Hybrid Dynamic Models of Bioprocesses Based on Elementary Flux Modes and Multilayer Perceptrons

The derivation of minimal bioreaction models is primary importance to develop monitoring and control strategies cell/microorganism culture production. These can be obtained based on the selection a basis elementary flux modes (EFMs) using an algorithm starting from relatively large set EFMs progressively reducing their numbers geometric least-squares residual criteria. reaction rates associated with selected usually have complex features resulting combination different activation, inhibition saturation effects several species. Multilayer perceptrons (MLPs) are used in order undertake representation these rates, hybrid dynamic model combining mass-balance equations provided by rate described MLPs. To further reduce number kinetic parameters model, pruning algorithms for MLPs also considered. whole procedure ends up reduced-order macroscopic that show promising prediction results, as illustrated data perfusion cultures hybridoma cell line HB-58.