Protein Design Using Structure-Prediction Networks: AlphaFold and RoseTTAFold as Protein Structure Foundation Models
Problem
Designing proteins with tailored structures and functions is a long-standing unsolved goal in bioengineering.
Why it matters
Custom-designed proteins could enable new binders, enzymes, and therapeutics.
What the AI did
Review covers methods that repurpose structure-prediction neural networks (AlphaFold, RoseTTAFold) for protein design via activation maximization, inpainting, and denoising diffusion.
Results
not stated in source (major improvements in wet-lab success rates claimed, no numbers given)
Impact
Reports major wet-lab success-rate improvements for binders, metalloproteins, enzymes, and assemblies.
Record
| Venue | Cold Spring Harbor Perspectives in Biology |
|---|---|
| Published | 2024 |
| Work type | preprint |
| Authors | Jue Wang, Joseph L. Watson, Sidney Lyayuga Lisanza |
| Citations | 34 (OpenAlex, at import time) |
| References | 118 |
| Model or method | AlphaFold, RoseTTAFold |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1101/cshperspect.a041472 |
| Open access | bronze |
| Keywords | protein design, AlphaFold, RoseTTAFold, deep learning, structure prediction |
Importance rating
major — millions of lives, an industry or a field
Abstract explicitly frames tailored protein design as a long-standing bioengineering goal and reports major wet-lab success-rate improvements for binders, enzymes, assemblies; foundational field with therapeutic reach.