Protein tertiary structure prediction and refinement using deep learning and Rosetta in CASP14
Problem
Predicting and refining protein 3D tertiary structures from sequence remains difficult and computationally expensive.
Why it matters
Accurate protein structure prediction accelerates biomedical research and drug design.
What the AI did
Deep learning (trRosetta) predicts residue-residue distance and orientation distributions to build 3D models, refined via DeepAccNet-guided recombination of template-free and template-based models.
Results
Considerable improvement over original trRosetta; full modeling in median <3 h in CASP14
Impact
Improved pipeline was validated in blind CASP14 community assessment, an independent benchmark.
Record
| Venue | Proteins Structure Function and Bioinformatics |
|---|---|
| Published | 2021 |
| Work type | article |
| Authors | Ivan V. Anishchenko, Minkyung Baek, Hahnbeom Park, Naozumi Hiranuma, David E. Kim, Justas Dauparas, Sanaa Mansoor, Ian R. Humphreys et al. |
| Citations | 50 (OpenAlex, at import time) |
| References | 37 |
| Model or method | trRosetta with DeepAccNet |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1002/prot.26194 |
| Open access | hybrid |
| Keywords | protein structure prediction, deep learning, trRosetta, CASP14, DeepAccNet |
Importance rating
major — millions of lives, an industry or a field
Protein structure prediction is a foundational challenge in structural biology and drug design; abstract shows blind CASP14 validation and major speedup, but states no scale of beneficiaries.