EUBCE 2026 - Nichakorn FUNGPRASERTKUL - Steps Toward A Digital Twin For Optimizing Biorefinery Process and Developing Culture Mode for Productive Lipid Production from an Oleaginous Yeast

Steps Toward A Digital Twin For Optimizing Biorefinery Process and Developing Culture Mode for Productive Lipid Production from an Oleaginous Yeast

Short Introductive summary

Bayesian regression with Gaussian processes (GP) was used to upgrade culture mode from batch to fed-batch or continuous by estimating a change of lipid production by Metchnikowia pulcherrima to achieve the high yield. Active learning was be used to construct gaussian process regression in the small datasets. In this study, GP was generated from the common oleaginous yeast database (Yarrowia lipolytica) to optimize the lipid production of the under-explored oleaginous yeast (M. pulcherrima). The GP was progressively developed from the experimental dataset by considering the acquisition function in order to find to optimal culture mode setting (i.e., substrate addition time and substrate concentration) to get the maximum lipid production (g/l) in the limited fermentation time. In addition, the responses of biomass (g/l) and lipid content (%) was investigated for the deeper understand in the oleaginous yeast metabolism. In addition, Rapeseed Meal (RSM) which is an industrial waste was used as a feedstock. It was biologically hydrolyzed to amino acids by solid-state fermentation of Aspergillus oryzae.

N. Fungprasertkul, The University of Manchester, UNITED KINGDOM
J. Winterburn, The University of Manchester, UNITED KINGDOM
P. Martin, The University of Manchester, UNITED KINGDOM
H. Moss, University of Cambridge, UNITED KINGDOM