Rapid protein stability prediction using deep learning representations
Problem
Predicting how mutations change protein thermodynamic stability is needed in protein engineering and disease research, but biophysics-based methods are slow.
Why it matters
Protein stability underpins protein engineering and understanding of evolution and genetic disease.
What the AI did
RaSP leverages deep learning representations to predict stability changes on-par with biophysics methods, enabling saturation mutagenesis predictions in under a second per residue.
Results
On-par with biophysics-based methods; under one second per residue; ~230 million stability changes computed across the human proteome.
Impact
Enables proteome-scale stability analyses; tool freely available including a web interface.
Record
| Venue | eLife |
|---|---|
| Published | 2023 |
| Work type | article |
| Authors | Lasse M. Blaabjerg, Maher M. Kassem, Lydia L. Good, Nicolas Jonsson, Matteo Cagiada, Kristoffer E. Johansson, Wouter Boomsma, Amelie Stein et al. |
| Citations | 174 (OpenAlex, at import time) |
| References | 75 |
| Model or method | RaSP |
| Organization | not stated in source |
| Primary source | https://doi.org/10.7554/elife.82593 |
| Open access | gold |
| Keywords | protein stability, deep learning, saturation mutagenesis, human proteome, RaSP |
Importance rating
significant — thousands to millions affected
Abstract states proteome-wide computation (~230 million stability changes) linking stability to genetic disease and protein engineering; broad biomedical utility, freely available.