Department of Biotechnology, Indian Institute of Technology Guwahati, Guwahati, 781 039, Assam, India.
Folia Microbiol (Praha). 2013 Sep;58(5):393-401. doi: 10.1007/s12223-012-0220-8. Epub 2013 Jan 12.
Response surface methodology (RSM) and artificial neural network-real encoded genetic algorithm (ANN-REGA) were employed to develop a process for fermentative swainsonine production from Metarhizium anisopliae (ARSEF 1724). The effect of finally screened process variables viz. inoculum size, oatmeal extract, glucose, and CaCl2 were investigated through central composite design and were further utilized for training sets in ANN with training and test R values of 0.99 and 0.94, respectively. ANN-REGA was finally employed to simulate the predictive swainsonine production with best evolved media composition. ANN-REGA predicted a more precise fermentation model with 103 % (shake flask) increase in alkaloid production compared to 75.62 % (shake flask) obtained with RSM model upon validation.
响应面法（RSM）和人工神经网络-实数编码遗传算法（ANN-REGA）被用于开发一种从金龟子绿僵菌（ARSEF 1724）发酵产苦马豆素的方法。通过中心复合设计研究了最终筛选出的工艺变量（接种量、燕麦片提取物、葡萄糖和 CaCl2）的影响，并将其进一步用于 ANN 的训练集，ANN 的训练和测试 R 值分别为 0.99 和 0.94。最后，采用 ANN-REGA 模拟最佳进化培养基组成的预测苦马豆素产量。通过验证，与 RSM 模型获得的 75.62%（摇瓶）相比，ANN-REGA 预测的发酵模型更精确，生物碱产量增加了 103%（摇瓶）。
