Comparison of response surface methodology and artificial neural network applied to enzymatic hydrolysis of rapeseed straw

Comparison of response surface methodology and artificial neural network applied to enzymatic hydrolysis of rapeseed straw

Special Abstracts / Journal of Biotechnology 150S (2010) S1–S576 hard woods and herbs, so its fermentation is essential for the economic conversion o...

62KB Sizes 1 Downloads 23 Views

Special Abstracts / Journal of Biotechnology 150S (2010) S1–S576

hard woods and herbs, so its fermentation is essential for the economic conversion of lignocellulose to ethanol, which may provide an ideal alternative fuel source in the future. To date, most efforts in the engineering of S. cerevisiae for xylose fermentation have focused on manipulation of the initial xylose metabolic pathway by borrowing heterogeneous genes from yeast or fungi in order to reconstruct an efficient xylose assimilation pathway in S. cerevisiae. However, the performance of recombinant strains is still inferior to that of native xylose-fermenting hosts. In order to enable the utilization of xylose in S. cerevisiae, firstly we improved the xylose uptake of this yeast by chemical mutagenesis. Since xylose uptake is usually suppressed by the uptake of glucose, the resistant mutant against 0.5% (w/v) deoxyglucose was collected, and we obtained M2 as interested mutant. This mutant showed 16-fold high xylose uptake ability compared to the wild-type. Next, we conducted the cell fusion of M2 with xyloseutilizable Candida intermedia by using Sorbitol/PEG method. Three fused yeast cells (FC) were obtained, and FC No.1 (FC-1) showed high growth in mineral medium (SD). The fermentation of FC-1 yeast cells using synthetic sugars such glucose (20 g/L), xylose (20 g/L) and arabinose (5 g/L) showed that the strain could convert the sugars except arabinose into ethanol at high yield of 0.5 g-ethanol/g-sugars and the productivity of 0.46 g/L.h was obtained. We also already confirmed, however, this FC-1 yeast strain possess the high ethanol productivity for 4 generation. When 80 g-rice straw were used as substrate, FC-1 strain could convert the hydrolysates of rice straw into 20 g-ethanol by simultaneous saccharification and fermentation at 30 ◦ C. This result indicates that the strain can be potentially used in industrial scale of ethanol using lignocellulosic biomass as its primary substate. doi:10.1016/j.jbiotec.2010.08.358 [P-B.6] Increasing ethanol productivity from xylose in recombinant Saccharomyces cerevisiae by protein engineering D. Runquist ∗ , B. Hahn-Hägerdal, M. Bettiga Lund University, Sweden Keywords: Bioethanol; Lignocellulose; Xylose reductase; Saccharomyces cerevisiae Second generation bioethanol is produced by fermentation of lignocellulose biomass derived from forest and agricultural byproducts, e.g. spruce chips and corn stover. The use of these substrates do not compete with food and feed production, however compared to starch based ethanol production conversion of lignocellulose is more complicated and a large fraction of the biomass consists of non-hexose carbohydrates. Next to glucose the second most abundant sugar in lignocellulose biomass is xylose, which may constitute as much as 40% of total carbohydrate content. Fermentation of xylose to ethanol has been achieved in the yeast S. cerevisiae by genetic engineering. A major limitation of xylose utilization is however the low ethanol productivity compared to glucose. In the current study, the cofactor specificity of the enzyme xylose reductase was altered in recombinant S. cerevisiae by protein engineering. The mutated enzyme displayed increased affinity for NADH compared to NADPH which improved cofactor recycling in the initial xylose catabolism. As a consequence of the mutation, the ethanol productivity from xylose was increased by an order of magnitude and the ethanol yield from xylose was improved to the same level as the one encountered for glucose. Due to the improved ethanol productivity, the strain harbor-


ing the mutated enzyme acquired the capacity of high anaerobic growth on xylose as sole carbon source. Since a rational genetic engineering approach was taken, the improved traits of the engineered strain are readily transferable to industrial S. cerevisiae strains. doi:10.1016/j.jbiotec.2010.08.359 [P-B.7] Comparison of response surface methodology and artificial neural network applied to enzymatic hydrolysis of rapeseed straw E. Castro ∗ , C. Cara, M.J. Jesus, V. Rivas University of Jaén, Spain Keywords: Rapeseed straw; Artificial networks; Enzymatic hydrolysis; Ethanol Rapeseed (Brassica napus) has traditionally been grown for the production of animal feed and vegetable oil for human consumption. In the last years, an increasing fraction of rapeseed oil has been used as raw material for biodiesel production. More than 30 million hectares of rapeseed are cultivated worldwide. After seed harvesting, rapeseed straw is left behind in the fields until it is eliminated. This agricultural residue is a renewable, low cost, and lacking of alternatives lignocellulosic material whose use as raw material for ethanol production has been proposed. Pre-treatment is a key step of the global conversion process by which the lignocellulosic structure is break down, thus facilitating enzyme access to the cellulose chains. Pre-treatment performance is usually assessed by enzymatic hydrolysis of the pretreated solid materials. Experimental design and Response Surface Methodology (RSM) are used to explore the influence of the main operational conditions. In this work, the pre-treatment temperature and residence time for rapeseed straw conversion are optimized, and the enzymatic hydrolysis yields at different operational conditions are predicted, using both RSM and Artificial Neural Network (ANN). For RSM, a rotatable central composite design was performed. For ANN, the Radial basis function neural network (RBFN) and Multilayer Perceptron have been used. Similarly to Multilayer Perceptrons, RBFN have been proved to be universal approximators. Nevertheless, its simple topology and the fact that its output depends on the spatial position of the inputs, make RBFN more interpretable than multilayer perceptrons. For this research, the Multilayer Perceptron and RBFN included in KEEL software package have been used. Estimated responses were compared with the experimental results and prediction capabilities of RSM and ANNs were determined. The coefficient of determination (R2 ) and average absolute deviation (AAD) values between actual and estimated responses were also evaluated. doi:10.1016/j.jbiotec.2010.08.360