Machine Learning Media Recipes
Using computers to optimize media composition
University of California, Davis, United States
Who: Zachary Cosenza, Ph.D. student in chemical engineering
When: 2019 – 2022
Institutes: University of California, Davis, United States
Supervisors: David Block, professor of chemical engineering at the University of California, Davis; Keith Baar, professor of functional molecular biology at University of California, Davis
Zachary is using machine learning to predict the best combination of ingredients to create a media that optimizes muscle cell growth. The media used for growing muscle cells is made up of several ingredients, but some of those ingredients may be better for cell growth than others. Conventional techniques involve changing one ingredient at a time, but Zachary’s lab uses an artificial neural network to learn from experiments and predict the best combination of ingredients.
Zachary’s work will help accelerate breakthroughs in the field of cellular agriculture by using machine learning to speed up the research process. In the future, his program could be used to create formula best suited for any purpose a future researcher would hope to obtain.
Zachary identified and explored, both experimentally and computationally, major advantages that machine learning and Bayesian optimization methods have over traditional design-of-experiments methods in quickly designing cell culture media for rapid product deployment. A Bayesian multi-assay approach to optimizing media to increase cell growth and reduce cost was particularly effective, and resulted in significantly more productive and less costly serum and serum-free media for cellular agriculture applications.
To find out more about this project, listen to our podcast where we talk to Zachary about computational tools for design optimization and parameter estimation for design of cell growth media.Listen Now
Biotechnology and Bioengineering, 2022