54 Optimization of Fermentation Media for Wine Lactic Acid Bacteria Using ANN and GA

A chemically defined medium is being developed to study the nutritional requirement of a commercial starter of Oenococcus oeni involved in the microbial deacidification of wine. The initial non-optimized medium provided lower growth than classic laboratory media. The relevance of individual medium compounds for growth was modeled using an Artificial Neural Network. A Genetic Algorithm utilized the generated model to evaluate each evolutionary set of media compounds for optimizing the media to produce the maximum optical density (OD) which is used to evaluate the growth of the bacteria. The predicted ODs were compared to the measured ODs for the same media compounds to validate the model. Even using a restricted number of training examples, a reasonable model of the growth media was achievable by the ANN and this allowed the GA to generate new media compositions that produced very good results.