental-theoretical strategy for a batch process for lead removal by photosynthetic consortium, conformed by algae and bacteria. Photosynthetic consortium, isolated from a treatment plant wastewater of Tecamac (Mexico), was used as inoculum in bubble column photobioreactors. The consortium was used to evaluate the kinetics of lead removal at different initial concentrations of metal (15, 30, 40, 50, and 60 mgL^-1), carried out in batch culture with a hydraulic residence time of 14 days using Bold's Basal mineral medium. The photobioreactor was operated under the following conditions: aeration of 0.5 vvm, 80 μmol m^-2 s^-1 of photon flux density and a photoperiod light/dark 12:12. After determining the best growth kinetics of biomass and metal removal, they were tested under different ratios (30 and 60%) of wastewater-culture medium. Additionally, the biomass growth (X), nitrogen consumption (N), chemical oxygen demand (COD), and metal removal (Pb) were quantified. Achieved lead removal was 97.4% when the initial lead concentration was up to 50 mgL^-1 using 60% of wastewater. Additionally, an unstructured-type mathematical model was developed to simulate COD, X, N, and lead removal. Furthermore, a comparison between the Levenberg-Marquardt (L-M) optimization approach and Genetic Algorithms (GA) was carried out for parameter estimation. Also, it was concluded that GA has a slightly better performance and possesses better convergence and computational time than L-M. Hence, the proposed method might be applied for parameter estimation of biological models and be used for the monitoring and control process.

Development of a Novel Kinetic Model for Cocoa Fermentation Applying the Evolutionary Optimization Approach

Traditional Mexican cocoa fermentation performed in batch was studied by applying kinetic modelling with experimental validation. Similar microbiological behaviour was observed up to 60 h, with a temperature increase at 72 h that remained constant (50 °C) until 156 h. Metabolite-production kinetics (ethanol and acetic acid) from degradable mucilage (glucose) was explored. Exploration involved applying different combinations of unstructured growth models, in order to consider the effect of temperature when predicting the concentration of metabolites in these microorganisms. Two methods were used to optimize model parameters: the Levenberg–Marquardt optimization approach and Genetic Algorithms (GAs). GAs which could be used to scale up the fermentation process indicated the applicability of this model for predicting fermentation quality. The maximum specific rate average for μmax and saturation constant (Ks) were 0.0961 h−1 and 1.4 mg/g m.s., respectively. The results obtained indicate the expediency of this technique for future application in the design and control of batch fermentation.

Comparison Tools for Parametric Identification of Kinetic Model for Ethanol Production using Evolutionary Optimization Approach

Living cells, type of substrate, enzymatic hydrolysis play an important role in the efficiency of ethanol production; however, the kinetic parameters of biochemical reactions necessary for modelling these processes are often not accessible directly through experiments. In this context, for the implementation of suitable operational strategies, it is necessary to have kinetic models able to describe the process as realistically as possible. This paper proposes a comparative study of two nonlinear techniques for parametric identification of a kinetic model for ethanol production from recycled paper sludge in order to improve process performance. The parameters of the model are optimized by two methods: using the Levenberg–Marquardt optimization approach and Genetic Algorithms. The performances of both techniques are evaluated using a numerical simulation. The optimal value of these parameters have been obtained based on Genetic Algorithm. Finally, the effect of parametric adjustment and dilution rate on productivity was demonstrated by changing the batch operation to the continuous operating model. The maximum ethanol concentration was about 13.25 g/l in batch process and about 13.9 g/l at Dilution rate: 0.005 1/h corresponding to a productivity of 0.327 in continuous process.

Simulation and prediction of protein production in fed-batch E. coli cultures: An engineering approach

An overall model describing the dynamic behavior of fed-batch E. coli processes for protein production has been built, calibrated and validated. Using a macroscopic approach, the model consists of three interconnected blocks allowing simulation of biomass, inducer and protein concentration profiles with time. The model incorporates calculation of the extra and intracellular inducer concentration, as well as repressor-inducer dynamics leading to a successful prediction of the product concentration. The parameters of the model were estimated using experimental data of a rhamnulose-1-phosphate aldolase-producer strain, grown under a wide range of experimental conditions. After validation, the model has successfully predicted the behavior of different strains producing two different proteins: fructose-6-phosphate aldolase and ω-transaminase. In summary, the presented approach represents a powerful tool for E. coli production process simulation and control.

Evaluation of Anaerobic Biofilm Reactor Kinetic Parameters Using Ant Colony Optimization

Fixed bed reactors with naturally attached biofilms are increasingly used for anaerobic treatment of industry wastewaters due their effective treatment performance. The complex nature of biological reactions in biofilm processes often poses difficulty in analyzing them experimentally, and mathematical models could be very useful for their design and analysis. However, effective application of biofilm reactor models to practical problems suffers due to the lack of knowledge of accurate kinetic models and uncertainty in model parameters. In this work, an inverse modeling approach based on ant colony optimization is proposed and applied to estimate the kinetic and film thickness model parameters of wastewater treatment process in an anaerobic fixed bed biofilm reactor. Experimental data of pharmaceutical industry wastewater treatment process are used to determine the model parameters as a consequence of the solution of the rigorous mathematical models of the process. Results were evaluated for different modeling configurations derived from the combination of mathematical models, kinetic expressions, and optimization algorithms. Analysis of results showed that the two-dimensional mathematical model with Haldane kinetics better represents the pharmaceutical wastewater treatment in the biofilm reactor. The mathematical and kinetic modeling of this work forms a useful basis for the design and optimization of industry wastewater treating biofilm reactors.

Time-dependent fermentation control strategies for enhancing synthesis of marine bacteriocin 1701 using artificial neural network and genetic algorithm

The artificial neural network (ANN) and genetic algorithm (GA) were combined to optimize the fermentation process for enhancing production of marine bacteriocin 1701 in a 5-L-stirred-tank. Fermentation time, pH value, dissolved oxygen level, temperature and turbidity were used to construct a "5-10-1" ANN topology to identify the nonlinear relationship between fermentation parameters and the antibiotic effects (shown as in inhibition diameters) of bacteriocin 1701. The predicted values by the trained ANN model were coincided with the observed ones (the coefficient of R(2) was greater than 0.95). As the fermentation time was brought in as one of the ANN input nodes, fermentation parameters could be optimized by stages through GA, and an optimal fermentation process control trajectory was created. The production of marine bacteriocin 1701 was significantly improved by 26% under the guidance of fermentation control trajectory that was optimized by using of combined ANN-GA method.

On-line evolutionary optimization of an industrial fed-batch yeast fermentation process

This paper presents two genetic algorithms based on optimization methods to maximize biomass concentration, and to minimize ethanol formation. The objective function is maximized according to the values of feed flow rate, using genetic search approaches. Five case studies were carried out for different initial conditions, which strongly influence the optimal profiles of feed flow rate for the fermentation process. The ethanol and glucose disturbance effects were examined to stress the effectiveness of proposed approaches. The proposed genetic approaches were implemented for an industrial scale baker's yeast fermentor which produces Saccharomyces cerevisiae known as baker's yeast. The results show that optimal feed flow rate was obtained in a satisfactory and successful way for fed-batch fermentation process.

Novel optimal temperature profile for acidification process of Lactobacillus bulgaricus and Streptococcus thermophilus in yoghurt fermentation using artificial neural network and genetic algorithm

The acidification behavior of Lactobacillus bulgaricus and Streptococcus thermophilus for yoghurt production was investigated along temperature profiles within the optimal window of 38-44 degrees C. For the optimal acidification temperature profile search, an optimization engine module built on a modular artificial neural network (ANN) and genetic algorithm (GA) was used. Fourteen batches of yoghurt fermentations were evaluated using different temperature profiles in order to train and validate the ANN sub-module. The ANN captured the nonlinear relationship between temperature profiles and acidification patterns on training data after 150 epochs. This served as an evaluation function for the GA. The acidification slope of the temperature profile was the performance index. The GA sub-module iteratively evolved better temperature profiles across generations using GA operations. The stopping criterion was met after 11 generations. The optimal profile showed an acidification slope of 0.06117 compared to an initial value of 0.0127 and at a set point sequence of 43, 38, 44, 43, and 39 degrees C. Laboratory evaluation of three replicates of the GA suggested optimum profile of 43, 38, 44, 43, and 39 degrees C gave an average slope of 0.04132. The optimization engine used (to be published elsewhere) could effectively search for optimal profiles of different physico-chemical parameters of fermentation processes.