Improving lipid production by Rhodotorula glutinis for renewable fuel production based on machine learning

• Published in: Frontiers of chemical science and engineering 2024-05, Vol.18 (5), Article 51
• Main Authors: Zhang, Lihe, Zhang, Changwei, Zhao, Xi, He, Changliu, Zhang, Xu
• Format: Article
• Language: English
• Subjects: Artificial neural networks; Biomass; Biorefinery and biomanufacturing; Chemistry; Chemistry and Materials Science; Cognition & reasoning; Fermentation; Fuel production; Genetic algorithms; Industrial Chemistry/Chemical Engineering; Lipids; Machine learning; Microorganisms; Multiple objective analysis; Nanotechnology; Optimization; Research Article; Support vector machines
• ISSN: 2095-0179; 2095-0187
• DOI: 10.1007/s11705-024-2410-8

Microbial lipid fermentation encompasses intricate complex cell growth processes and heavily relies on expert experience for optimal production. Digital modeling of the fermentation process assists researchers in making intelligent decisions, employing logical reasoning and strategic planning to optimize lipid fermentation. It this study, the effects of medium components and concentrations on lipid fermentation were investigated, first. And then, leveraging the collated data, a variety of machine learning algorithms were used to model and optimize the lipid fermentation process. The models, based on artificial neural networks and support vector machines, achieved R 2 values all higher than 0.93, ensuring accurate predictions of the fermentation process. Multiple linear regression was used to evaluate the respective target parameter, which were affected by the medium components of lipid fermentation. Lastly, single and multi-objective optimization were conducted for lipid fermentation using the genetic algorithm. Experimental results demonstrated the maximum biomass of 50.3 g·L −1 and maximum lipid concentration of 14.1 g·L −1 with the error between the experimental and predicted values less than 5%. The results of the multi-objective optimization reveal the synergistic and competitive relationship between biomass, lipid concentration, and conversion rate, which lay a basis for in-depth optimization and amplification.