基于人工神经网络的酶解波纹巴非蛤制备小分子肽的研究

结合人工神经网络(artificial neural networks，ANNs)的良好特性，利用正交试验获得的数据作为神经网络的训练样本，建立输入为酶解实验条件参数，输出为短肽产率的神经网络模型，并通过随机选取的样本检验了ANNs 模型的准确性。利用ANNs 模型所预测出的数据，再次结合正交试验法，对酶解波纹巴非蛤实验条件进一步优化。实验结果表明：人工神经网络优化结果的小分子肽产率为4.944%，优于正交试验4.670% 的小分子肽产率。将神经网络与正交试验结合用于酶解实验条件优化可以缩短优化实验参数的时间，获得比单纯的正交试验更优化的实验条件。

Artificial Neural Network-based Optimization of Enzymolysis of Paphia undulate Meat for Production of Small Peptides

Based on the training of artificial neural networks (ANNs) using orthogonal arrays, a model for the productivity of small peptides as the output of the input consisting of five technological parameters for papain hydrolysis of Paphia undulate meat developed and validated for reliability using arbitrarily selected specimens. The further optimization of optimal values of these parameters obtained using orthogonal array design was conducted based on the AAN model by means of small-step search. AAN-based optimization gave a productivity of small peptides of 4.944%, higher than the value of 4.670% from orthogonal array optimization. In conclusion, our results reveal that more optimized technological parameters and higher optimization efficiency can be obtained using combined ANNs and orthogonal array design than using orthogonal array design alone.