Precision fermentation
Engineered microbes brew specific proteins such as rennet, whey or egg and milk proteins that replace ingredients once taken from animals.
Open in the interactive tree →Chymosin made by engineered microbes got US approval in 1990 and now supplies over 90% of the world's rennet. Perfect Day's microbe-made beta-lactoglobulin (whey protein) received an FDA 'no questions' letter in 2020. Because no animals are involved, the same protein can be produced wherever fermenters stand.
As of now
As of 2024 no country had a legal definition or specific regulatory framework for precision fermentation; products go through existing novel-food or GRAS routes case by case. Further dairy proteins were cleared in the US and Israel (Remilk 2022, Imagindairy 2024). In September 2025 Strauss Group began selling CowFree milk and cream cheese made with cow beta-lactoglobulin grown in engineered fungi.
Open steps
- Codon design for secreted proteins High AI leverageExtend language models trained on codon usage to optimize secreted-protein expression, including signal peptides, in Pichia, Trichoderma and Saccharomyces.
- Gene targets for yield improvement Medium AI leverageMine fermentation logs and genotypes to identify rate-limiting genes (secretion helpers, stress response, overflow metabolism) as engineering targets.
- Predicting large-tank behavior Medium AI leverageUse learned heat-transfer and oxygen-diffusion models to predict cell growth in 100 m³+ tanks from 1 L lab data, avoiding failed large batches.
Where AI could help
Medium AI leverage. AI raises protein yields by design, but fermenter capacity and case-by-case food approvals set how fast products reach shelves.
- Optimise codons and signal sequences to raise secretion in yeast and fungi
- Predict strain-engineering steps that raise yield, such as secretion helper genes
- Design animal-free proteins that keep function but are easier to produce
- Model oxygen and heat transfer in large tanks to predict scale-up behaviour
Shown so far
- In February 2026, MIT researchers published Pichia-CLM (PNAS), a language model for codon design in industrial yeast that gave the best sequence for five of six test proteins against commercial tools. source