ProTEUS proposes a mutation-resolved model for protein evolution
Original title: ProTEUS: A mutation-resolved world model for closed-loop protein evolution
Authors report variants of aldehyde dehydrogenase and levansucrase with more than 10-fold higher activity than wild type.
ProTEUS is a mutation-resolved world model pretrained on deep mutational scanning assays. MutationFormer represents sequence, evolutionary, and structural information, and a physics-constrained all-atom predictor is described as more than 10 times faster than AlphaFold3 with guaranteed physical validity. Closed-loop experimental feedback is used in two multi-objective campaigns. The authors say it beats established models on zero-shot and supervised mutation prediction.
Unconfirmed: Activity gains and speed claims are the authors' experimental results, not independently verified here.
- Campaigns target aldehyde dehydrogenase and levansucrase.
- Authors claim more than 10-fold higher activity than wild type.
- The all-atom predictor is described as over 10 times faster than AlphaFold3.
- Pretraining uses deep mutational scanning assays.
- Products
- ProTEUSMutationFormer
- AI models
- AlphaFold3
From the original article
1 min read at huggingface.co
Protein engineering requires navigating rugged fitness landscapes, where small sequence changes can produce large functional differences, but experimental screening can only sample a small fraction of sequence space. Thus, a central challenge is to efficiently traverse the large combinatorial space of mutations in silico. We introduce ProTEUS, a mutation-resolved world model for closed-loop protein evolution.
By Yikun Zhang, Minghao Guo, Guangxue Xu · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.