AlphaFold as a prior: experimental structure determination conditioned on a pretrained neural network - PubMed
10 hours ago
- #ROCKET
- #protein structure determination
- #AlphaFold2
- ROCKET integrates experimental data (cryo-EM, cryo-ET, X-ray crystallography) into AlphaFold2 to refine predicted protein structures.
- It optimizes structures in the coevolutionary embedding space, capturing biological variations that AlphaFold2 alone cannot, especially in low signal-to-noise scenarios.
- This approach allows for scalable, automated model building without retraining, combining experimental observables with machine learning for better interpretability.