Penicillin fermentation: mechanisms and models for industrial-scale bioreactors

Multiple feature fusion transformer for modeling penicillin fermentation process with unequal sampling intervals

The quality prediction of batch processes is an important task in the field of biological fermentation. However, dynamic nonlinearity, unequal sampling intervals, uneven duration, and multiple features of a batch process make this task challenging. Thus, the multiple-feature fusion transformer (MFFT) model is proposed for the time series quality prediction of a batch process. First, the application of sequence-to-sequence architecture enables MFFT to perform a wide range of sequence prediction tasks. Second, the transformer parallel operation model imposes no rigid requirement for the order of sequence input, allowing the model to deal with problems of unequal interval sampling and utilize the sequence information. Third, MFFT integrates a pretrained ResNet50 as a mycelium status classifier for fusing image information into the features. Moreover, a multiple-feature encoding structure is proposed to integrate sampling time and mycelium status. Finally, multiple tasks in penicillin fermentation have shown that MFFT significantly outperforms existing methods for time series prediction.

A cybernetic model to predict the effect of freely available nitrogen substrate on rifamycin B production in complex media

It is well-known that secondary metabolite production is repressed by excess nitrogen substrate available in the fermentation media. Although the nitrogen catabolite repression has been known, quantitative process models have not been reported to represent this phenomenon in complex medium. In this paper, we present a cybernetic model for rifamycin B production via Amycolatopsis mediterranei S699 in complex medium, which is typically used in industry. Nitrogen substrate is assumed to be present in two forms in the medium; available nitrogen (S(ANS)) such as free amino acids and unavailable nitrogen (S(UNS)) such as peptides and proteins. The model assumes that an inducible enzyme catalyzes the conversion of S(UNS) to S(ANS). Although S(ANS) is required for growth and product formation, high concentrations were found to inhibit rifamycin production. To experimentally validate the model, five different organic nitrogen sources were used that differ in the ratio of S(ANS)/S (UNS). The model successfully predicts higher rifamycin B productivity for nitrogen sources that contain lower initial S(ANS). The higher productivity is attributed to the sustained availability of S(ANS) at low concentration via conversion of S(UNS) to S(ANS), thereby minimizing the effects of nitrogen catabolite repression on rifamycin production. The model can have applications in model-based optimization of substrate feeding recipe and in monitoring and control of fed batch processes.