Vireo Advisors Contributes to AI-derived Open-Source Growth Factor Thermostability Dataset
We are excited to share a new open-source dataset on growth factor thermostability for cultured meat and seafood safety research! The dataset was developed in collaboration with New Harvest, the Alberta Machine Intelligence Institute (Amii), and Defined Bioscience, a San Diego-based biotechnology company specializing in animal-free cell culture products, and is described in a manuscript submitted to Data in Brief. This work was conducted as part of the Cultured Meat Safety Initiative (CMSI), a joint initiative led by New Harvest and Vireo Advisors.
Growth factors are essential inputs in cultured meat production. Native forms of some growth factors occur naturally in foods, but the recombinant growth factors used in cultured meat production do not have specific approval for use in food and generally lack a history of safe use in food. A key safety question is whether any residual growth factors remaining in the final product would retain activity after processing and cooking. Most cultured meat products will be cooked before eating, and growth factors are sensitive to heat; they denature and lose biological activity at higher temperatures. Understanding the temperature at which different growth factors break down is therefore critical to demonstrating the safety of cultured meat products. Thermostability data are available for some commonly used growth factors, but these data have not been systematically complied and can be difficult to access. For others, data remain limited and new measurements are needed.
To address this gap, our team generated an open-source dataset combining three sources of thermostability data: experimental measurements of growth factors commonly used in cultured meat production using thermal shift assays; computational stability calculations using FoldX, a protein stability prediction tool; and over 500 melting temperature values curated from published literature using a custom AI-assisted extraction workflow. The combined dataset covers 32 unique GFs and their variants across 42 species, and is now publicly available on Zenodo.
A highlight of this project was the development of an LLM-based literature extraction workflow, which successfully extracted 91% of melting temperature values correctly from published papers. This demonstrates that AI tools can meaningfully accelerate safety data curation, with appropriate human validation.
This project was funded by Food System Innovations (FSI) and is part of New Harvest's AI and Machine Learning in Cellular Agriculture Initiative and the Cultured Meat Safety Initiative (CMSI). We thank FSI for supporting this important work, and our collaborators at Amii and Defined Bioscience for making it possible!