A generalized platform for artificial intelligence-powered autonomous enzyme engineering
Problem
Engineering proteins with desired functions is slow, expensive, and dependent on specialist expertise.
Why it matters
Engineered enzymes have numerous applications in energy, health, and sustainability.
What the AI did
A platform integrating machine learning and large language models with biofoundry automation autonomously engineered enzymes, requiring only a protein sequence and a fitness measurement.
Results
90-fold improvement in AtHMT substrate preference, 16-fold improvement in ethyltransferase activity, and 26-fold improvement in YmPhytase activity at neutral pH, in four rounds
Impact
Demonstrated autonomous engineering of multiple enzymes with large measured activity improvements.
Record
| Venue | Nature Communications |
|---|---|
| Published | 2025 |
| Work type | article |
| Authors | Nilmani Singh, Stephan Lane, Tianhao Yu, Jingxia Lu, Adrianna Ramos, Haiyang Cui, Huimin Zhao |
| Citations | 111 (OpenAlex, at import time) |
| References | 57 |
| Model or method | machine learning and large language models (specific models not stated in source) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1038/s41467-025-61209-y |
| Open access | gold |
| Keywords | enzyme engineering, machine learning, large language models, biofoundry, autonomous platform, protein engineering |
Importance rating
major — millions of lives, an industry or a field
Abstract states proteins have numerous applications in energy, health, sustainability, and the platform paves way for advances across medicine, biotech, renewable energy, sustainable chemistry.