Integration of dynamical systems and machine learning for cell factory design Machine learning has emerged as a promising paradigm for model-based design in synthetic biology, particularly for the optimization of cellular systems with applications in the textile, food and pharmaceutical sector. In this talk, I will describe some of our recent work on model-based design of metabolic pathways, including the use of discontinuous dynamical systems, Bayesian optimization, and neural networks. I will focus on the long-term dynamics of complex arrays of positive and negative feedback systems, nonlinear mixed-integer optimisation of metabolic pathways, as well as surrogate machine learning for integration of local models with metabolism at the genome-scale. These results showcase the many opportunities offered by the combination of data-driven and mechanistic modelling in synthetic biology. The talk will be based on these works: https://www.biorxiv.org/content/10.1101/2024.04.09.588720v1 https://doi.org/10.1016/j.automatica.2018.10.046 https://pubs.acs.org/doi/10.1021/acssynbio.3c0012 This article was published on 2025-04-22
