New research is advancing the fundamental science behind cellular agriculture, driving exciting advances in the future of food production.
The two new papers highlight the promise that machine learning has in the field of cellular agriculture, which involves producing livestock food products, like meat, by growing animal cells directly instead of from the whole animal. This research builds upon a long-term research collaboration between Amii and New Harvest, a non-profit dedicated to advancing cellular agriculture.
To date there has been limited collaboration and cross pollination between the Cell Ag and ML communities. When we started learning more about this field we realized lots of areas ripe for optimization with ML,” says Dave Staszak, Lead Machine Learning Scientist at Amii.
“By ensuring the data and models we develop are open, these projects begin to build a foundation for the entire field to utilize ML solutions.”
Bre Duffy
Director of Responsible Research & Innovation at New Harvest.
While still an emerging technology, cell-based agriculture has the potential to address some of the greatest challenges faced by the global food industry, including sustainability, food-borne diseases and ethical concerns around animal husbandry.
The new research uses machine learning techniques to build on the fundamental tools and approaches used in cellular agriculture. In keeping with New Harvest’s open-source mandate and Amii’s philosophy of “AI for good and for all”, both of the research papers are intended to offer useful new resources to the field of cellular agriculture, supporting further research into the area beyond just this collaboration.
“These projects address the two main barriers to realizing ML's potential in cellular agriculture: a lack of public data and researchers with expertise in both domains,” says Bre Duffy, Director of Responsible Research & Innovation at New Harvest.
“By ensuring the data and models we develop are open, these projects begin to build a foundation for the entire field to utilize ML solutions.”
Better understanding of growth mediums
The first paper, Growth factor thermostability dataset derived from experimental and computational approaches to accelerate cultured meat development, was published in June. It is co-authored in part by Kiana Aghakasiri (University of Alberta student working with Amii), Dave Staszak, and Breanna Duffy (New Harvest’s Director of Responsible Research & Innovation). The research was funded by Food System Innovations, a non-profit dedicated to investing in solutions for more sustainable food protein.
In this paper, the team used a combination of laboratory testing, computer modelling, and LLM-assisted research reviews to build a dataset containing information about the heat tolerance of the growth factors used to develop cultured meat. Growth factors are an essential part of cellular agriculture — they act as a signal that tells cells to grow and differentiate into different structures making up meat, like muscle and fat.
These growth factors can come from various plant- and animal-based sources, and different growth factors can break down at different cooking temperatures. Knowing how these growth factors react to different temperatures is critical for food safety and food regulation. By gathering this data from different sources and compiling it into this dataset, New Harvest and Amii researchers offer an important tool for the wider scientific community and help advance the future of food production.
More effective use of growth cultures
The second paper, published last month, is A data-driven evaluation of optimization techniques in cell culture media, which was co-authored by Staszak, Ali Parsaee (formerly a Machine Learning Resident at Amii) and Miranda Stahn (formerly New Harvest’s Program Manager, Western Canada). It was supported by funding from Schmidt Sciences, a philanthropic initiative founded by Eric and Wendy Schmidt.
This research focused on optimizing the culture media — the nutrients that the cells are grown in. Finding the best mix of nutrients and growth factors is an extremely difficult task, with billions of possible combinations to consider. It’s also one of the most costly parts of the process. The team’s research used a form of Bayesian optimization to develop a model that could predict which culture media would be more effective for a specific production. Their model was able to find optimal formulations of the culture media much faster than traditional statistical methods, sometimes in as few as half the trials. This kind of optimization could dramatically lower the production costs in cellular agriculture and speed up project timelines.
The two projects we focused on here are very practical, applied problems and will help solve two of the critical bottlenecks holding this technology back from larger adoptions: cost efficiency and safety,” Staszak says.
Both papers were part of the second phase of Amii’s research collaboration with New Harvest to explore how machine learning could advance the science of cellular agriculture. The first phase focused on a review of the current research on ML in the field, while the second phase brought on experts to apply ML solutions to the biggest challenges in cellular agriculture.