Bayesian network and response surface methodology for prediction and improvement of bacterial metabolite production

Abstract

The optimization of antifungal activity production by Bacillus amyloliquefaciens was carried out using Response Surface Methodology (RSM) in two steps. The first step involved the screening of cultural parameters affecting the production. The second step involved the optimization of significant ones. In this study, we used Bayesian network to predict the results of the experiments required for the second step. Then, by RSM, using the predicted values by BN, we defined the composition of a culture medium allowing 56% improvement in antifungal activity production over the basal medium. Such medium composition and improvement were shown to be similar to that obtained in the previous study demonstrating that, when coupled with RSM, BN permitted improvement of antifungal activity production with a much reduced number of experiments.