Enhancing fermentation development procedures via artificial neural networks

This paper describes how artificial neural networks can aid in recombinant fermentation process development. Two specific areas are addressed. Firstly, neural networks are used to increase the quality of information available during the course of a run. Available on-line measurements, together with a network model, are used to estimate key bioprocess parameters. Secondly, neural networks are used to formulate process models to aid in the specification of fermentation operational procedures. The ability to capture non-linear bioprocess characteristics is particularly significant and is an enhancement to existing experimental design procedures. Both the off-line experimental design and online parameter estimation techniques can aid in the progression from shake flask scale to large pilot scale operation.