Modeling of Continuous PHA Production by a Hybrid Approach Based on First Principles and Machine Learning

Polyhydroxyalkanoates (PHA) are renewable alternatives to traditional oil-derived polymers. PHA can be produced by different microorganisms in continuous culture under specific media composition, which makes the production process both promising and challenging. In order achieve large productivities while maintaining high yield efficiency, needs operated so-called dual nutrient limitation condition, where nitrogen carbon sources kept at very low concentrations. Mathematical models greatly assist design operation of bioprocess, but challenged complexity system, particular nutrient-limited growth phenomenon, cells undergo a metabolic shift that abruptly changes their behavior. Traditional, non-structured mechanistic based on Monod uptake kinetics used describe bioreactor conditions. However, absence model description phenomena inside cell, extrapolation broader domain (e.g., feeding concentrations dilution rates) may present mismatches between predictions actual outcomes. Such detailed require almost perfect knowledge cell metabolism omic-level measurements, hampering development. On other hand, purely data-driven learn correlations from experimental data do not any prior therefore unbiased flexible. many more required for development ability is limited conditions similar ones training. An attractive alternative combination power first principles with flexibility machine learning methods. This approach results hybrid rates predict dynamic bioreactor. Here we develop Pseudomonas putida GPo1 culture. After training, gained medium compositions, demonstrate how wide range operating conditions, including single growth.