SCIENCE UNDER AUTOMATION
Autonomous enzyme engineering platform
A biofoundry couples ML and language models to enzyme design and measurement over four rounds, with fewer than 500 variants per enzyme.
What this result supports
The input sequence and measurable fitness objective are supplied. Two enzyme demonstrations do not establish general organism-level experimental autonomy.
Evidence record: Peer-reviewed physical biofoundry experiments. Checked: 2026-10-21.
Where this evidence is used
Biological sciences
- Propose candidates for a measurable objective: Autonomous enzyme design starts from an input sequence and a defined fitness measurement.
- Construct and measure candidates: The integrated biofoundry executes iterative construction and characterization.
- Choose subsequent rounds from measured feedback: Four rounds of design and testing connect feedback to new candidate selection.
- Validate causality beyond the original assay: Researchers provide the physical validation; success in a configured assay does not resolve new biological contexts.