Using genetic algorithms coupling neural networks in a study of xylitol production: medium optimisation

doi:10.1016/S0032-9592(02)00094-8

Process Biochemistry

Volume 38, Issue 7, 28 February 2003, Pages 979-985

https://doi.org/10.1016/S0032-9592(02)00094-8 Get rights and content

Abstract

Genetic algorithms (GAs) coupling neural networks (NNs) were developed for the medium optimisation of xylitol fermentation by Candida mogii with high yield. The concentrations of 6 medium components were optimised within 50 experiments and 40 emulate experiments to give a maximum yield with minimum use of medium content. With the optimised medium composition in a stirred tank bioreactor for batch production of xylitol, under the same operating conditions, xylitol yield was improved by 15%, compared to results with the medium used previously.

Introduction

Xylitol is a polyol with some interesting properties which make it of high value to pharmaceutical, food and chemical industries etc. Despite this wide range of applications, the use of xyitol is limited by comparatively high production costs and this has encouraged the development of new technologies able to lower the production costs. The microbiological conversion of xylose solutions is a selective and promising process [1].

The bioconversion of pentose into polyalcohols is a complex process influenced by several factors. To improve xylitol fermentation, designing a fermentation medium is of critical importance because medium composition can significantly affect product concentration, yield, volumetric productivity, and the ease and cost of downstream product separation [2].

The optimisation of medium composition uses mainly the ‘one-at-a-time’ method and statistical approaches such as response surface methodology based on factorial design experiments [3], [4], [5], [6], [7], orthogonal array design based on statistical approaches [8] or uniform design based on numeral theory [9]. The ‘one-at-a-time’ method ignores interactions among the different medium components, optimal conditions may be missed and a large number of experiments are required. For the application of response surface methodology, a model must be assumed to determine the relative influence of the various medium components. This model is then used to calculate the optimal concentrations for the objective function. The common model used is a full second-order polynomial. The number of experiments is given by L^N (N factors at L levels), therefore in practice only two or three levels can be applied and plotting is limited to two variables at a time. The level of orthogonal array design or uniform design is also limited by the factor.

Genetic algorithms (GAs) use evolutionary natural selection processes, where selection results in species that fit the best. GAs have proved to be extremely suitable for the optimisation of highly non-linear problems with many variables and for reported problems the global optimum was found [10]. Using GAs for medium optimisation, Freyer et al. [11] succeeded in increasing the biomass productivity. Deuster-Botz et al. [12], [13] optimised 12 concentrations of 12 medium components within 144 experiments to give maximum specific hydrocortisone dehydrogenase activity and biomass yield, and 14 medium components within 125 experiments to give a maximum formate-dehydrogenase activity with minimal use of nutrient salts. In the study of mevinolin biosynthesis by Aspergillus terreus, within four generations of fermentation experiments the productivity has increased nearly three times [14]. With GAs however no insight from the designer is used and historical data are abandoned on each iteration.

Neural networks (NNs) may provide considerable advantages in prediction against different biological backgrounds. Their strong points are that they work well with large amounts of data and require no mechanistic description of the system. This makes NNs particularly well suited to medium design [15].

In this paper, a new medium optimisation method based on GAs coupled to NNs was developed for experimental design of xylitol fermentation (Fig. 1).

Section snippets

Micro-organism, medium and culture conditions

Candida mogii ATCC 18364 used in this work was maintained on potato dextrose agar slants at 4 °C. The inoculum was prepared using (g dm⁻³): glucose, 20; xylose, 40; (NH₄)₂HPO₄, 4; (NH₄)₂SO₄, 1; KH₂PO₄, 4; MgSO₄, 1; Bacto-yeast extract, 6 and Bacto-peptone, 6.5. The medium was autoclaved at 121 °C for 20 min after pH adjustment to 5 with 1 M NaOH at pH 5. The culture was grown in 300 cm³ shake-flasks with 170 cm³ medium at 200 rpm and 3 ° C for 48 h. After 20 min of sterile centrifugation at 3000

Medium optimisation

In parallel and standardised shake-flask experiments, medium optimisation was performed for four generations using the GA on the basis of the medium used by Chen et al. [17].

Fig. 2, Fig. 3 represent the course for each composition created by the GA in generation 0 and generation 3 for optimisation of xylitol fermentation by C. mogii. The concentration range is divided into 10 identical subunits, each of which represents the sum of the ‘individuals’ of 10 concentration levels.

In generation 0

Acknowledgements

This work was supported by grant No. 29776025 from NNSFC (National Natural Science Foundation of China) and No. C96068 from NSFFC (Natural Science Foundation of Fujian, China).

References (19)

P. Nigam et al.
Processes for fermentative production of xylitol—a sugar substitute
Process Biochem.
(1995)
S.-L. Lee et al.
Optimisation of medium composition for the production of glucosyltransferase by Aspergillus niger with response surface methodology
Enzyme Microb. Technol.
(1997)
S. Bloor et al.
Elucidation and optimisation of the medium constituents controlling antibiotic production by the cyanobacterium Nostoc muscorum
Enzyme Microb. Technol.
(1991)
R.G. Silveira et al.
Medium optimisation by an orthogonal array design for the growth of Methanosarcina barkeri
J. Ferm. Bioeng.
(1991)
D. Weuster-Botz et al.
Medium optimisation by genetic algorithm for continuous production of formate-dehydrogenase
Process Biochem.
(1995)
S. Sirisansaneeyakul et al.
Screening of yeasts for production of xylitol from d-xylitol
J. Ferm. Bioeng
(1995)
M. Kennedy et al.
Strategies for improving fermentation medium perfornance: a review
J. Ind. Microbiol. Biotechnol.
(1999)
I. Roberto et al.
Influence of media composition on xylitol fermentation by Candida guilliermondii using response surface methodology
Biotechnol. Lett.
(1995)
P.V. Rao et al.
Production of lipase by Candida rugosa in solid state fermentation. 2: Medium optimisation and effect of aeration
Process Biochem.
(1993)

There are more references available in the full text version of this article.

Cited by (53)

Optimization of medium components for protein production by Escherichia coli with a high-throughput pipeline that uses a deep neural network
2024, Journal of Bioscience and Bioengineering
To optimize rapidly the medium for green fluorescent protein expression by Escherichia coli with an introduced plasmid, pRSET/emGFP, a single-cycle optimization pipeline was applied. The pipeline included a deep neural network (DNN) and mathematical optimization algorithms with simultaneous optimization of 18 medium components. To evaluate the DNN data sampling method, two methods, orthogonal array (OA) and Latin hypercube sampling (LHS), were used to design 64 initial media for each sampling method. The OA- and LHS-based data sampling resulted in green fluorescent protein fluorescence intensities of 0.088 × 10³−1.85 × 10⁴ and 3.30 × 10³−1.50 × 10⁴, respectively. Fifty DNN models were built using the OA and LHS datasets. Hold-out validation was performed using 15 % test of OA and LHS data. Mean square errors of the DNN models were 0.015–0.64, indicating the estimation accuracies were sufficient. However, the sensitivities of components in the DNN models varied and were grouped into six major classes by the index of k-means clustering. A representative model was selected for each class. Mathematical optimization algorithms using Bayesian optimization and genetic algorithm were applied to the representative models, and representative optimized medium (OM) compositions were selected by k-means clustering from the proposed OMs. A total of 54 OMs were obtained from the OA and LHS datasets. In the validating cultivation, the best OMs of OA and LHS were 2.12-fold and 2.13-fold higher, respectively, than those of the learning data.
Challenges in developing cell culture media using machine learning
2024, Biotechnology Advances
Microbial and mammalian cells are widely used in the food, pharmaceutical, and medical industries. Developing or optimizing culture media is essential to improve cell culture performance as a critical technology in cell culture engineering. Methodologies for media optimization have been developed to a great extent, such as the approaches of one-factor-at-a-time (OFAT) and response surface methodology (RSM). The present review introduces the emerging machine learning (ML) technology in cell culture engineering by combining high-throughput experimental technologies to develop highly efficient and effective culture media. The commonly used ML algorithms and the successful applications of employing ML in medium optimization are summarized. This review highlights the benefits of ML-assisted medium development and guides the selection of the media optimization method appropriate for various cell culture purposes.
Improvement of ultrasound-assisted extraction of polyphenolic content of the plant Cytisus triflorus L'Her as a natural resource using artificial neural network modeling and multi-objective optimization
2023, Sustainable Chemistry and Pharmacy
The improvement of environmental performance of production processes is a critical problem right now. Processes that are effective in terms of productivity and environmental effects are the center of interest of researchers. This study established such a method for an ultrasound-assisted extraction procedure. The plant Cytisus triflorus L'Her was used to extract polyphenolic chemicals using ultrasound-assisted extraction as a quick, less energy consuming, high-yield technology. This work offers a new experimental planning approach using Artificial Neural Networks, that enables the extraction process to be optimized using multi-objective optimization in order to identify the best extraction conditions to obtain the maximum polyphenolic compounds possible. Advanced modeling using Artificial intelligence and statistical optimization techniques was used to determine the best combinations of ethanol concentration, extraction temperature and time for the maximum extraction of total phenols, total flavonoids, total flavonols, and total tannins. An artificial neural network (ANN) was used to build the models for each of the four goals, then they were multi-objectively optimized using a genetic algorithm technique. The maximum polyphenolic contents including TPC (4.986 ± 0.006 mg GA/g DE), TFC (14.026 ± 0.070 mg Q/g DE), TFolC (5.225 ± 0.010 mg R/g DE), and TTC (6.240 ± 0.021 mg C/g DE) were obtained using an extraction process under optimal conditions, which included a solvent concentration of 78%, an extraction temperature of 52 °C, and an extraction time of 55 min which were obtained using the MOO Pareto optimum solution selection approach Gray Relational Analysis GRA.
Machine learning for biochemical engineering: A review
2021, Biochemical Engineering Journal
The field of machine learning is comprised of techniques, which have proven powerful approaches to knowledge discovery and construction of ‘digital twins’ in the highly dimensional, nonlinear and stochastic domains common to biochemical engineering. We review the use of machine learning within biochemical engineering over the last 20 years. The most prevalent machine learning methods are demystified, and their impact across individual biochemical engineering subfields is outlined. In doing so we provide insights into the true benefits of each technique, and obstacles for their wider deployment. Finally, core challenges into the application of machine learning in biochemical engineering are thoroughly discussed, and further insight into adoption of innovative hybrid modelling and transfer learning strategies for development of new digital biotechnologies is provided.
Increased mixing intensity is not necessary for more efficient cellulose hydrolysis at high solid loading
2021, Bioresource Technology
Citation Excerpt :
To improve the efficiency of high solid enzymatic hydrolysis, considerable efforts have been made to develop more efficient hydrolysis strategies, such as pre-hydrolysis, fed-batch operation and multi-feeding strategy (Koppram et al., 2014; López-Linares et al., 2014), or develop more efficient reactors, such as stirring reactor (with impeller) and shaking/rotating reactor (with no impeller) (Chen and Liu, 2017; Liu et al., 2015). For the high solid loading hydrolysis in stirring or shaking reactors, it has been generally considered that sufficient mixing within the reactors is essential as it could enhance mass transfer and increase the hydrolysis efficiency (Baishan et al., 2003). Compared with the stirring reactor, Ghorbanian et al. (2014) and Dasari and Berson (2007) reported that the use of a horizontal rotating reactor improved the mixing efficiency, enhanced the mass and heat transfer, and minimized the enzyme deactivation induced by mechanical and thermal stress.
To enhance the cellulose hydrolysis at high solid loadings, an increased mixing intensity is generally required for the high solid loading hydrolysis, while it leads to higher energy consumption. In this study, the impact of mixing intensity on cellulose conversion during hydrolysis at different solid loadings were systematically studied. It was found that the increased mixing intensity is not necessary for more efficient cellulose hydrolysis. For cellulose hydrolysis at higher solid loadings, a lower mixing intensity is needed for a higher cellulose conversion. Although the increased mixing intensity promoted enzyme adsorption, it strengthened product inhibition and caused severer enzyme deactivation. Besides, mixing at the initial stage of cellulose hydrolysis was more crucial, while continuous mixing throughout the hydrolysis was not required for more efficient cellulose hydrolysis. Based on the mechanism study, a combined mixing strategy was developed to achieve efficient cellulose hydrolysis with about two-third reduction in energy consumption.
Improvement of ε-polylysine production by marine bacterium Bacillus licheniformis using artificial neural network modeling and particle swarm optimization technique
2017, Biochemical Engineering Journal
Citation Excerpt :
The level of orthogonal array design is confined and RSM model is only full second order polynomial. In the present days, artificial neural networks have merged as viable means to construct models and to predict outputs accurately for a new input data set via applying to multivariate non-linear bioprocesses [33–38]. One of the important advantages of ANN-based model is that it can be constructed without detailed information and has the capacity to learn from the old data.
ε-Polylysine is water soluble, biodegradable, edible and non-toxic homopolymer of l-lysine linked by the peptide bond between the carboxyl and ε-amino groups. ε-polylysine and its derivatives being used for past few decades for a broad range of industrial applications. However, the yields of ε-polylysine using wild type strains are comparatively low with respect to what desired for industrial production. Hence, in this study, an advanced modeling and optimization technique was applied to optimize medium parameters for enhanced ε-polylysine production by marine bacterium Bacillus licheniformis. The critical nutrients including glucose, yeast extract, magnesium sulphate and ferrous sulphate were incorporated in artificial neural networks (ANN) as input variables and ε-polylysine as the output variable. The ANN topology of 4-10-1 was found to be optimum upon training the model with feed-forward back propagation algorithm and on application of the developed model to particle swarm optimization (PSO) resulted in 3.56 ± 0.16 g L⁻¹ of ε-polylysine under the following optimal conditions: glucose, 34 g L⁻¹; yeast extract, 2.3 g L⁻¹; magnesium sulphate, 0.44 g L⁻¹ and ferrous sulphate, 0.08 g L⁻¹. Thus, this optimization technique could significantly improve ε-polylysine by 196.7%, as compared to un-optimized medium. The potential significance of this study lies in the development of a suitable production medium for improved ε-polylysine production by an advanced optimization approach, ANN-PSO.

View all citing articles on Scopus

View full text

Article preview

Process Biochemistry

Abstract

Introduction

Section snippets

Micro-organism, medium and culture conditions

Medium optimisation

Acknowledgements

References (19)

Processes for fermentative production of xylitol—a sugar substitute

Process Biochem.

Optimisation of medium composition for the production of glucosyltransferase by Aspergillus niger with response surface methodology

Enzyme Microb. Technol.

Elucidation and optimisation of the medium constituents controlling antibiotic production by the cyanobacterium Nostoc muscorum

Enzyme Microb. Technol.

Medium optimisation by an orthogonal array design for the growth of Methanosarcina barkeri

J. Ferm. Bioeng.

Medium optimisation by genetic algorithm for continuous production of formate-dehydrogenase

Process Biochem.

Screening of yeasts for production of xylitol from d-xylitol

J. Ferm. Bioeng

Strategies for improving fermentation medium perfornance: a review

J. Ind. Microbiol. Biotechnol.

Influence of media composition on xylitol fermentation by Candida guilliermondii using response surface methodology

Biotechnol. Lett.

Production of lipase by Candida rugosa in solid state fermentation. 2: Medium optimisation and effect of aeration

Process Biochem.

Cited by (53)

Optimization of medium components for protein production by Escherichia coli with a high-throughput pipeline that uses a deep neural network

Challenges in developing cell culture media using machine learning

Improvement of ultrasound-assisted extraction of polyphenolic content of the plant Cytisus triflorus L'Her as a natural resource using artificial neural network modeling and multi-objective optimization

Machine learning for biochemical engineering: A review

Increased mixing intensity is not necessary for more efficient cellulose hydrolysis at high solid loading

Improvement of ε-polylysine production by marine bacterium Bacillus licheniformis using artificial neural network modeling and particle swarm optimization technique