Therapeutic enzyme engineering using a generative neural network
Problem
Finding enzyme variants that improve mRNA therapeutic potency is extremely slow because the possible sequence space is too vast to search experimentally or accurately computationally.
Why it matters
More potent enzyme therapeutics could allow lower, less frequent dosing for patients with rare metabolic diseases.
What the AI did
A generative neural network (deep latent variable model of sequence co-evolution) combined with automated protein library design to rapidly identify improved enzyme variants, applied to ornithine transcarbamylase.
Results
Abstract claims variants that are both more thermally stable and more catalytically active; specific numbers not stated in source
Impact
Peer-reviewed demonstration of AI-guided engineering of a therapeutic enzyme for a rare urea-cycle disease.
Record
| Venue | Scientific Reports |
|---|---|
| Published | 2022 |
| Work type | article |
| Authors | Andrew J. Giessel, Athanasios D. Dousis, Kanchana R. Ravichandran, Kevin Christopher Smith, Sreyoshi Sur, Iain J. McFadyen, Wei Zheng, Stuart L. Licht |
| Citations | 40 (OpenAlex, at import time) |
| References | 65 |
| Model or method | generative neural network / deep latent variable model |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1038/s41598-022-05195-x |
| Open access | gold |
| Keywords | enzyme engineering, generative neural network, mRNA therapeutics, protein design, ornithine transcarbamylase |
Importance rating
real but narrower advance
Abstract states enhancing mRNA therapeutic potency may enable lower, less-frequent dosing and addresses a rare but serious metabolic disease; meaningful therapeutic benefit but rare-disease scope is narrow.