Artificial Neural Network-Assisted Spectrophotometric Method for Monitoring Fructo-oligosaccharides Production

作者：Erdos Balazs; Grachten Maarten; Czermak Peter; Kovacs Zoltan^*

来源：Food and Bioprocess Technology, 2018, 11(2): 305-313.

DOI：10.1007/s11947-017-2011-3

• Fructo-oligosaccharides
• Enzyme membrane reactor
• Artificial neural network
• UV spectrophotometry
• Prebiotics

Short-chain fructo-oligosaccharides (FOS) are considered as low-calorie carbohydrates with prebiotic function. They can be produced from sucrose by fructosyltransferase activity, resulting in a mixture of saccharides with different chain lengths. Current practice for carbohydrate analysis involves the use of time-costly and off-line chromatographic procedures. This study is dedicated to the development of an artificial neural network (ANN) model for predicting carbohydrate composition from the direct measurement of UV spectra. A total of 182 samples were generated by operating an enzyme membrane reactor (EMR) under both optimal and suboptimal settings. The concentration data determined by HPLC and corresponding absorbance readings were used to train a two-layer feed-forward neural network. The optimized model was then validated by using new observations that were not involved in the training. The model explained 98, 97, and 88% of the variation in the composition of the new observations regarding the main components sucrose, kestose, and glucose with a mean squared error of prediction of 6.59, 3.40, and 2.81, respectively. The results indicate that the proposed UV-ANN method has a great potential to be used for the real-time monitoring of the bioconversion.

• 出版日期2018-2